scieee Science in your language
[en] (orig)

Einsatz von digitalen Entscheidungshilfen für das Management von heterogenen Pflanzenbeständen im ökologischen Landbau

Reuter, Tobias · 2026

Abstract

The global agricultural sector faces the dual challenge of ensuring food security for a growing population while contending with a steady decline in biodiversity. Although organic farming offers significant environmental benefits—in terms of improved soil health and increased biodiversity—its yields are, on average, 20 % lower than those achieved with conventional practices. Precision Farming (PF) offers a promising solution to mitigate this yield gap by optimizing input applications (e.g., fertilizers) based on spatial variability within fields, which most fields exhibit. This dissertation presents possible solutions for enhancing the productivity and sustainability of organic farming. PF technologies, such as Unmanned aerial vehicles (UAVs) and Decision Support Systems (DSS), were integrated with traditional organic management practices including mechanical weeding, crop rotation and intercropping. To test this approaches a series of trials were conducted in northwestern Germany, reported in three peer reviewed papers. Site-specific mechanical weeding reduced the treated area by 58 % to 83 % in maize (depending on the year). Weeds were detected based on image recognition of UAV multispectral images. The Relative Weed Cover was a successful decision support, as it helped to limit the weeding areas by considering both weed and maize cover. UAVs were also effectively employed for delineating of three clover–grass fields into distinct management zones. The used NDRE (Normalized Difference Red Edge)-Maps and fuzzy C-means clustering algorithms were appropriated for this task. These zones exhibited significant differences in clover proportion, biomass and the yield of subsequent cereal crops—driven primarily by variations in soil properties and topography. Such delineation supports the tailored management of sub-fields for optimal fertilization and crop selection. In addition, intercropping trials demonstrated that mixing malting barley with peas (Pisum sativum) enhanced protein content and land-use efficiency, whereas intercropping with linseed (Linum usitatissimum) resulted in reduced protein levels; nevertheless, sole stands of barley generally performed comparably to mixed cropping systems. Intercropping can therefore be used to influence the quality by partner crop to mitigate the spatial heterogeneity within the field. Collectively, these approaches advance sustainable agriculture by reducing unnecessary weed control and enabling more efficient resource use. This thesis contributes to the few research projects that combines PF-technology with organic farming practice. The results add to more efficient resource use and sustainable agriculture. The continued development of field robots and advanced DSS holds the potential to further modernize agriculture toward a biodiversity-based system. Future practices may involve subdividing fields into smaller management units that are integrated with landscape elements to promote biodiversity. However, additional long-term, user-focused research is necessary to fully understand the complex interactions within environmentally friendly agricultural landscapes.

Read accessible full text

Einsatz von digitalen Entscheidungshilfen für das Management von heterogenen Pflanzenbeständen im ökologischen Landbau

Author: Reuter, Tobias
Year: 2026
DOI: 10.48693/901
Source: https://osnadocs.ub.uni-osnabrueck.de/bitstream/ds-2026033114799/2/thesis_reuter.pdf
Tobias Reu e
Use o Digi al Decision Suppo Tools o Managing
He e ogeneous Fields in O ganic Fa ming
Einsa z on digi alen En scheidungshil en ü das
Managemen on he e ogenen P lanzenbes änden im
ökologischen Landbau
PhD Thesis
Osnab ück Uni e si y
2025
Use o Digi al Decision Suppo Tools o Managing
He e ogeneous Fields in O ganic Fa ming
Einsa z on digi alen En scheidungshil en ü das
Managemen on he e ogenen P lanzenbes änden im
ökologischen Landbau
Disse a ion
zu E langung des Dok o g ades Dok o de Na u wissenscha en (D . e . na .)
des Fachbe eichs Kul u - und Sozialwissenscha en de Uni e si ä
Osnab ück
in Koope a ion mi de Hochschule Osnab ück
Fakul ä Ag a wissenscha en und Landscha sa chi ek u
o geleg on
Tobias Reu e
Aus Bad Godesbe g
am 05.09.2025, Osnab ück
Index
i
Index
Index ........................................................................................................................................................ i
Abs ac .................................................................................................................................................. ii
Abb e ia ions ........................................................................................................................................ iii
Figu e index .......................................................................................................................................... i
Table index .............................................................................................................................................
Chap e 1 Gene al in oduc ion ....................................................................................................... 6
1.1 Backg ound and objec i e .......................................................................................................... 7
1.2 S uc u e o he hesis ................................................................................................................. 8
1.3 S udy a ea ................................................................................................................................ 10
1.4 O ganic ag icul u e in Ge many and Eu opean Union ............................................................. 11
1.5 He e ogeneous ield condi ions ................................................................................................ 13
1.6 P ecision Fa ming and decision suppo .................................................................................. 16
1.6.1 Da a sensing and analysis........................................................................................................ 17
1.6.2 Decision making ....................................................................................................................... 20
1.6.3 Resou ce applica ions .............................................................................................................. 22
1.6.4 C op mapping and e alua ion .................................................................................................. 23
1.7 Resea ch ques ions and hypo heses ....................................................................................... 24
Chap e 2 Scien i ic publica ions wi hin he con ex o his wo k.............................................. 26
2.1 Si e-speci ic mechanical weed managemen in maize (Zea mays) in No h-Wes Ge many .. 27
2.2 Delinea ion o managemen zones in clo e -g ass o si e-speci ic managemen o subsequen
c ops .................................................................................................................................................. 45
2.3 E ec s o mixed in e c opping on he ag onomic pa ame e s o wo o ganically g own mal ing
ba ley cul i a s (Ho deum ulga e) in No hwes Ge many .................................................................. 66
Chap e 3 Gene al discussion ........................................................................................................ 67
3.1 In eg a ion o P ecision Fa ming and o ganic ag icul u e: po en ial and challenges ............... 68
3.2 Va ia ion o ag icul u al pa ame e s be ween yea s and si es ................................................. 71
3.3 Ou look: Syne gies be ween p oduc ion and ecology .............................................................. 73
Chap e 4 Conclusions ................................................................................................................... 77
Summa y .............................................................................................................................................. 78
Zusammen assung (Ge man summa y) ............................................................................................ 80
Re e ences ........................................................................................................................................... 82
Acknowledgemen s ........................................................................................................................... 103
Appendix ............................................................................................................................................ 105
Lis o publica ions ............................................................................................................................... 105
Abs ac
ii
Abs ac
The global ag icul u al sec o aces he dual challenge o ensu ing ood secu i y o a
g owing popula ion while con ending wi h a s eady decline in biodi e si y. Al hough o ganic
a ming o e s signi ican en i onmen al bene i s—in e ms o imp o ed soil heal h and
inc eased biodi e si y—i s yields a e, on a e age, 20 % lowe han hose achie ed wi h
con en ional p ac ices. P ecision Fa ming (PF) o e s a p omising solu ion o mi iga e his yield
gap by op imizing inpu applica ions (e.g., e ilize s) based on spa ial a iabili y wi hin ields,
which mos ields exhibi .
This disse a ion p esen s possible solu ions o enhancing he p oduc i i y and
sus ainabili y o o ganic a ming. PF echnologies, such as Unmanned ae ial ehicles (UAVs)
and Decision Suppo Sys ems (DSS), we e in eg a ed wi h adi ional o ganic managemen
p ac ices including mechanical weeding, c op o a ion and in e c opping. To es his
app oaches a se ies o ials we e conduc ed in no hwes e n Ge many, epo ed in h ee pee -
e iewed pape s. Si e-speci ic mechanical weeding educed he ea ed a ea by 58 % o 83 %
in maize (depending on he yea ). Weeds we e de ec ed based on image ecogni ion o UAV-
mul ispec al images. The Rela i e Weed Co e was a success ul decision suppo , as i
helped o limi he weeding a eas by conside ing bo h weed and maize co e . UAVs we e also
e ec i ely employed o delinea ing o h ee clo e –g ass ields in o dis inc managemen
zones. The used NDRE (No malized Di e ence Red Edge)-Maps and uzzy C-means
clus e ing algo i hms we e app op ia ed o his ask. These zones exhibi ed signi ican
di e ences in clo e p opo ion, biomass and he yield o subsequen ce eal c ops—d i en
p ima ily by a ia ions in soil p ope ies and opog aphy. Such delinea ion suppo s he ailo ed
managemen o sub- ields o op imal e iliza ion and c op selec ion. In addi ion, in e c opping
ials demons a ed ha mixing mal ing ba ley wi h peas (Pisum sa i um) enhanced p o ein
con en and land-use e iciency, whe eas in e c opping wi h linseed (Linum usi a issimum)
esul ed in educed p o ein le els; ne e heless, sole s ands o ba ley gene ally pe o med
compa ably o mixed c opping sys ems. In e c opping can he e o e be used o in luence he
quali y by pa ne c op o mi iga e he spa ial he e ogenei y wi hin he ield. Collec i ely, hese
app oaches ad ance sus ainable ag icul u e by educing unnecessa y weed con ol and
enabling mo e e icien esou ce use.
This hesis con ibu es o he ew esea ch p ojec s ha combines PF- echnology wi h
o ganic a ming p ac ice. The esul s add o mo e e icien esou ce use and sus ainable
ag icul u e. The con inued de elopmen o ield obo s and ad anced DSS holds he po en ial
o u he mode nize ag icul u e owa d a biodi e si y-based sys em. Fu u e p ac ices may
in ol e subdi iding ields in o smalle managemen uni s ha a e in eg a ed wi h landscape
elemen s o p omo e biodi e si y. Howe e , addi ional long- e m, use - ocused esea ch is
necessa y o ully unde s and he complex in e ac ions wi hin en i onmen ally iendly
ag icul u al landscapes.
Abb e ia ions
iii
Abb e ia ions
AI: A i icial In elligence
AMSL: Abo e mean sea-le el
ANN: A i icial Neu al Ne wo k
BBCH: Biological Fede al Ins i u e
o Ag icul u e and Fo es y
( ega ding plan g ow h
s ages)
CCCI: Canopy Chlo ophyll Con en
Index
CGZ: Clo e -g ass zones
CNN: Con olu ional Neu al
Ne wo k
Con: Con ol ea men
CU: Ce eal uni s
DA: Digi al Ag icul u e
DAS: Days a e sowing
DSS: Decision Suppo Sys ems
EU: Eu opean Union
GNDVI: G een No malized
Di e ence Vege a ion Index
GNSS: Global Na iga ion Sa elli e
Sys em
GS: G ow h s ages
LCCI: Lea Chlo ophyll Con en
Index
LER: Land Equi alen Ra io
LiDAR: Ligh De ec ion and
Ranging
LLM: La ge Language Models
MCARI: Modi ied Chlo ophyll
Abso p ion Ra io Index
Mg: Megag am
ML: Machine Lea ning
MZ: Managemen zone
N: Ni ogen
NDRE: No malized Di e ence Red
Edge
NDVI: No malized Di e ence
Vege a ion Index
NUE: Nu ien use e iciency
OA: O e all accu acy
PA: P ecision Ag icul u e
PF: P ecision Fa ming
PSRI: Plan Senescence
Re lec ance Index
RF: Random o es
RGB: Red, G een, Blue
RWC: Rela i e weed co e
RYT: Rela i e yield o al
SMN: Soil mine al ni ogen
SNC: Soil ni ogen con en
SOC: Soil o ganic ca bon
SOM: Soil o ganic ma e
SSWM: Si e-speci ic weed
managemen
TKW: Thousand ke nel weigh
UAV: Unmanned ae ial ehicles
VI: Vege a ion Index
VRA: Va iable a e applica ion
VSWC: Volume ic soil wa e
con en
WC: Weed co e
XAI: eXplainable A i icial
In elligence

Figu e index
i
Figu e index
Fig 1. O e iew o h ee pee - e iewed pape s on empo al and spa ial scales. ................................... 9
Fig 2. Geog aphical dis ibu ion o expe imen al si es .......................................................................... 10
Fig 3. To al (poin , solid line) and sha ed ( iangle, dashed) a ea unde o ganic managemen ........... 13
Fig 4. O e iew o sou ces o spa ial he e ogenei y. ............................................................................ 14
Fig 5. Wo k low o P ecision Fa ming wi h examples in each s ep. ...................................................... 17
Fig 6. (Sub)plo a angemen and placemen o measu ing a eas. ...................................................... 31
Fig 7. Mean ai empe a u e [°C] (lines) and sum o p ecipi a ion [mm] (ba s) ..................................... 31
Fig 8. Linea eg ession be ween olume ic soil wa e con en [%] and weed co e [%]. ................... 37
Fig 9. Weed co e [%] es ima ed by image ecogni ion be o e i s si e-speci ic weed managemen .. 38
Fig 10. Compa ison o ea ed a ea [m²] due o second hoeing be ween ea men ............................ 39
Fig 11. Compa ison o maize d y ma e yield [g/m-2] a op and weed biomass [g m-2] a bo om ...... 39
Fig 12. Geog aphical Dis ibu ion o Expe imen al Si es ...................................................................... 48
Fig 13. Te ain based on digi al ele a ion model o he ial ields. ....................................................... 49
Fig 14. Mean ai empe a u e [°C] (lines) and sum o p ecipi a ion [mm] (ba s) ................................... 51
Fig 15. NDRE-Maps (no malized di e ence ed edge) o clo e -g ass o he di e en ields and
selec ed da es. ...................................................................................................................................... 54
Fig 16. Maps o he delinea ed o clo e -g ass zones (CGZ) o each ield. ....................................... 55
Fig 17. Yield o clo e , g ass, and weeds a e compa ed be ween he wo clo e -g ass zones ........... 55
Fig 18. Yield p opo ions o clo e , g ass, and weeds a e compa ed be ween he wo clo e -g ass
zones .................................................................................................................................................. 56
Fig 19. Ce eal plan heigh [cm] in compa ison be ween he wo clo e -g ass zones (CGZ) ............... 57
Fig 20. Ce eal ke nel yield [g m-2] in compa ison be ween he wo clo e -g ass zones (CGZ) a
ha es . .................................................................................................................................................. 57
Fig 21. De elopmen o olume ic soil wa e con en in compa ison be ween wo clo e -g ass zones ..
................................................................................................................................................ 58
Fig S 1. Indexes o inding bes numbe o clus e s ............................................................................. 65
Fig S 2. Maps o he delinea ed o clo e -g ass zones (CGZ) o he single obse ed da es ............. 65
Pic u e 1: Sampling o win e whea (T i icum aes i um) ial, 17 h May 2023. ..................................... 6
Pic u e 2: Summe spel (T i icum aes i um subsp. Spel a) ial, 7 h July 2021 .................................. 26
Pic u e 3: Summe spel (T i icum aes i um subsp. Spel a) ial, 19 h July 2022 ................................ 67
Table index
Table index
Table 1: Soil nu ien con en . ............................................................................................................... 30
Table 2: De ailed ield his o y and ial managemen o all expe imen al seasons. ............................ 32
Table 3: Pa ame e s o he weed ecogni ion o he di e en sampling da es. ................................... 33
Table 4: Compa ison o olume ic wa e con en ................................................................................ 35
Table 5: Compa ison o maize plan heigh [cm], weed co e [%] and numbe o weed species ........ 36
Table 6: Accu acies o UAV-based image classi ica ion o weeds, maize and soil. ............................ 37
Table 7: Field desc ip ion. ..................................................................................................................... 48
Table 8: Field managemen and soil sampling da es ........................................................................... 49
Table 13: Tes ed hypo hesises in he h ee pee - e iewed pape . ....................................................... 69
Table S 1: Sampling da es ield da a and UAC campaigns. ................................................................. 63
Gene al in oduc ion
6
Chap e 1 Gene al in oduc ion
Pic u e 1: Sampling o win e whea (T i icum aes i um) ial, 17 h May 2023.
Gene al in oduc ion
7
1.1 Backg ound and objec i e
The global ag icul u al sec o aces he challenge o eeding an an icipa ed popula ion o
nea ly 10.000.000.000 people by 2050 (Wille e al., 2019), while simul aneously con ending
wi h dec easing ag icul u al land pe capi a (FAO, 2024). This si ua ion is compounded by
signi ican declines in bo h plan and animal wildli e popula ions (Geige e al., 2010; Wille e
al., 2019). In ensi e use o e ilize s and pes icides has led o a p omo ion o plan and animal
species (F ancksen e al., 2022; Milla d e al., 2021), which a e o en mo e ha m ul, han a
di e se communi y (Esposi o e al., 2023). Addi ionally, homogeneous ag icul u al
landscapes—bo h in e ms o c op o a ion and landscape elemen s— ail o suppo di e se
wildli e, as each species has speci ic equi emen s such as nu ien supply, illage ime and
hos s (Meye e al., 2019; Tscha n ke e al., 2021). Di e se landscapes con ibu e o a ied
wildli e. This has bo h empo al and spa ial scales. The empo al scale e e s o a ying and
ex ended elemen s o c op o a ion (Da is e al., 2012). The spa ial scale includes landscape
elemen s ( lowe s ips, ees and hedge ows), habi a connec i i y, c op mix u es, smalle ield
sizes and highe ield edge densi y (Raa z e al., 2019).
The concep o plane a y bounda ies summa ies he global challenge o a s able ea h
sys em. I delinea es c i ical pa ame e s essen ial o main aining ea h's sys em s abili y and
esilience. Ou o hese nine p ocesses, six cu en ly exceed hei bounda ies. Ag icul u e
signi ican ly impac s many o hese p ocesses, pa icula ly biogeochemical lows, no el en i ies
like pes icides, land-sys em change due o ag icul u al managemen and biosphe e in eg i y
(Richa dson e al., 2023).
O ganic ag icul u e holds he po en ial o educe he ag icul u al impac . I aims o p oduce
ood in an en i onmen ally iendly manne , p omo ing closed nu ien cycles and minimizing
he use o pes icides and e ilize s. This p ac ice leads o a e age o 30 % highe species
ichness, especially plan s (Tuck e al., 2014) and pollina o s like bees p o i om o ganic
a ming (Walke e al., 2024). Howe e , his app oach o en esul s in yields ha a e 17 % o
20 % lowe han con en ional a ming p ac ices (de la C uz e al., 2023; De Pon i e al., 2012),
due o limi a ion o N e ilise (Dö ing and Neuho , 2021). The p ohibi ion o chemical
pes icides necessi a es he use o mechanical weeding, which is less e icien in weed con ol
and can lead o soil e osion and plan inju ies (Machleb e al., 2020; Melande e al., 2015).
Ag icul u al ields, bo h con en ional o o ganic managed, o en exhibi a a ie y o
condi ions, including di e ences in soil ex u e, o ganic ma e con en , opog aphy, nu ien
a ailabili y and wa e accessibili y (Aksakal e al., 2019; Zhu e al., 2013). These a ia ions,
in luenced by ield his o y (Schulp and Ve bu g, 2009) and landscape elemen s, such as
lowe ing s ips, hedge ows o ees (Raa z e al., 2019), esul in di e se plan g ow h pa e ns
ha impac bo h p oduc i i y and wildli e (Pä zold e al., 2020). P ecision ag icul u e add esses
hese issues by ailo ing managemen p ac ices o he speci ic condi ions o sub- ields (Eli-
Chukwu, 2019). Mode n ag icul u al p ac ices o en u ilize senso s o moni o plan and soil
condi ions, wi h he collec ed da a being analysed o p o ide managemen ecommenda ions
h ough digi al Decision Suppo Sys ems (DSS; Balasund am e al., 2023). These sys ems
le e age ag icul u al knowledge alongside s a is ical models o A i icial In elligence o op imize
decision-making p ocesses (Eli-Chukwu, 2019).
Gene al in oduc ion
14
Fig 4. O e iew o sou ces o spa ial he e ogenei y. The g ey a ea ep esen s pa en ma e ial, he b own a ea
ep esen s soil, and he g een a ea ep esen s he su ace. Soil ex u e (a) and soil o ganic ma e (SOM) (b)
in luence wa e and nu ien e en ion. Soil dep h (c) limi s oo ing space and a ailabili y o wa e and nu ien s.
Topog aphy (d) a ec s wa e low and exposu e o sunligh ; ypically, he lowe slope is mo e e ile due o he
accumula ion o ine ex u ed soil and SOM. Land managemen p ac ices (e) impac soil e ili y and nu ien le els
due o illage and e ilisa ion. Gene ally, o ganic a ming and no- ill managemen inc ease SOM and ea hwo m
abundance. Pas land use ( ) a ec s yield and SOM con en in he cu en pe iod. Landscape elemen s like ees
and hedge ows compe e wi h c ops o ligh , nu ien s, and wa e . Landscape elemen s (g), such as ield bo de s,
es ic e ilisa ion and pes icide use. Weeds (h) occu in he e ogeneous pa ches and compe e wi h c ops o
nu ien s, wa e , and ligh . Some pes species (i) a e dis ibu ing om an in ec ion poin and building pa ches.
Soil condi ion
Soil ex u e is one o he mos impo an soil pa ame e s because i signi ican ly a ec s soil
chemical p ope ies and c op yield (Bölenius e al., 2017; F anzluebbe s e al., 2025). In
pa icula , he clay con en o soil in luences he e en ion o wa e and nu ien s due o i s
pa icle s uc u e and nega i e cha ge (Boenecke e al., 2018; Gi z and Ma ila, 2024;
Habib‑u ‑Rahman e al., 2022). Thus, a highe clay con en can p omo e c op yields— o
example, in g ass leys (da Sil a e al., 2022), as well as in ice (O yza sa i a), co n (Zea mays),
co on (Gossypium hi su um) (Ouazaa e al., 2022) and win e whea (T i icum aes i um; G oß
e al., 2023; Usowicz and Lipiec, 2017). Mo eo e , because biological ni ogen ixa ion is
co ela ed wi h he biomass p oduc ion o legumes (Hoeks a e al., 2015), soil ex u e indi ec ly
in luences ni ogen con en . I should be no ed ha c ops espond di e en ly o soil ex u e; o
ins ance, soybeans (Glycine max) a e signi ican ly mo e a ec ed by soil ex u e han pas u e
c ops (Oldoni e al., 2025). In addi ion, SOM has a s ong posi i e e ec on soil e ili y and
c op yield because i supplies nu ien s and can e ain wa e and ni ogen (G oß e al., 2023).

Gene al in oduc ion
15
Soil dep h cons ains he oo ing space o plan s and hus in luences he a ailabili y o
nu ien s and wa e . In a eas wi h deepe soils, mo e esou ces a e a ailable, which is
gene ally associa ed wi h highe yields (Odone e al., 2024; Pe al a e al., 2015; Sad as and
Cal iño, 2001). No ably, he esponse o soil dep h is s onge in maize (Zea mays) han in
whea (T i icum aes i um) (Sad as and Cal iño, 2001).
Many ields ea u e a dis inc elie . In many cases, he lowe pa o a slope is mo e
p oduc i e han he uppe slope (Bölenius e al., 2017; Pe al a e al., 2015; Rod iguez Mi anda
e al., 2021) due o he accumula ion o SOM and a ine soil ex u e ha esul s in deepe soils
downslope, he eby enhancing soil e ili y. Addi ionally, downslope a eas o en exhibi highe
soil wa e con en because wa e na u ally lows o lowe ele a ions (Pe al a e al., 2015;
Rod iguez Mi anda e al., 2021). In con as , Habib‑u ‑Rahman e al. (2022) epo ed ha yields
a he op o he slope we e 53 – 88 % highe han hose downslope—whe e ele a ions we e
2 o 5 m lowe . This disc epancy was explained by he accumula ion o SOM in a small
dep ession wi hin a con ex slope. Mo eo e , opog aphy in luences exposu e o sunligh
(Pe al a e al., 2015).
Managemen in luence
Human ac i i ies, including his o ical land use, also in luence soil p ope ies and e ili y.
Fo example, he egula applica ion o e ilize s is a common p ac ice (F anzluebbe s e al.,
2025). Con e sely, he use o hea y machine y can compac he soil, hus limi ing oo ing
dep h and educing i s capaci y o s o e wa e —nega i ely a ec ing yields (Gi z and Ma ila,
2024). The a oidance o ploughing (also known as no- ill managemen ) can imp o e soil bulk
densi y and s abili y (F anzluebbe s e al., 2025), as well as inc ease SOM by app oxima ely
21 %. Bu he e ec on SOM a ies be ween c ops, wi h no e ec in maize (Zea Maize) o
s ong inc ease in ba ley (Ho deum ulga e) (Bai e al., 2018). O he s udy showed a signi ican
highe SOM con en only in he op soil laye (0 – 15 cm), a deepe soil laye s, no e ec
occu ed (Jakab e al., 2023; Wulanning yas e al., 2021). Such p ac ices addi ionally s imula e
soil biological ac i i y, wi h ea hwo m abundance inc easing by up o 50 % when pes icides
a e no applied (Bai e al., 2018).
O ganic a ming is conside ed o ha e a posi i e e ec on soil e ili y as men ioned in
sec ion 1.4. Pa o his bene i is a ibu able o he use o o ganic e ilize s, which can
inc ease SOM by 21 – 47 % (Bai e al., 2018). A mo e di e se c op o a ion has been shown
o inc ease ea hwo m abundance by 60 % and SOM by 20 % (Bai e al., 2018). Mo eo e ,
pe manen g asslands ypically exhibi highe soil agg ega e s abili y, enhanced biological
ac i i y and lowe soil bulk densi y compa ed o a able lands (F anzluebbe s e al., 2025). The
e ec s on SOM and soil ni ogen can pe sis o decades; o example, changes in land use o
a able land can s ill be de ec ed 50 yea s la e and con inue o in luence yields (Schus e e
al., 2024). Simila ly, land use changes om o es and plaggen ag icul u e in he Ne he lands
ha e p oduced measu able e ec s e en a e mo e han 100 yea s (Schulp and Ve bu g,
2009).
Field bo de s ha e also e ol ed his o ically as ields we e spli o combined. Land-use
changes— o example, con e ing pe manen g assland o a able land—we e some imes
implemen ed a he sub- ield le el (Schus e e al., 2024). Consequen ly, his o ical land-use
decisions and c op managemen p ac ices con ibu e o inc eased spa ial a iabili y.
All ields a e cha ac e ized by ce ain bo de ing elemen s, including ield bo de s,
ag icul u al oads, neighbou ing ields, ees and hedge ows, all o which a ec c op
de elopmen on si e (Raa z e al., 2019). Gene ally, bo de a eas yield less (wi hin 6 – 7 m o
Gene al in oduc ion
16
he edge) compa ed o mid- ield a eas due o es ic ions in pes icide and e ilize applica ions
(Raa z e al., 2019). Na u al landscape elemen s such as ees and hedge ows can lead o
yield educ ions o 17.5 % and 7.9 %, espec i ely, o e anges o app oxima ely 18 m—e ec s
ha a e a ibu ed o shading and compe i ion o wa e and nu ien s (Jose e al., 2000; Raa z
e al., 2019). I should be no ed ha ees can also posi i ely a ec yields ou side he immedia e
shading zone (beyond 20 m; Raa z e al., 2019). Beyond hei in luence on ligh , ees and
hedge ows unc ion as wind ba ie s and help educe e apo anspi a ion (Jacobs e al., 2022).
These e ec s a y om yea o yea , in d y yea s, shading and educed e apo anspi a ion
can be bene icial, whe eas in we yea s he opposi e may occu (Raa z e al., 2019). This is
especially impo an unde u u e condi ions o clima e change wi h inc easing hea . T ees can
u he p o ide p o ec ion agains c op damage o ex eme p ecipi a ion (Jacobs e al., 2022).
The he e ogeneous condi ions in luence no only c ops bu also weeds (Usowicz and
Lipiec, 2017), esul ing in a pa chy dis ibu ion o wild plan s (Me cal e e al., 2019). Each
species has i s own speci ic en i onmen al equi emen s (Pä zold e al., 2020) and he
a iabili y in condi ions in luences plan compe i i eness. This u he con ibu es o he
he e ogeneous dis ibu ion o bo h c ops and weeds (Mhlanga e al., 2016). Many pa hogens
like Fusa ium g aminea um (Fusa ium head bligh ) o Puccinia s ii o mis . sp. i ici (Ps )
(Yellow us ) a e sp eading om ini ial in ec ion poin . Cen e a ound his in ec ion poin , nes s
a e building, leading o a pa chy dis ibu ion (Gao e al., 2023; A. Guo e al., 2021).
1.6 P ecision Fa ming and decision suppo
P ecision Fa ming (PF), also known as P ecision Ag icul u e, in ol es using senso s o
gain insigh s in o c op de elopmen , nu ien supply and g owing condi ions. Resou ces a e
hen applied acco ding o hese condi ions (Mo an e al., 1997). The in e p e a ion o his
in o ma ion can be ei he manual o au oma ic, leading o managemen decisions ha adap
o he spa ial and empo al he e ogenei y wi hin a ield o landscape. This app oach aims o
use esou ces mo e e icien ly (Balasund am e al., 2023; FAO, 2022; Ka una hilake e al.,
2023).
The amewo k o PF is based on h ee key ag icul u al s eps: (1) diagnosis, (2) decision
making and (3) pe o ming (FAO, 2022). The wo k low includes: (i) da a sensing, (ii) da a
analysis, (iii) decision making, (i ) esou ce applica ion, ( ) c op mapping and ( i) e alua ion
(Balasund am e al., 2023). Fig 5 illus a es he main wo k low and p o ides examples. PF
encompasses a ious opics, including c op a ming, ag onomics, li es ock, aquacul u e and
ag o o es y (Ka una hilake e al., 2023).
Sma a ming, digi al ag icul u e o Ag icul u e 4.0 is he u he de elopmen o PF and
some imes used synonymously (Leddin e al., 2023). This concep p o ides a digi al
“ecosys em”, wi h he aim o connec and analyse da a ac oss he whole ag icul u al p oduc ion
chain. Digi al ag icul u e combines a wide a ie y o senso s, au onomous sensing and
applica ion, big da a analysis (AI and DSS) wi h In e ne o hings (IoT: a ne wo k o
in e connec ed i ems and echnologies; Basso and An le, 2020; Goumagias e al., 2021).
Unmanned ae ial ehicle (UAV) and Unmanned g ound ehicles (UGV), such as ield obo s,
a e commonly used in his concep (Balasund am e al., 2023).
This s udy ocuses on he managemen o he e ogeneous c op ields using digi al DSS. In
his con ex , he e m "P ecision Fa ming" (PF) pe ains speci ically o c op a ming, excluding
o he po en ial a eas. The subsequen sec ions will p o ide a de ailed examina ion o da a
sensing and analysis, decision making and esou ce applica ion.
Gene al in oduc ion
17
Fig 5. Wo k low o P ecision Fa ming wi h examples in each s ep. UAV = unmanned ae ial ehicle.
1.6.1 Da a sensing and analysis
The sensing and analysis o da a ega ding c op de elopmen and soil condi ions o m he
basis o PF applica ions. A wide ange o ools and da a sou ces is a ailable—each wi h a
dis inc scope and applica ion. In he ollowing a selec ion o s a e o he a me hods a e
desc ibed.
Senso s a e commonly employed, many based on op ical p inciples measu ing
e lec ance. Fo example, chlo ophyll me e s and g een seeke s assess he chlo ophyll con en
o lea es, which se es as an indica o o plan nu ien le els. These senso s a e a ailable as
handheld de ices o can be moun ed on ac o s (Sha ma and Bali, 2018).
Remo e sensing in ol es acqui ing in o ma ion abou an objec wi hou physical con ac
(Fussell e al., 1986). Many emo e sensing applica ions deploy senso s on sa elli es o UAVs,
each wi h i s own bene i s. Sa elli es a e o en low-cos o e en ee and p o ide global
co e age (Vidican e al., 2023). In con as , UAVs can yield images wi h pixel esolu ions unde
1 cm (Nah s ed e al., 2024), compa ed o sa elli es such as Plane Scope (3 m), Sen inel-2
(10 m), o Landsa (30 m). Addi ionally, UAVs o e g ea e lexibili y—especially unde cloudy
condi ions (Ba e h e al., 2019). In esea ch, sa elli e image y is he mos common, accoun ing
o app oxima ely 65 % o s udies, ollowed by ae ial ehicles a 22 % (Vidican e al., 2023).
Gene al in oduc ion
18
Vege a ion Indices
Many emo e sensing app oaches u ilize Vege a ion Indices (VIs), which a e spec al
ans o ma ions combining wo o mo e bands o iden i y ege a ion and quan i y biomass. The
mos commonly used index is he No malized Di e ence Vege a ion Index (NDVI) because o
i s sui abili y o a wide ange o applica ions. This index co ela es well wi h g een biomass
(B eunig e al., 2020; Munna e al., 2022; Sapko a e al., 2024; Vidican e al., 2023). Howe e ,
NDVI is known o o e sa u a e a highe biomass alues (Ka una a ne e al., 2020; Xu e al.,
2020). In addi ion, indices such as G een No malized Di e ence Vege a ion Index (GNDVI),
NDVI, Modi ied Chlo ophyll Abso p ion Ra io Index (MCARI) and Plan Senescence
Re lec ance Index (PSRI) can se e a b oad ange o objec i es, while o he s like he Canopy
Chlo ophyll Con en Index (CCCI) and Lea Chlo ophyll Con en Index (LCCI) a e mo e speci ic
o applica ions such as c op classi ica ion and chlo ophyll es ima ion (Vidican e al., 2023).
VIs ha e been employed o a wide ange o applica ions. They ha e been used o
dis inguish among c ops—such as beans (Phaseolus ulga is), bee oo (Be a ulga is), g ass
(Poaceae), maize (Zea mays), po a o (Solanum ube osum) and whea (T i icum aes i um)—
a a landscape le el (Sonobe e al., 2018) and o de ec s ess (Skendži e al., 2023). Because
nu ien con en and o he quali y pa ame e s in luence he spec al esponse, VIs can be used
o assess hese aspec s, as demons a ed by Zeng and Chen (2018) o o ages and by Ba zin
e al. (2020) o maize.
Yield es ima ion and p edic ion ep esen a b oad esea ch a ea, ha ing been applied o
many c ops including maize (Zea mays; Ba zin e al., 2020), whea (T i icum aes i um; Cheng
e al., 2022), po a o (Solanum ube osum), suga bee (Be a ulga is) (Vannoppen and Gobin,
2022) and o ages (Zeng and Chen, 2018). Combining VIs wi h o he measu emen s (e.g.,
plan heigh ) can imp o e biomass es ima ion in g asslands (Viljanen e al., 2018). RGB indices
ha e also been success ully used o es ima ing o age biomass (Lussem e al., 2019).
Al hough RGB came as a e mo e a o dable, be e esul s a e ypically achie ed when
combining RGB da a wi h nea -in a ed (NIR) image y, gi en s ong co ela ion o NIR wi h
chlo ophyll con en (Viljanen e al., 2018).
Despi e hei b oad u ili y, VIs ace se e al challenges, including: (1) a iabili y in c op
g ow h s ages (e.g., lowe ing) due o di e ences in lea s uc u e; (2) o e lapping lea es; (3)
a iabili y wi hin and be ween ields; (4) cloud co e in e e ence and (5) simila spec al
esponses ac oss di e en plan species. These ac o s can complica e he in e p e a ion o VI
da a (Vidican e al., 2023).
Image Analysis and Machine Lea ning
Image analysis me hods ha use he o iginal spec al and ex u al in o ma ion can help
mi iga e some o he challenges associa ed wi h VI ans o ma ions, which may lead o
in o ma ion loss. While many analyses ha e adi ionally elied on s a is ical models, machine
lea ning (ML) app oaches—including Random Fo es (RF), Suppo Vec o Machines (SVM)
and A i icial Neu al Ne wo ks (ANN)—a e gaining p ominence. These me hods can handle
complex, mul idimensional da a by independen ly co ela ing, weigh ing and ex apola ing
mul iple ea u es (Ge ha ds e al., 2020; Júnio e al., 2020; Lee e al., 2017).
Fo ins ance, Azimi e al. (2021) compa ed Con olu ional Neu al Ne wo ks (CNN) wi h
classical Machine lea ning me hods and achie ed highe accu acy when measu ing s ess
le els due o nu ien de iciency. Machine lea ning me hods such as Suppo Vec o Machines
and Random Fo es ha e also been used o es ima e whea yield, wi h R² alues a ound 0.9
Gene al in oduc ion
19
(Cheng e al., 2022). In g ass swa d s udies, Random Fo es was used o success ully es ima e
d y ma e yield and ni ogen up ake, achie ing R² alues o 0.85 and 0.89, espec i ely
(Oli ei a e al., 2020). Viljanen e al. (2018) achie ed an R² o 0.97 o d y ma e es ima ion
using a combina ion o swa d heigh , RGB-spec al alues and VIs, while Random Fo es has
also been deployed o iden i y c op p opo ions in g asslands using UAV image y (Nah s ed
e al., 2024).
CNNs p o ide accu acy ad an ages o e classical ML app oaches o weed de ec ion,
al hough hey ypically equi e longe aining imes (Júnio e al., 2020). Fo example, weed
ecogni ion in maize achie ed an accu acy o 88 % using a CNN applied o high- esolu ion
RGB images cap u ed nea g ound le el (Hasan e al., 2024). Compa able esul s ha e been
ob ained du ing he seedling s age (V1) o maize wi h UAV-cap u ed RGB images, despi e
lowe pixel esolu ion (Pei e al., 2022). In g assland s udies, UAV mul ispec al images and
CNN models ha e been used o map wild plan species, achie ing an o e all accu acy o 88 %
(Pö ke e al., 2023).
Mul ispec al UAV image y has also been used o iden i y Ce cospo a Lea Spo (caused
by Ce cospo a be icola) in suga bee (Be a ulga is), whe e pa ial leas squa es disc iminan
analysis achie ed an o e all accu acy o 86 % (Ba e o e al., 2023). Fo de ec ing Fusa ium
head bligh (caused by Fusa ium g aminea um) in whea (T i icum aes i um), a combina ion
o VIs and ex u e indices achie ed 93 % accu acy wi h UAV-moun ed mul ispec al images
(Gao e al., 2023). A simila app oach was used o he de ec ion o yellow us (caused by
Puccinia s ii o mis . sp. i ici) in whea (T i icum aes i um), whe e hype spec al UAV images
yielded R² alues anging om 0.55 o 0.88 depending on he in ec ion s age, wi h lowe
accu acy obse ed du ing ea ly in ec ions (A. Guo e al., 2021).
O e iew o soil analysis me hods
Soil o ganic ca bon has been es ima ed wi h R² alues o 0.89 using hype spec al images
and 0.69 using ime se ies o mul ispec al images. These esul s, achie ed ia a Neu al
Ne wo k model (Ex eme Lea ning Machine), ou pe o med hose ob ained om pa ial leas
squa es eg ession (Guo e al., 2021). T adi ional soil sampling emains a aluable me hod o
p oducing soil maps and o en unde pins emo e sensing app oaches (Ouazaa e al., 2022);
howe e , i is limi ed by high cos s (Buladaco II e al., 2024) and low spa ial esolu ion (Ouazaa
e al., 2022).
An al e na i e is he use o p oximal senso s, which ope a e ei he on an op ical o an
elec ochemical p inciple. An example o an elec ochemical senso is he de ice de eloped
by Smolka e al. (2017). This measu emen echnique is based on capilla y elec opho esis, in
which ions in a liquid sample a e sepa a ed by an elec ic ield, allowing hem o pass
sequen ially pas a de ec o . E alua ions demons a e a s ong co ela ion o ni a e (NO₃-)
and po assium (K+) measu emen s; howe e , o ammonium (NH₄+) and phospha e (PO43-),
he concen a ions we e oo low o be eliably de ec ed (Smolka e al., 2017).
Fa mlab (S enon, Ge many), a handheld de ice ha u ilizes impedance measu emen s
and spec al analysis ( anging om ul a iole o NIR) o es ima e a ious ni ogen o ms in he
soil. While e ec i e a de ec ing ield he e ogenei y, his de ice ends o o e es ima e mine al
ni ogen (Nmin) by 38 kg N ha−1 in 75 % o cases compa ed o labo a o y analyses (Vikuk e al.,
2024). Addi ionally, elec ical conduc i i y me e s (e.g., he EM38 by Geonics L d.,
Mississauga, Canada) measu e soil elec ical conduc i i y ia elec omagne ic induc ion,
p o iding a good es ima e o soil e ili y since he alues co ela e s ongly wi h soil wa e
con en and clay con en (Cimpoiaşua e al., 2020; Fo es e al., 2015).

Gene al in oduc ion
20
Soil wa e con en se es as an excellen indica o o soil he e ogenei y and p o ides
insigh in o he wa e supply a ailable in a ield. Time Domain Re lec ome y (TDR) is
commonly used o hese measu emen s (Engels e al., 2025). One key ad an age o TDR is
i s high accu acy— ypically wi hin 1 – 2 % o olume ic con en —along wi h minimal
calib a ion equi emen s. Addi ionally, TDR o e s excellen spa ial and empo al esolu ion,
wi h op ions o au oma ion and con inuous moni o ing. The de ices a e easy o use, as and
non-des uc i e (Skie ucha e al., 2012).
Soil compac ion is a c ucial ac o a ec ing oo g ow h, as well as soil wa e and ai
a ailabili y. Pene ome e can measu e soil pene abili y o pene a ion esis ance on a ou ine
basis and a e used o es ima e soil compac ion (He ick and Jones, 2002).
His o ical Yield Da a
His o ical yield da a om combined ha es e s can o e aluable insigh s in o ield
a iabili y, as yield o en co ela es wi h soil e ili y. Key o his analysis a e he global
na iga ion sa elli e sys em (GNSS) coo dina es, which p ecisely loca e yield da a and mul i-
yea eco ds ha help mi iga e he e ec s o annual wea he a iabili y (Spe anza e al., 2023).
1.6.2 Decision making
A e da a a e collec ed and analysed, hey ha e o be in e p e ed, leading o a
managemen decision. The da a om yield maps o o he echnology men ioned abo e can
al eady gi e aluable insigh in o he spa ial a iabili y and help a me s make in o med
decisions abou he op imal iming o ha es (Ghazal e al., 2024). In gene al, decisions a e
ca ego ized in o h ee ca ego ies: (1) s a egic decisions a e on a high-le el and complex like
business decisions; (2) ac ical decision o managemen decision a e done weekly o mon hly,
like e ilize use; and (3) ope a ional decision, his is on a daily basis like wo ke dis ibu ion
(Leddin e al., 2023). Fa me s ha e o do se e al decisions each day and hey canno be an
expe o e e y opic (Saikai e al., 2020). Managemen decisions a e complex due o he need
o b oad knowledge and in e ac ion in he ag icul u al ield. Fu he mo e hey a e cha ac e ized
by unce ain y and isk due o esponse o c ops o wea he di e ences wi hin and be ween
seasons. An addi ionally unce ain y is he p ice de elopmen o esou ces like uel, e ilize
and p oduc p ices (Leddin e al., 2023). Decision Suppo Sys ems (DSS) and A i icial
In elligence (AI) can help wi h decision making due o he a ailabili y o handling highly complex
and dynamic condi ions in ag icul u e (Ka una hilake e al., 2023). The challenge o DSS is o
ep esen he eali y wi h unique condi ions a each a m (Saikai e al., 2020).
Me hods o Decision Suppo Sys ems
Va ious me hods o DSS a e employed, including bo h s a is ical and AI-based
app oaches (Jha e al., 2019). S a is ical modelling in ol es he de elopmen o ma hema ical
models o ep esen ela ionships be ween a iables using da a. These models a e u ilized o
comp ehend pa e ns, make p edic ions and es hypo heses. Examples include eg ession
analysis, k-means clus e ing and classi ica ion (Bocks alle and Gi a din, 2003). Fuzzy logic,
on he o he hand, is a compu a ional heo y based on "g adual u hs," as opposed o he
con en ional Boolean logic ha elies solely on ue o alse (1 o 0) cons uc s, o ming he
ounda ional s uc u e o all compu a ional ope a ions (Mamdani, 1974). Machine lea ning is
desc ibed as he scien i ic ield ha enables machines o lea n wi hou explici p og amming
(Samuel, 2000). I is o en used o compu e ision bu also o decision making (Jha e al.,
2019; Saikai e al., 2020). A i icial Neu al Ne wo ks (ANN) a e also sel -lea ning algo i hms.
Gene al in oduc ion
21
The a chi ec u e o ANN ypically includes h ee o mo e laye s: 1. Inpu laye 2. A leas one
hidden (middle) laye 3. Ou pu laye . This is oughly inspi ed by he connec i i y o he human
b ain. They a e especially bene icial in he p edic ion o highly complex and dynamic se ings
(Song and He, 2005). Neu o- uzzy logic is he combina ion o ANN wi h uzzy logic (Wie man
and Dob ansky, 1993), he ad an age is he implemen a ion o human expe ience and
knowledge (Papageo giou e al., 2011). Expe sys ems a e one he oldes app oaches o AI.
This sys em has wo componen s: (1) he knowledge base con ains he symbolic knowledge
o he expe in he o m o ules and heu is ics. (2) An In e ence engine ha sol es p oblems
by in e p e ing he ules o he knowledge base. An expe sys em can be lexible, adjus ed
wi h new expe knowledge. As i is based on human knowledge, i is easie o explain
(McKinion and Lemmon, 1985), especially compa ed o ANNs.
Selec ed DSS applica ions in ag icul u e
DSS we e de eloped o a wide ange o goals and applica ion and ha e di e en le els
o complexi y. The DSS “F uch olge” (Ge man o c op o a ion) was de eloped o c op
ecommenda ion and coa se manu e applica ion in Ge many. The sys em uses cus ome
e e ence numbe o ield geome y, p e ious c ops and soil da a. Based on a c op model,
egional yield and c op p ocess, i sugges s wha c op o plan and how much manu e o apply.
Fo his an in ege linea p og amming model is used. The case s udy a m shows a s ong
inc ease in p o i s (Pahmeye e al., 2021).
The In eg a ed Fa m Sys em Model (IFSM) is a esea ch model o he dai y o age a m
sys em. Unlike mos a m models, IFSM simula es all majo a m componen s on a p ocess
le el and can he e o e simula e economic pe o mance. I simula es he g ow h o di e en
c ops including al al a (Medicago sa i a), g ass, maize (Zea mays), soybean (Glycine max)
and small g ain c ops based on wea he da a. Nu ien supply, illage and ha es ing a e
conside ed. Based on a ailable eeds and he nu ien equi emen s, he animal esponses a e
compu ed wi h a cos -minimizing linea p og amming app oach, o use he eed in an op imal
way. Ano he componen is he nu ien low due o manu e p oduc ion, biological N- ixa ion
and losses (deni i ica ion and leaching). This model is o emos o esea ch pu poses bu can
help wi h decision suppo (Ro z e al., 2018).
A model, de eloped o suppo a me s decision o ha es in I eland is ”Pas u eBase”.
This calcula es he yield o each paddock based on a g ow h model using pas o co e ( isual
assessmen o pla eme e ) and he use o paddock (g azed o silage). The yield was p edic ed
wi h R² o 0.84 o g azed paddocks. The DSS achie ed a R² o only 0.58 o silage used
paddock due o lowe numbe o obse a ions. None heless, his helps he a me o ind he
igh ha es ime, as he yield is an impo an p edic o (Han ahan e al., 2017).
Ano he expe sys em o pas u e ha es managemen was de eloped by Reu e e al.
(2023). I in eg a es wea he o ecas s, pas u e heigh , g ow h s age, clo e p opo ion and
c ude ib e con en (de i ed om a g ow h model) o de e mine he op imal ha es da e and
o m—ei he hay o silage. A ha es da e is ecommended only i a d y pe iod o a leas i e
days ( o silage) o se en days ( o hay) is o ecas ed. Valida ion o his ule se on 26 ields
p oduced an o e all R² o 0.75, while in ensi ely managed ields o silage achie ing an R² o
0.9.
A Machine lea ning-model was es ablished o si e-speci ic applica ions in gene al wi h he
app oach ha each a me can cons uc a unique model o he own a m. The inpu s can be
d awn om his o ical da a. The algo i hm u ilizes Bayesian op imiza ion and can handle a wide
ange o managemen and en i onmen al a iables, adap ing o un o eseen in e ac ion e ec s.
Gene al in oduc ion
22
This DSS was es in simula ion and yielded posi i e esul s o lea nabili y o complex si e-
speci ic managemen and he highe p o i abili y compa ed o uni o m managemen (Saikai e
al., 2020).
Si e-speci ic e ilisa ion is o en based on yield po en ial maps, soil p ope ies, o sa elli e-
and UAV-images, as desc ibed in sec ion 1.6.1. Mos commonly, he e ilize is dis ibu ed o
e en ou di e ences o imp o e he yield in speci ic a eas (Hagn e al., 2025; Mo a i e al.,
2018). A DSS, based on soil da a, was de eloped o he dis ibu ion o manu e wi h he aim
o inc ease he s able pool o SOM (Co i e al., 2023). The nu ien use e iciency and yield
depend on he a ying wea he condi ions du ing he g owing season. One app oach uses
mid- e m wea he o ecas s combined wi h c op o ecas o plan he si e- and season-speci ic
N-managemen , leading o less used N- e ilize wi hou economic losses (Co i e al., 2023;
Palka and Manschadi, 2024). Co i e al. (2023) showed ha he e ec o a iable a e manu e
applica ion a ied be ween he yea s, bu o e all can imp o e he ni ogen use e iciency
(NUE).
Ano he complex decision opic is he managemen o weeds, as iming can be c ucial, as
well as he igh he bicide choice. Much esea ch ocus on he si e-speci ic applica ion o
educe he bicides o weeding. This is o en based on a simple weed h eshold sys ems ei he
o whole weed co e (Nikolić e al., 2021) o species speci ic h eshold (Hamouz e al., 2014).
Some s udies ollow e en a ze o- ole ance s a egy (Allmendinge e al., 2024; Spae h e al.,
2024). Th esholds can ei he base on expe knowledge o expe imen a ion and es ima ion o
yield losses by weeds (Longchamps e al., 2014). The weed co e o numbe is mos ly
es ima ed wi h UAV o ac o moun ed came as as desc ibed in sec ion 1.6.1.
Some app oaches a e mo e complex. Fo he ha owing in ensi y a non-linea model was
de eloped. The op imum angle o he weed ha ow ines was p edic ed based on h ee
a iables: (1) he p e-ha ow measu ed weed co e , (2) he d a o ce measu ed on he ines
and (3) he a ge weed con ol e icacy equi ed o lowe he weed damage o below he
biological h eshold (Be ge e al., 2024). The igh iming o image acquisi ion o si e-speci ic
weeding can be es ima ed by a weed me gence p edic ion model wi h soil empe a u e and
p ecipi a ion as inpu s (Nikolić e al., 2021). Niemeye e al. (2024) de eloped a easoning
sys em based on an expe sys em. I ca ego izes weeds in o he classes “ha m ul” and “no
ha m ul”, he classi ica ion depends on he dis ance be ween c op-plan and weed-plan and
ela i e ge mina ions ime. Weeds classi ied as “non ha m ul” can be spa ed o ole a ed in
highe co e age han “ha m ul” species. In each g id cell he numbe o each species is
coun ed and compa ed wi h he speci ic h eshold. Zingsheim and Dö ing (2024) sugges ha
weeding obo s should ha e he abili y o measu e weed co e and iden i y indi idual weed
species o op imize c op yield and biodi e si y.
1.6.3 Resou ce applica ions
PF is he p ac ice o applying esou ces based on he speci ic condi ions wi hin each ield,
esul ing in mo e e icien esou ce managemen . These esou ces include soil, space, e ilize
and wa e (Ghazal e al., 2024). Adjus ing esou ce applica ions o accoun o ield a iabili y
is known as si e-speci ic managemen o a iable a e applica ion (VRA). An impo an
echnology in his con ex is he global na iga ion sa elli e sys em (GNSS), which enables
p ecise loca ion mapping o a ge ed esou ce deli e y (Ka una hilake e al., 2023).
One widely used p ac ice is si e-speci ic nu ien managemen , o en implemen ed as
a iable a e e iliza ion. By applying e ilize s in acco dance wi h c op needs and soil nu ien -
supplying capaci y, ni ogen use e iciency (NUE) can be signi ican ly inc eased, while
Gene al in oduc ion
23
educing nu ien losses and g eenhouse gas emissions (Balasund am e al., 2023). Fo
ins ance, esea ch in Ge many demons a ed ha senso -based VRA could educe ni ogen
applica ion by 38 kg ha-1, esul ing in a 15 % highe NUE (Hagn e al., 2025). Howe e , he
e ec i eness o VRA is no consis en ac oss all si es and yea s. T ials in Aus ia ha e shown
ha he posi i e e ec s o si e-speci ic ni ogen applica ion— ega ding yield, p o ein con en
and economic p o i — a y by loca ion (Palka and Manschadi, 2024). Simila ly, si e-speci ic
manu e applica ion esul ed in highe NUE, hough ou comes luc ua ed om yea o yea
(Co i e al., 2023).
Si e-speci ic weeding educes he use o chemicals, he eby enhancing en i onmen al
sus ainabili y and educing he bicide s ess (Ghazal e al., 2024; Spae h e al., 2024). When
mechanical weeding is applied, i can also mi iga e nega i e impac s such as plan inju ies and
soil deg ada ion (Sei z e al., 2019; Woźniak, 2020). The educ ion in ea ed a ea can a y
be ween 10 % and 90 %, wi h many s udies epo ing sa ings be ween 40 % and 60 %
(Allmendinge e al., 2024; Cas aldi e al., 2017; Nikolić e al., 2021; Peña e al., 2013). These
a ia ions depend on ac o s such as wea he condi ions (López-G anados e al., 2016), weed
in es a ion le els (Spae h e al., 2024) and he iming and h esholds applied (Nikolić e al.,
2021) and g id size (López-G anados e al., 2016).
Li e al. (2022) es ed an in- ield, eal- ime sp aying sys em ha used indus ial-g ade RGB
came as o de ec ield his le (Ci sium a ense a . in eg i olium) plan s and immedia ely apply
he bicide. The sys em’s de ec ion accu acy and ela i e hi a e we e a ec ed by d i ing speed;
inc easing he speed om 2 km ha-1 d opped he accu acy om 91 % o 80 % and he hi a e
om 95 % o 93 %. Economic bene i s om educed he bicide applica ion ha e been
es ima ed o ange om 16 € ha-1 o 33 € ha-1 (Nikolić e al., 2021) and up o 150 € ha-1 (Rajmis
e al., 2022).
As no ed in sec ion 1.6.1, pes de ec ion also lays he ounda ion o si e-speci ic
managemen . Fo example, po en ial ungicide educ ions o a ound 18 % ha e been obse ed
(Rajmis e al., 2022). Ano he ial conduc ed in g assland achie ed ungicide educ ions
be ween 51 % and 65 % (Boo h e al., 2021).
Seeding a es can be adjus ed o local ield condi ions, as demons a ed wi h maize using
NDVI and soil maps. Si e-speci ic seeding esul ed in 3 % o 7 % highe yield compa ed wi h
uni o m seeding, he eby imp o ing he g oss ma gin by 3 % o 9 % (Munna e al., 2022). Fo
po a oes, yields inc eased o 17 Mg ha-1 wi h a iable a e seeding compa ed o 14 Mg ha-1
wi h uni o m seeding. Seeding maps based on ei he elec ical conduc i i y o NDVI each
esul ed in highe g oss ma gins han uni o m seeding (Munna e al., 2020a).
Ag icul u e is one o he la ges use s o eshwa e ; howe e , wa e -use e iciency emains
low. P ecision i iga ion helps educe wa e use by p e en ing o e wa e ing and wa e logging
(Neupane and Guo, 2019). Fo si e-speci ic i iga ion, c op models can p edic wa e
equi emen s a di e en g ow h s ages. Using his app oach, yields o co on and yeg ass
inc eased by 4.9 % and 8.5 %, espec i ely, while wa e applica ion was educed by
app oxima ely 5.5 % (Mcca hy e al., 2023).
1.6.4 C op mapping and e alua ion
C op mapping and he success o PF can be e alua ed using he echnology desc ibed in
sec ion 1.6.1, such as yield maps om combine ha es e s (Hagn e al., 2025). Ano he
app oach in ol es es ima ing c op biomass h ough emo e sensing da a ob ained om
sa elli es and UAVs, as demons a ed o a ious c ops including whea (T i icum aes i um;
Cheng e al., 2022), po a o (Solanum ube osum) and suga bee (Be a ulga is; Vannoppen
Scien i ic publica ions wi hin he con ex o his wo k
30
Table 1: Soil nu ien con en . Mg and pH- alue es ima ed wi h CaCL2-Me hod. P2O5 and K wi h Calcium-Ace a -
Lac a -Me hod.
Yea
P2O5
[mg/100 g soil]
K2O
[mg/100 g soil]
Mg
[mg/100 g soil]
NO3 + NH4
[kg ha-1]
pH-
alue
Co g [%]
2021
21.8
17.4
4.0
30.0
5.7
1.4
The p edomi al weed species in bo h yea s was Chenopodium album. S ella ia media, Poa
i ialis (L.), Polygonum con ol ulus (L.) and Galinsoga cilia a (RAF.) we e common as well.
The species Ama an hus, Equise um a ense (L.), Spe gula a ensis (L.), Capsella bu sa-
pas o is (L.), Ve onica ag es is (L.), Echinochloa c us-galli (L.), Lamium pu pu eum (L.)
occu ed only a some places and no a all imes.
Two di e en weed con ol h esholds we e compa ed wi h a uni o m weed managemen
as con ol ea men (Con). This ea men is equi alen o con en ional weeding, whe e e e y
subplo is ea ed du ing all applica ions ega dless o he weed p essu e.
The second ea men was a decision-making sys em based on a Weed Co e Th eshold
(WC) wi h he h esholds o 0.25 %, 0.5 % and 1.0 % weed co e . I he h eshold o a subplo
was exceeded, weed managemen was conduc ed. A weed co e o 0.06 o 0.31 % is he
economic h eshold o he bicide applica ion in maize (Longchamps e al., 2014). As
mechanical weeding is mo e cos ly han he bicide applica ion, a h eshold o 0.5 % was
chosen, complemen ed by a mo e conse a i e h eshold o 0.25 % and a less conse a i e
h eshold o 1.0 % weed co e .
The hi d ea men conside ed o decision-making he Rela i e Weed Co e (RWC,
Ngouajio e al. (1999). The RWC is a dimensionless alue and was calcula ed as ollows:
𝑅𝑊𝐶 = 𝑊𝑒𝑒𝑑 𝑐𝑜𝑣𝑒𝑟 (%)
𝑊𝑒𝑒𝑑 𝑐𝑜𝑣𝑒𝑟 (%)+𝐶𝑟𝑜𝑝 𝑐𝑜𝑣𝑒𝑟 (%) Equa ion 1
The ea men RWC had h ee h esholds: 0.1, 0.2 and 0.4 RWC. I he h eshold o a
subplo was exceeded, weed managemen was conduc ed. A RWC o 0.2 is equal o yield
losses o a ound 10 % in compa ison wi h a weed ee plo (Ali e al., 2015; Ngouajio e al.,
1999). In o de o be able o e alua e he e ec s o di e en ole ance h esholds, a mo e
conse a i e (RWC = 0.1) and a mo e ole an (RWC = 0.4) ea men we e also es ed he e.
The expe imen was s uc u ed as a andomized block design. The se en ea men s we e
applied ac oss ou eplica es, esul ing in a o al o 28 plo s. Each plo measu ed 30 m in
leng h and 3 m in wid h. To mi iga e he in luence o p io si e-speci ic weeding, he ial’s
loca ion wi hin he s udy si e was al e ed in he subsequen yea , and he plo s we e e-
andomized.
Fo si e-speci ic managemen , each plo was subdi ided in o h ee subplo s, each
ex ending 10 m in leng h and main aining he wid h o 3 m. Wi hin each subplo , a 1 m²
measu ing a ea was designa ed o he assessmen o plan ai s. This a ea was posi ioned
a he cen e, spanning he wo cen al ows o he subplo . Each measu ing a ea including wo
0.1 m2 a eas, one co e s he in a- ow space, he o he he in e - ow space as shown in Fig 6.
Thei loca ion was pe manen o each ial yea .

Scien i ic publica ions wi hin he con ex o his wo k
31
Fig 6. (Sub)plo a angemen and placemen o measu ing a eas.
Selec ion o each measu emen a ea was me iculously conduc ed o ensu e
ep esen a ion o he en i e plan communi y wi hin each subplo . This app oach yielded a o al
o 168 samples on each designa ed measu emen da e.
Wea he condi ions
The wea he da a was eco ded by a wea he s a ion o he Hochschule Osnab ueck
Uni e si y o Applied Science a 60 m AMSL abou 1 km away om he ield. Ai empe a u e
was measu ed a a heigh o 2 m. Fig 7 shows he weekly wea he condi ions o 2021 and
2022.
Fig 7. Mean ai empe a u e [°C] (lines) and sum o p ecipi a ion [mm] (ba s) o each week o ial pe iod (Ap il –
Oc obe ) in compa ison o he long- e m pe iod o 1996 – 2022 in g ey in he ial yea s 2021 and 2022. Ve ical
do ed lines indica e selec ed managemen .
Wi h an annual ain amoun o 646 mm in 2021 and 631 mm in 2022, he p ecipi a ion was
much lowe han he a e age o 872 mm o he pe iod o 1996 o 2022. In he pe iod o Ap il
o Oc obe , he yea 2021 shows mo e ain han 2022 wi h 385 mm in compa ison o 324 mm.
In he i s ial yea , 56.8 mm mo e p ecipi a ion occu ed in he sowing pe iod (week 14 o
Scien i ic publica ions wi hin he con ex o his wo k
32
22). The mean empe a u e o 2021 is compa able o he long- e m pe iod a e age. Howe e ,
wi h a empe a u e o 11 °C, 2021 was wa me han 2022 and he long- e m a e age o 10 °C.
T ial managemen
To es he si e-speci ic managemen unde condi ions o high weed p essu e, maize was
cul i a ed as p e ious c op. Be o e he p e-c op, i e yea s o Phacelia anace i olia was g own.
Maize, a ie y “Se e een” (FAO class 230), was seeded wi h he sowing machine P acea
(Amazone) and eal ime kinema ic au os ee ing sys em. This a ie y is usable bo h o o age
and g ain ha es (BSA, 2021). 8 ke nels m-2 we e sown a 6 cm dep h and 75 cm owing
space. Seeding da es and o he managemen measu es is shown in Table 2.
Table 2: De ailed ield his o y and ial managemen o all expe imen al seasons. Days a e sowing = DAS. G ow h
s ages (GS) ma ked: V0 – sowing, V5 – ge mina ion, V13/15/16 – hi d/ i h/six h lea e, R4 –dough y ke nels.
Managemen
2021
2022
Da e [MM-DD]
DAS
GS
Da e [MM-DD]
DAS
GS
Ploughing
05-10
-4
03-27
-40
Seedbed p epa a ion
05-14
0
04-19
-17
Sowing
05-14
0
V0
05-06
0
V0
P e-eme gence uni o m weeding
05-20
6
V5
05-11
5
V5
Fi s si e-speci ic mechanical weeding
06-09
26
V13
05-31
25
V13
Second si e-speci ic mechanical weeding
06-25
42
V15
06-15
40
V16
Ha es
06-10
144
R4
09-14
131
R4
In 2022, 10 ha-1 o d y chicken manu e we e applied be o e seeding.
Weed de ec ion
UAV-based mul ispec al came as we e used o weed and maize plan de ec ion. Di e en
came a sys ems and ligh al i udes we e es ed o image da a acquisi ion o de elop a lexible
and sys em-independen concep o spa ial di e en ia ion o maize and weeds. A MicaSense
Al um came a moun ed on a DJI Ma ice 210 as well as a DJI P4 Mul ispec al wi h a
pe manen ly ins alled mul ispec al came a we e used o acqui e UAV-based mul ispec al
images. The MicaSense Al um eco ds he su ace e lec ance in he isible (475 ± 32, 560 ±
27 and 668 ± 14 nm) and nea in a ed (717 ± 12 and 840 ± 57 nm) wa eleng hs. Spec al
in o ma ion is eco ded in compa able wa eleng hs by he P4 mul ispec al came a ( isible:
450 ± 16, 560 ± 16 and 650 ± 16 nm; nea in a ed: 730 ± 16 and 840 ± 26 nm). Fo obse a ion
da es in 2021, a ligh al i ude o 10 m was de ined which led o a spa ial esolu ion o 0.4 cm.
In he ollowing yea , he ligh al i ude was inc eased o 25 m o be able o simula e la ge a ea
ou pu s and as e image acquisi ion, ega ding a p ac ice-o ien ed applica ion con ex . The
esul ing images we e p ocessed using Agiso Me ashape (Ve sion 1.7.2). Va ious wea he
condi ions du ing da a acquisi ion esul ed in di e ences in exposu e, which we e co ec ed
using an illumina ion senso on he d ones. A e wa ds indi idual images we e aligned o an
o hopho o. Fo geome ic co ec ion o he image da a eigh e e ence panels dis ibu ed o e
he es a ea we e measu ed as g ound con ol poin s using a RTK GNSS ecei e (S onex
S9III).
The o hopho os o each eco ding da e we e classi ied using machine lea ning me hods,
o de e mine weed in es a ion in maize. The used me hods and sys ems o bo h yea s a e
shown in Table 3.
Scien i ic publica ions wi hin he con ex o his wo k
33
Table 3: Pa ame e s o he weed ecogni ion o he di e en sampling da es.
Da e
Fligh al i ude
Pixel esolu ion
Came a sys em
Machine lea ning me hods
[YY-MM-DD]
[m]
[cm]
2021-06-03
10
0.4
MicaSense Al um
Con olu ional neu al ne wo k
2021-06-17
10
0.4
MicaSense Al um
Con olu ional neu al ne wo k
2022-05-19
25
1.0
MicaSense Al um
Random o es
2022-06-10
25
1.0
DJI P4 mul ispec al
Random o es
In he i s ial yea a simple 2D con olu ional neu al ne wo k was implemen ed, his
algo i hm is inspi ed by he biological b ain and uses mul iple con olu ional laye s o ex ac
ea u es, such as ex u e, om images o iden i y objec s (Zhu e al., 2017). The CNN
a chi ec u e consis ed o h ee con olu ional laye s, each employing a 3 x 3 pixel ke nel o
cap u e impo an image ea u es. As inpu , he model ecei ed image pa ches measu ing 32
x 32 pixels, ex ac ed om he ield ial plo s showcasing di e ences in weed p essu e. The
aining o he CNN encompassed 25 epochs and was execu ed using he Adam op imize , a
s ochas ic g adien descen a ian . The aining p ocess ea u ed a ba ch size o 16 and
ini ia ed wi h an ini ial lea ning a e se o 0.0001.
The second me hod is RF classi ie , which is based on a numbe o unco ela ed decision
ees a ange in an ensemble. Wi hin his ensemble, a classi ica ion is ca ied ou by a majo i y
decision on all decision ees (B eiman, 2001). The RF implemen a ion was done using he
sciki -lea n so wa e lib a y (Ped egosa e al., 2011). The de aul se ings we e main ained o
he classi ie pa ame e s, as he RF has been shown o achie e good esul s wi h hem
(Fe nández-Delgado e al., 2014). The spec al in o ma ion o ial plo s was used o he
classi ica ion. In con as o he CNN, he RF does no conside ex u e ea u es.
Fo image pixel classi ica ion, dis inc ion was made be ween he classes “maize”, “weeds”
and “soil”. The esul ing classi ica ion maps should enable he localiza ion and quan i ica ion
o weed p essu e. The sampling o aining and alida ion da a was spa ially independen o
each o he . To compa e classi ica ion esul s o he conside ed da es o each o he classes,
p ecision and ecall, as well as he esul ing F1-sco e and o e all accu acy we e calcula ed.
The sampling o he alida ion da a was andomly d a ed om he eco ded images.
𝑃
𝑟=𝑇𝑃
𝑇𝑃+𝐹𝑃 Equa ion 2
𝑅𝑒=𝑇𝑃
𝑇𝑃+𝐹𝑁 Equa ion 3
𝐹1 = 2∗ 𝑃𝑟∗ 𝑃𝑒
𝑃𝑟+𝑃𝑒 Equa ion 4
𝑂𝐴 = 𝑇𝑃+𝑇𝑁
𝑇𝑃+𝑇𝑁+𝐹𝑃+𝐹𝑃 Equa ion 5
P ep esen s he p ecision, Re he ecall, F1 he F1-sco e and OA he o e all accu acy.
TP ( ue posi i e) indica es he numbe o posi i e samples co ec ly classi ied and TN ( ue
nega i e) he numbe o nega i e samples co ec ly classi ied. The numbe o inco ec
classi ied posi i e o nega i e samples is ep esen as FP ( alse posi i e) and FN ( alse
nega i e) espec i ely.
Based on he classi ica ion maps, he numbe o pixels o each class we e calcula ed o
each subplo wi h he unc ion “Zonal s a is ic” (Quan um GIS, Ve sion 3.22). Subsequen ly
WC and RWC we e calcula ed o each subplo wi h he “Field Calcula o ” unc ion o Quan um
GIS using he Equa ion 6 – Equa ion 8.
Scien i ic publica ions wi hin he con ex o his wo k
34
𝑀𝐶 =∑ 𝑃𝑀
∑𝑃𝑆+𝑃𝑊+𝑃𝑀 ∗100 Equa ion 6
𝑊𝐶 =∑𝑃𝑊
∑𝑃𝑆+𝑃𝑊+𝑃𝑀 ∗100 Equa ion 7
𝑅𝑊𝐶 = 𝑊𝐶
𝑊𝐶+𝑀𝐶 Equa ion 8
whe e MC ep esen s he maize co e , WC he weed co e , RWC he ela i e weed co e .
PS, PM and PW co espond o he classi ied pixels o he classes Soil, Maize o Weed
espec i ely.
The WC and RWC we e compa ed o he h eshold o he co esponding ea men le el
o each subplo . I he alue was highe han he h eshold o a subplo , weed egula ion was
done as desc ibed in sec ion Weed managemen , i no , no ha owing was conduc ed o his
subplo and imes amp. Due o he ime equi ed o p ocessing and analysing he UAV images,
as well as a ying wea he condi ions, weed egula ion was applied se en o en days a e he
ligh s. The classi ied maps we e u he used o show he weed dis ibu ion wi hin he ield.
Weed managemen
Be ween seeding and maize seed eme ging, all plo s we e ea ed wi h a p ecision ine
ha ow (T e le ) a a speed o 10 o 12 km h-1, as common in o ganic p ac ice o educe he
compe i ion o weeds in he ea ly g owing pe iod. This ea men was uni o m o each a ian
as e e y subplo was managed. Two si e-speci ic weed egula ions wi h a hoe ollowed,
ma ching he GS V13 and V16. The selec ion o he a eas o be ea ed was based on weed
de ec ion. Subplo s we e hoed only i hei weed co e pe subplo exceeded he speci ic
h eshold o each ea men . An excep ion was he Con- ea men , whe e e e y plo was
ea ed uni o mly. Hoeing was done wi h a Kombi-PP (Schmo ze ) in 2021 and a Chops a
(Einböck) in 2022. The change o equipmen was done due o a ailabili y a he esea ch
s a ion. Bo h machines we e equipped wi h duck oo sweeps, inge weede and au oma ic
came a s ee ing sys em. The de ices we e se up in he same way. D i ing speed was 5 o
8 km h-1. Wi h his equal equipmen , compa able weeding success we e a chi ed wi h bo h
machines. No chemical plan p o ec ion was applied o simula e o ganic a ming condi ions.
Soil and plan analyses
Visual assessmen s we e pe o med i e o se en days be o e and a e each weed
managemen in e en ion o es ima e weed co e age and species di e si y. Concu en ly, he
mean plan heigh wi hin each subplo was asce ained by a e aging he heigh s o ou
ep esen a i e maize plan s. Soil olume ic wa e con en was gauged a a dep h o 12 cm
using a TDR-150 (Field Scou ) de ice. Fo each measu emen a ea, wo eadings we e aken,
one in he in a- ow and one in he in e - ow space, o calcula e he a e age alue. These soil
mois u e eadings we e aken alongside o he measu emen s o cha ac e ize plan wa e
a ailabili y and ield he e ogenei y. Addi ionally, he phenological s age o he maize was
eco ded.
Maize yield was app oxima ed by manually ha es ing a 1 m² a ea a a heigh o
app oxima ely 15 cm abo e he g ound wi hin each subplo . The coun o maize cobs was
eco ded. The ha es ed plan s we e hen pa i ioned in o cob and lea -s em ac ions and
p ocessed using a Schliesing 220 ZX wood choppe . The esh weigh o each ac ion was
measu ed, ollowed by a d ying p ocess in an o en a 105 °C. The d y weigh was de e mined
Scien i ic publica ions wi hin he con ex o his wo k
35
pos a 48-hou pe iod. Weed biomass was ea ed simila ly, being weighed esh and
subsequen ly d ied.
Calcula ions & s a is ics
The i s s ep was he calcula ion o a i hme ic means o each plo o he pa ame e
( olume ic wa e con en , maize pan heigh , weed co e , numbe o weed species, ea ed
a ea, whole plan yield, cob yield, numbe o cobs and weed biomass a ha es ). In case o
maize yield, cob yield and weed biomass, he d y ma e was used. All pa ame e s (maize
yield, cob yield, weed biomass, weed co e , maize heigh , ea ed a ea, numbe o weed
species, olume ic soil wa e con en ) we e es ed o no mal dis ibu ion wi h he Shapi o-
Wilk es (α = 0.05; Roys on, 1995) and o homogenei y o a iance wi h Le ene’s es (α =
0.05; Fox, 2008). All pa ame e s showed a homogenei y o a iance. An analysis o a iance
(ANOVA; Chambe s e al., 1992) ollowed wi h an alpha o 0.05. I he al e na i e hypo hesis
was accep ed, a subsequen Tukey HSD-pos hoc es s (α = 0.05; Holland, 1988) ollowed.
Fo he a iables maize yield, cob yield and weed biomass he ixed ac o s we e he ea men
and yea , block and plo we e andom ac o s. Fo he a iable maize heigh , numbe o weed
species, olume ic soil wa e con en he ixed ac o s we e he ea men and da e, block and
plo we e andom ac o s. Coe icien o de e mina ion (R²) and a linea model a e compu ed
be ween olume ic soil wa e con en and weed co e .
All s a is ical analyses we e ca ied ou in R (Ve sion 4.2.2; R Co e Team, 2020) using he
packages “nlme”, “emmeans”, “mul comp”, “ s a ix” and “mul comp”.
Resul s
Plan de elopmen
The olume ic soil wa e con en (VSWC) was wi h 11.8 % (± 3.57 (s anda d de ia ion))
in 2021, o e all highe han 7.3 % (± 4.90) in 2022. Be ween he ea men s no di e ences
occu ed bu he imes amps showed al e a ions (ANOVA, F- alue = 6.38E+01, p- alue =
8.55E-14). No in e ac ion be ween yea and ea men was obse ed. A e he i s si e-speci ic
weeding a DAS 26/25 o 30, he VSWC was highe in 2021 han in 2022 as shown in Table
4. In he ime pe iod a e he second si e-speci ic hoeing (DAS 40 – 50), he soil mois u e
condi ions we e we e in 2021 han in 2022.
Table 4: Compa ison o olume ic wa e con en [%] be ween imes amp. DAS = Days a e sowing. SD = S anda d
de ia ion. Di e en le e s indica e signi ican (Tukey HSD-pos hoc es , alpha = 0.05) di e ences be ween
imes amps. N = 28.
Yea
DAS
Volume ic soil wa e con en
Mean [%]
SD
G oup
2021
30
8.65
1.12
B
2021
44
9.99
2.52
B
2021
46
9.85
2.52
B
2021
60
11.73
1.89
C
2022
34
11.56
3.01
C
2022
47
3.04
1.46
A
The maize plan heigh de eloped as e in 2021 han in 2022, as shown in Table 5. In he
i s ial yea he plan heigh was 11 cm la ge a DAS 34 han in he second yea and 64 cm
la ge a DAS 48. Be ween he ea men s, no signi ican di e ences we e obse ed.
In 2021, he weed g ow h was slow and only inc eased signi ican ly 49 days a e sowing
(Table 5). The hoeing a DAS 26 and 42 did no dec ease he weed co e , bu he g ow h

Scien i ic publica ions wi hin he con ex o his wo k
36
s agna ed. The yea 2022 showed a s ong g ow h o weed be ween DAS 17 and 24. Be o e
he i s hoeing a DAS 25, he weed co e was highe han in 2021. A e he weeding i
dec eased signi ican ly ill DAS 34. The weed co e inc eased by 5 % poin s om DAS 34 o
47, bu wi hou signi ican di e ence. The single ea men s did no di e in weed co e .
Table 5: Compa ison o maize plan heigh [cm], weed co e [%] and numbe o weed species be ween ime s amps.
DAS = Days a e sowing. SD = S anda d de ia ion. Di e en lowe case le e s indica e signi ican (Tukey HSD-
pos hoc es , alpha = 0.05) di e ences be ween imes amps. N = 28.
Yea
DAS
Maize plan heigh
Weed co e
Numbe o weed species
Mean [cm]
SD
G oup
Mean [%]
SD
G oup
Mean
SD
G oup
2021
20
11.63
0.67
A
2.59
1.46
A
3.66
0.56
E
2021
34
40.90
3.20
C
3.50
2.94
A
3.21
0.98
CDE
2021
49
114.26
5.88
E
11.41
10.13
BC
2.41
1.12
B
2021
60
192.45
7.92
F
14.00
11.05
CD
3.00
0.92
BCD
2022
17
12.57
1.23
A
4.80
3.52
AB
1.54
0.41
A
2022
24
12.57
1.23
A
19.89
14.78
D
3.51
0.61
DE
2022
34
29.28
3.73
B
10.44
4.10
BC
2.64
0.75
BC
2022
47
63.67
7.38
D
15.41
6.28
CD
1.49
0.37
A
In bo h yea s he mean numbe o weed species Table 5 dec eased a e he i s and
second si e-speci ic weeding (ANOVA, F-Value = 3.87E+01, p- alue = 2.65E-09, Table 5). In
2021 he mean weed species coun declined om 3.5 o 3.0 species, whe eas in 2022, a
dec ease om 3.5 o e 2.7 o 1.7 was obse ed.
The p edominan weed species ac oss he yea s was Chenopodium album (L.) wi h 196
(2021) o 576 (2022) coun s o e all plo s and imes amps. In addi ion, S ella ia media (L.)
(150/144 coun s) and Poa i ialis (141/113 coun s) we e p esen in la ge numbe s. Polygonum
con ol ulus (L.) occu ed manly in 2022 (138 in 2022 and only 44 in 2021). Galinsoga cilia e
(RAF.) we e ound 43 o 70 imes in 2021 and 2022 espec i ely. Few indi iduals o Ama a hus
( wo imes in bo h yea s), Capsella bu sa-pas o is (L.) (only in 2021, six imes), Echinochloa
c us-galli (L.) ( wo and 15 imes), Equise um a ense (L.) ( wo imes in 2022), Lamium
pu pu eum (L.) (ele en and 17 imes), Spe gula a ensis (L.) ( h ee imes in 2021) and
Ve onica ag es is (L.) we e ound a ou o se en spo s pe yea .
In 2021, i was obse ed ha weed co e inc eased wi h highe olume ic soil wa e
con en , as indica ed by a posi i e slope o 3.06 (Fig 8). The eg ession analysis yielded a
signi ican esul wi h an R² o 0.526 (p- alue = 2.2E-16). Howe e , his end was no ound in
2022. In ha yea , he ela ionship be ween weed co e and soil wa e con en emained
signi ican (p- alue = 7.28E-3), bu he s eng h o associa ion was ela i ely weak (R² = 0.118).
Unlike p e ious yea s, he ela ionship was nega i e, wi h a slope o -0.405. No ably, he
eg ession analysis o bo h yea s emained signi ican (R² = 0.034, p- alue = 1.99E-2).
O e all, he slope o 0.421 sugges s ha weed co e ends o inc ease wi h ising soil wa e
con en .
Scien i ic publica ions wi hin he con ex o his wo k
37
Fig 8. Linea eg ession be ween olume ic soil wa e con en [%] and weed co e [%]. Lines a e he eg ession
line wi h con idence in e al o 95 %. Colou o poin s indica es he yea s. N = 170, n 2021 = 110, n 2022 = 58.
Du ing he alida ion o he esul ing classi ica ion maps, an o e all accu acy o a leas
85.1 % was achie ed ac oss all imes amps (Table 6). In addi ion, F1-sco es o a leas 81.6 %
and 82.3 % we e achie ed o he a ge classes “maize” and “weeds”. Bo h machine lea ning
algo i hms showed simila accu acy, wi h CNN achie ing an OA o 85.1 % and 86.7 % and RF
ob aining an OA o 88.8 % and 92.0 %. Nei he he ligh al i ude no he came a sys em
a ec ed he esul s.
Table 6: Accu acies o UAV-based image classi ica ion o weeds, maize and soil. OA = O e all accu acy.
Times amp
o acquisi ion
Came a
sys em
F1-sco e [%]
OA [%]
Machine lea ning me hods
Maize
Weeds
Soil
2021-06-03
Micasense
81.6
86.1
87.9
85.1
Con olu ional neu al ne wo k
2021-06-17
Micasense
85.2
87.0
87.9
86.7
Con olu ional neu al ne wo k
2022-05-19
Micasense
85.6
82.3
98.3
88.8
Random o es
2022-06-10
Phan om
MS
88.4
88.3
99.3
92.0
Random o es
Fig 9 depic s he weed co e dis ibu ion de i ed om image classi ica ion be o e he i s
si e-speci ic weeding in e en ion in each yea . The mean weed co e was 2.24 % (± 1.60) in
2021 and 1.73 % (± 0.86) in 2022. The wes e n sec ion o he ield was co e ed consis en ly
highe wi h weed (2021: 2.7 %; 2022: 2.2 %) han he eas e n sec ion (2021: 1.58 %; 2022:
1.47 %). The wes e n sec ion also had he pa wi h he highes weed co e wi h 11 % (2021)
and 7.35 % (2022), as well as highe s anda d de ia ions o 1.78 and 0.97, e sus 0.99 and
0.62 in he eas e n sec ion. This indica es he p esence o some weed clus e s wi h highe
in ensi y in he wes e n sec ion, which we e mo e dispe sed in 2022 han in 2021. Be o e he
second hoeing he si ua ion was compa able bu on a highe le el. The wes e n pa showed
highe weed co e wi h 3.43 % (± 3.1) e sus 1.51 % (± 2.23) and 13.6 % (± 7.95) e sus
8.09 % (± 9.82) in 2021 and 2022 espec i ely.
Scien i ic publica ions wi hin he con ex o his wo k
38
Fig 9. Weed co e [%] es ima ed by image ecogni ion be o e i s si e-speci ic weed managemen in bo h ial
yea s. Da ke colou s indica e highe weed co e . The whole ield is displayed. In 2021, a he no he n pa o he
ield, a co e c op mix u e was sown, in his pa no weed de ec ion was conduc ed and he e o e no displayed.
The posi ion o he ial ield a ies in each yea o educe he e ec o SSWM on he ollowing yea .
As all h eshold alues we e exceeded in bo h yea s be o e he i s hoeing, all subplo s
we e ini ially ea ed equally. Fo he second hoeing, signi ican di e ences be ween he
a ian s we e obse ed. In 2021, RWC0.4 showed signi ican less ea ed a ea han Con,
WC0.25, WC0.5 and WC1.0 (Fig 10), none o he plo s o RWC0.4 we e hoed. RWC0.1,
RWC0.2, WC0.5 and WC1.0 ook an in e media e posi ion.
In 2022, he WC-T ea men s we e weeded comple ely, as well as Con and RWC0.1. Only
RWC0.2 showed some spa ed plo s, bu wi hou signi ican di e ences o he con ol
ea men . Less plo s o RWC0.4 we e hoed han he plo s o he o he ea men s, expec
RWC0.2 in 2022. O e all, wi h RWC0.4 less a ea was ea ed. In a e age, mo e plo s we e
hoed (ANOVA, F-Value = 2.15E+01, p- alue = 3.79E-05) in 2022 han in 2021.
Scien i ic publica ions wi hin he con ex o his wo k
39
Fig 10. Compa ison o ea ed a ea [m²] due o second hoeing be ween ea men (as shown as colou s): Con ol
(Con), Rela i e Weed Co e (RWC), Weed Co e (WC) and ial yea s. Whiske s show he minimum/maximum
alues. Ba s indica e he in e qua ile ange and columns he sum. Line inside he box symbolises he median.
Single alues a e shown as poin s. Lowe case le e s indica e signi ican (Tukey HSD-pos hoc es , alpha = 0.05)
di e ences be ween ea men s and yea s. N=4.
Maize yield and weed biomass a ha es
The alle g ow h in 2021 led o 3.5 imes highe maize yield in 2021 han in 2022 wi h
signi ican di e ences in yield (ANOVA, F- alue = 1.02E+03, p- alue = 4.02E-30) (Fig 11).
Fig 11. Compa ison o maize d y ma e yield [g/m-2] a op and weed biomass [g m-2] a bo om a ha es da e
be ween ea men (as shown as colou s): Con ol (Con), Rela i e Weed Co e (RWC), Weed Co e (WC) and he
ial yea s. Whiske s show he minimum/maximum alues. Ba s indica e he in e qua ile ange. Line inside he box
symbolises he median. Single alues a e shown as poin s. Le e s indica e signi ican (Tukey HSD-pos hoc es ,
alpha=0.05) di e ences be ween ea men s and yea s, lowe case le e s: maize d y ma e yield, uppe case le e :
weed biomass. N=4.
Scien i ic publica ions wi hin he con ex o his wo k
46
In oduc ion
Mode n ag icul u e aces he challenge o eeding a wo ld popula ion o abou 10 billion
people by 2050 and main aining ood secu i y (Wille e al., 2019). A he same ime, he
in ensi e ood p oduc ion has signi ican nega i e e ec s on he en i onmen , including a
decline in plan and animal species due o land use changes and in ensi e use o ni ogen (N)
e ilize s and pes icides (Geige e al., 2010; Wille e al., 2019). O ganic ag icul u e can
enhance wildli e by p omo ing soil e ili y and educing he use o N e ilize s and pes icides.
O ganic a ming, howe e , ypically achie es only 80 % o 83 % o he yields o con en ional
a ming, wi h a wide ange o 40 – 130 % (de la C uz e al., 2023; De Pon i e al., 2012). This
yield gap is mainly due o lowe inpu s o N e ilize s (Dö ing and Neuho , 2021). In o ganic
a ming, N e ilize sou ces a e limi ed o o ganic op ions such as biological N ixa ion, g een
manu e, compos , e c. The limi ed supply necessi a es highe N use e iciency and be e
dis ibu ion wi hin he ield and o e ime.
Mos ag icul u al land is cha ac e ized by he e ogeneous soil condi ions, including soil
ex u e, o ganic ma e , elie , nu ien , and wa e supply. Consequen ly, plan de elopmen is
in luenced by spa ial di e ences (Usowicz and Lipiec, 2017). While soil ex u e and soil o ganic
ma e (SOM) a e ela i ely s able o e ime (Behe a e al., 2018), nu ien and wa e supply
can a y g ea ly wi hin yea s (Aksakal e al., 2019; Zhu e al., 2013). Addi ionally, land use
his o y can in luence cu en c op de elopmen (Schulp and Ve bu g, 2009). The main N
sou ce in o ganic ag icul u e is he symbio ic N ixa ion by legumes, mainly clo e -g ass
(Obe son e al., 2024). The amoun o ixed N depends s ongly on biomass p oduc ion and
a ies wi h he wea he condi ions (Hoeks a e al., 2015). Thus, he de elopmen o clo e
a ec s he N supply o cu en and subsequen c ops.
Sub ield adap ed managemen can add ess issues o he e ogeneous plan de elopmen .
Fields can be delinea ed in o managemen zones (MZ), whe e each sub ield has simila plan
and/o soil p ope ies (Diacono e al., 2014; Pe al a e al., 2015). This di ision in sub ields can
be done using a ious algo i hms, wi h he uzzy C-means clus e ing algo i hm being widely
used ecen ly (Damian e al., 2020; Oldoni e al., 2025; Rod iguez Mi anda e al., 2021). In
con en ional ag icul u e, spa ial di e ences can be mi iga ed wi h as -dissol ing a i icial
e ilize s, bu in o ganic ag icul u e, he legal e ilize like g een manu e o cow manu e wo k
slowly. The e o e, ield di e ences mus be balanced in ad ance wi h o ganic e ilize s (Co i
e al., 2023). Ano he app oach is o adjus seeding densi y o use c op mix u es o e en ou
spa ial he e ogenei y (Munna e al., 2022). The e is a discussion abou whe he lowe -
pe o ming a eas should ecei e esou ces like seeds o e ilize , he "Robin Hood" app oach,
o i high-po en ial a eas should ecei e mo e esou ces o maximize hei po en ial he so
called "King" app oach.
Typically, he e ogeneous condi ions a e es ima ed by soil sampling, bu his can be
uneconomical due o labou and analysis cos s (Buladaco II e al., 2024; Ouazaa e al., 2022).
As an al e na i e, emo e senso s can be employed o iden i y spa ial di e ences wi hin plan
communi ies (B eunig e al., 2020) and can subs i u e soil sampling (Spe anza e al., 2023),
o e ing a less labou -in ensi e solu ion. Vege a ion Indices (VI), such as he No malized
Di e ence Red Edge Index (NDRE), a e pa icula ly e ec i e o his pu pose and a e widely
u ilized (Damian e al., 2020; Gi z and Ma ila, 2024; Leo e al., 2023). The NDRE conside s
he e lec ance in o ma ion om NIR and RedEdge o highligh di e ences in chlo ophyll
concen a ion wi hin plan s ands (Raymond Hun e al., 2011). Unmanned ae ial ehicles
(UAVs) play a c ucial ole in his p ocess, as hey can swi ly and lexibly cap u e images

Scien i ic publica ions wi hin he con ex o his wo k
47
needed o NDRE calcula ion, making i a p ac ical subs i u e o adi ional soil sampling
me hods (Rasmussen e al. 2021).
Al hough UAV images ha e been success ully used o delinea e ields o co e and ce eal
c ops, only a ew s udies ha e analysed managemen zones o c op o a ion (Oldoni e al.,
2025; Ouazaa e al., 2022). To ou knowledge, none ha e conduc ed ials wi h clo e -g ass
ollowed by a subsequen ce eal c op unde o ganic a ming condi ions. Consequen ly, his
s udy aims o use UAV-based NDRE maps o clo e -g ass ields o delinea e managemen
zones and e alua e hei impac on subsequen c op p oduc i i y. The indings migh p o ide
insigh s in o op imizing N use e iciency in o ganic a ming sys ems. This s udy aligns wi h he
Uni ed Na ions' Sus ainable De elopmen Goals (SDGs), pa icula ly Goal 2: Ze o Hunge ,
and Goal 15: Li e on Land. By p omo ing sus ainable ag icul u al p ac ices, we aim o
con ibu e o hese global objec i es. The ollowing hypo heses we e e alua ed:
• Hypo hesis I: High p oduc i i y zones du ing he clo e -g ass pe iod a e also highly
p oduc i e o he subsequen c op.
• Hypo hesis II: NDRE images a e use ul o delinea ing managemen zones in clo e -
g ass ields.
• Hypo hesis III: Highe clo e yields lead o highe ixed N and, he e o e, highe soil N
con en (SNC) du ing bo h he clo e -g ass pe iod and he subsequen c op.
To es hese hypo heses, h ee ials we e conduc ed in no hwes e n Ge many unde
o ganic on- a ming condi ions.
Scien i ic publica ions wi hin he con ex o his wo k
48
Ma e ials & Me hods
T ial design and s udy si e
The s udy was conduc ed nea Osnab ück in Lowe Saxony, in he no hwes o Ge many,
om 2020 o 2023. Th ee o ganically managed ields we e selec ed, each app oxima ely
1 hec a e in size. The dis ances be ween hem anged be ween 500 and 4,100 m (Fig 12).
Fig 12. Geog aphical Dis ibu ion o Expe imen al Si es in No hwes e n Ge many nea Osnab ück (Backg ound
Sou ce: Google Maps, 2024)
The soil alue is a Ge man indica o o yield po en ial, wi h 0 indica ing he lowes and 100
he highes e ili y (A bei sg uppe Boden (Soil wo kingG oup) 2024 Table 7).
Table 7: Field desc ip ion. Soil ypes and soil alues a e based on NIBIS® Map Se e (2023). The Soil alue is a
Ge man indica o yield po en ial, he e 0 indica es he lowes and 100 he highes e ili y (A bei sg uppe Boden
(Soil wo king G oup), 2024). Da a sou ce o soil alue and soil ype: (LBEG, 2023). AMSL = abo e mean sea le el.
Field name
Yea
Geog aphical
coo dina es
Soil ype
Soil ex u e
AMSL [m]
Soil
alue
Powe Weg
2020,
2021
52°18'59.3"N,
8°06'18.7"E
Cambisol
Loamy sand / sandy loam
113 – 115
52 – 55
Kiesschach
2021,
2022
52°19'25.8"N,
8°09'28.9"E
Cambisol
Loamy sand / sandy loam
88 – 92
30 – 40
B eme S aße
2022,
2023
52°19'34.3"N,
8°09'46.7"E
Regosol
Loamy sil / sandy loam
121 – 123
42 – 62
Powe Weg has a low slope g adien and is nea ly la wi h an abo e mean sea le el
(AMSL) o 113 o 115 m (Fig 13). The Kiesschach ield has he s eepes elie , wi h ele a ions
anging om 88 o 92 m AMSL. Especially he wes e n pa has a s eep slope. B eme S aße
ield has di e ences in slope compa able o he Powe Weg.
Scien i ic publica ions wi hin he con ex o his wo k
49
Fig 13. Te ain based on digi al ele a ion model o he ial ields. AMSL (abo e mean sea le el) [m] is indica ed
by colou : blue, lowe alues; yellow, highe alues. Rec angles show he plo s. Each map has a sepa a e scale
o easie iden i ica ion o di e ences.
The ial was conduc ed as an on- a m expe imen , whe e he p ima y managemen
decisions ( ime o ha es , c op choice, e c.) we e made by he a me . All ields ha e been
managed o ganically since 1995. Each ield was obse ed du ing wo seasons: he clo e -
g ass pe iod and he subsequen ce eal c op season. The seedbed o clo e -g ass was
p epa ed using a plough and disk ha ow a e he ha es o he p e-c op in summe . A seed
mix u e o 32 kg ha−1 was sown, consis ing o he ollowing pe cen ages: 11.11 % ed clo e
(T i olium p a ense) 'Luc um bio', 11.11 % ed clo e 'Ti us bio', 11.11 % alsike clo e (T i olium
hyb idum L.) 'Au o a', 11.1 1 % whi e clo e (T i olium epens) 'Li lex', 7.78 % pe ennial
yeg ass (Lolium pe enne) 'Mel os ', and 47.78 % pe ennial yeg ass 'Vale io bio'. The clo e -
g ass swa d was ha es ed wo o h ee imes pe yea , wi h all biomasses emo ed om he
ield. The subsequen c op was a ce eal. Powe Weg and Kiesschach we e illed wi h a plough
and disk ha ow. Summe spel (T i icum aes i um subsp. Spel a) 'Flaude ' (180 kg ha−1) was
sown in Ma ch 2020 and 2021. B eme S aße was illed in he same way in Oc obe 2022,
ollowed by he sowing o 195 kg ha−1 o win e whea (T i icum aes i um L.) 'G annosos'.
To sion ha owing was applied a Powe Weg a g ow h s age (GS) 15. On he ield Powe Weg
10 Mg ha-1 o cow manu e was applied and ploughing o clo e -g ass was pe o med. A
B eme S aße, 13 Mg ha-1 o cow manu e was applied du ing GS 37. A de ailed lis o he
managemen p ac ices can be ound in Table 8.
Table 8: Field managemen and soil sampling da es. Da e o ma YYYY-MM-DD.
Field name
Clo e -g ass pe iod
Ce eal pe iod
Tillage
Seeding
Ha es
Tillage
Seeding
Ha es
Powe Weg
2019-08-01
2019-08-01
2020-04-26
2020-07-11
2020-09-20
2021-03-24
2021-03-24
2021-08-11
Kiesschach
2020-08-20
2020-08-25
2021-05-14
2021-07-16
2022-03-09
2022-03-14
2022-07-25
B eme S aße
2021-07-27
2021-08-01
2022-06-01
2022-08-03
2022-10-13
2022-10-17
2023-06-18
A he Powe Weg ield, he p edominan weed species du ing he clo e -g ass pe iod we e
Papa e hoeas (common poppy), Cen au ea cyanus (co n lowe ), and Ma ica ia chamomilla
(chamomile). Du ing he summe spel c opping, addi ional weed species such as S ella ia
Scien i ic publica ions wi hin he con ex o his wo k
50
media (chickweed), T i olium spp. (clo e ), and Con ol ulus a ensis ( ield bindweed) we e
p esen . One applica ion o ine ha ow was conduc ed a GS 59. In he Kiesschach ield, he
mos p e alen weeds we e Lamium pu pu eum ( ed deadne le), Cen au ea cyanus
(co n lowe ), Vicia spp. ( e ch), and Rumex spp. (dock). Du ing he subsequen c opping, Vicia
spp. ( e ch), Cen au ea cyanus (co n lowe ), Papa e hoeas (common poppy), Lolium ssp.
(g ass), and S ella ia media (chickweed) we e ound. In he clo e -g ass ield B eme S aße,
species such as Ci sium spp. ( his le), Lamium pu pu eum ( ed deadne le), B assica napus
( apeseed), Rumex spp. (dock), Ve onica spp. (speedwell), Alopecu us spp. ( ox ail), and
Ma ica ia chamomilla (chamomile) occu ed. Du ing he win e whea pe iod, he ield mos ly
ea u ed species such as T i olium spp. (clo e ), Ve onica spp. (speedwell), S ella ia media
(chickweed), and Galium apa ine (clea e s).
Plan and soil sampling
On each ield, 48 plo s we e es ablished in a g id, wi h each plo being a squa e wi h a side
leng h o 4 me e s (Fig 13). A RTK GNSS e o caused es ima ion-based placemen o plo s
a Kiesschach ield, wi h exac posi ions eco ded la e . Wi hin each plo , a sampling a ea o
0.25 m² was designa ed o clo e -g ass, and 1 m² was designa ed o he subsequen ce eal
c op. The posi ion o he clo e -g ass sampling a ea was sligh ly shi ed a e each sampling
o ensu e an unha es ed sampling a ea be o e each obse a ion. In con as , he ce eal
measu emen s a ea emained cons an o he en i e season, as he sampling was non-
des uc i e wi h he excep ion o he ha es . The GPS posi ion o each sampling a ea was
eco ded using he RTK GNSS ecei e Helix M7 (Helix Geospace, Uni ed Kingdom).
Righ be o e each clo e -g ass ha es and e e y en o 14 days a e wa ds, each
sampling a ea was obse ed. The imes amp immedia ely p eceding he ha es will be
e e ed o as “ha es ”, while all o he imes amps will be designa ed as “sampling da es”.
Plan heigh was calcula ed as he a e age o ou measu emen s wi h a olding ule.
Volume ic soil wa e con en (VSWC) was measu ed a wo poin s using a Field Scou TDR-
150 (SPECTRUM TECHNOLOGIES, USA) a a dep h o 120 mm. Due o lack o access o
measu ing de ice, his was only used in ials s a ing in 2021. Visual assessmen s we e
conduc ed o es ima e he yield p opo ion (Klapp and S ählin, 1936) and GS (Moo e e al.,
1991) o g ass and clo e species. Addi ionally, he swa d was cu and sepa a ed in o he
ac ions "G ass", "Clo e ", and "Weeds". Fo each sampling a ea, he o al d y biomass and
he d y mass o he ac ions we e de e mined by d ying a 85 °C o 48 hou s.
Du ing he ce eal pe iod, a ele an GS (de e mined acco ding o BBCH (Maie , 2018)),
plan heigh was de e mined o each sampling a ea using he same p ocedu e as o clo e -
g ass. A ha es ime, he leng h o ea s was measu ed in he same manne . The numbe o
plan s pe me e was coun ed and hen con e ed o a pe -squa e-me e basis by mul iplying
he coun by se en. VSWC was measu ed wi h in he same way as du ing he clo e -g ass
pe iod. Addi ionally, weed species and weed co e we e es ima ed by isual assessmen o
an a ea o 0.1 m², as isual assessmen s a e mo e p ecise o smalle a eas. Yield was
de e mined by hand-ha es ing o all plan s. Subsequen ly, ce eals and weeds we e
sepa a ed, he esh weigh o each ac ion was measu ed, hen d ied in an o en a 85 °C o
48 hou s, a e which he d y weigh was de e mined. Ke nels we e sepa a ed om s aw using
a HALDRUP LT-35 labo a o y h eshe (Hald up, Ge many). A Con ado seed coun e
(P eu e , Ge many) was used o de e mine housand-ke nel weigh (TKW). Ke nel and s aw
samples we e hen milled o 0.5 mm wi h a ZM 200 (Re sch, Ge many), ollowed by d ying o
48 hou s a 105 °C. N es ima ion was pe o med wi h he Elemen al Analysis Leco FP 628
(Leco, Ne he lands). Finally, p o ein con en was calcula ed by mul iplying he N con en by
Scien i ic publica ions wi hin he con ex o his wo k
51
6.25, based on he assump ion ha p o ein con ains an a e age o 16 % N (ISO 16634-2,
2016).
Du ing he clo e -g ass and ce eal pe iods, soil sampling was conduc ed a ele an
imes amps; a leas one sampling be o e illage o clo e -g ass and du ing he ege a i e
g owing o ce eal. Th ee soil samples wi hin each plo we e collec ed om dep hs o 0 o 0.3,
0.3 o 0.6, and 0.6 o 0.9 m. Due o se e e d y soil condi ions, i was no always possible o
sample he deepe laye s. Each laye was homogenized and cooled un il analysis. A
comme cial labo a o y analysed he samples o soil mine al N (NH4+ and NO3−) based on
VDLUFA I, A 6.1.4.1:2002 (VDLUFA, 2007). Soil ex u e and clay con en we e es ima ed by
he same labo a o y using VDLUFA I, D 2.1:1997 (VDLUFA, 2007). The exac days o da a
sampling and UAV campaigns can be ound in Table S 1 (Supplemen a y Ma e ial).
Wea he condi ions
Wea he da a was eco ded by a Ge many's Na ional Me eo ological Se ice (Deu sche
We e diens ) wea he s a ion a 103 me e s abo e mean sea le el, loca ed app oxima ely 1.0
o 4.3 km away om he ields (52°19'01.2"N, 8°10'09.8"E). Ai empe a u e was measu ed a
a heigh o wo me e s. Fig 14 illus a es he weekly wea he condi ions o he ial yea s in
compa ison o he long e m pe iod.
Fig 14. Mean ai empe a u e [°C] (lines) and sum o p ecipi a ion [mm] (ba s) o each week o ial pe iod in
compa ison o he long- e m pe iod o 2011 – 2019 in g ey. Da a sou ce: (DWD, 2024).
Du ing he g owing pe iod om Ap il o Sep embe , he yea 2020 had an a e age
empe a u e o 15.5 °C, which was 0.3 °C wa me han he long- e m a e age o 15.2 °C. In
2021, he empe a u e was 14.3 °C, making i 1 °C coole han he long- e m a e age. The
yea 2022 expe ienced an a e age empe a u e o 15.7 °C, which was 0.5 °C wa me han he
long- e m a e age. The inal ial pe iod was compa able o he a e age, wi h a empe a u e
o 15.1 °C. Peak empe a u es we e obse ed in all yea s, pa icula ly du ing he summe
mon hs. The long- e m a e age p ecipi a ion du ing he g owing pe iod was 461 mm. The yea s
2020 and 2022 expe ienced wa e de ici s, wi h p ecipi a ion le els o 313 mm and 408 mm,
espec i ely. Con e sely, p ecipi a ion in 2021 and 2023 was highe han he long- e m
a e age, a 575 mm and 654 mm, espec i ely. Ne e heless, pe iods o se e e d ough
occu ed in all yea s.

Scien i ic publica ions wi hin he con ex o his wo k
52
Image acquisi ion and calcula ion o ege a ion indices
Pa allel o he sampling in clo e -g ass, mul ispec al images we e acqui ed ia UAV. In
2020, an Al um came a (MicaSense, USA) moun ed on a Ma ice 210 (DJI, China) was used
in ligh , wi h a i ude be ween 25 m and 46 m. In 2021 and 2022, a DJI P4 Mul ispec al (DJI,
China) wi h a pe manen ly-ins alled mul ispec al came a was used in a ligh wi h an a i ude
o 10 m o 14 m. The MicaSense Al um cap u es su ace e lec ance in he isible wa eleng h
ange (475 nm ± 32 nm, 560 nm ± 27 nm, and 668 nm ± 14 nm) as well as in he nea -in a ed
(717 nm ± 12 nm and 840 nm ± 57 nm). Simila ly, he P4 mul ispec al came a eco ds spec al
da a in compa able wa eleng hs ( isible: 450 nm ± 16 nm, 560 nm ± 16 nm, and 650 nm ±
16 nm; nea -in a ed: 730 nm ± 16 nm and 840 nm ± 26 nm).
The a ying wea he condi ions du ing da a collec ion led o di e ences in exposu e, which
we e adjus ed using an illumina ion senso moun ed on bo h UAVs. Subsequen ly, he
indi idual images we e aligned o an o hopho o. To co ec he geome y o he image da a,
en e e ence panels sca e ed ac oss he es a ea we e used as g ound con ol poin s and
measu ed wi h an RTK GNSS ecei e S onex S9III (S onex, Ge many). Fo his, he so wa e
Agiso Me ashape (Ve sion 1.7.2) was used.
The images we e pos -p ocessed using Quan um GIS (Ve sion 3.34). Based on he
o hopho os, NDRE images (Tucke , 1979) we e calcula ed using Equa ion 9 wi h he as e
calcula o unc ion.
𝑁𝐷𝑅𝐸 = 𝑁𝐼𝑅−𝑅𝑒𝑑𝐸𝑑𝑔𝑒
𝑁𝐼𝑅+𝑅𝑒𝑑𝐸𝑑𝑔𝑒 Equa ion 9
In he NDRE calcula ion, NIR ep esen s he e lec ance in he nea -in a ed (840 nm), and
RedEdge deno es o he e lec ance in he ed edge egion (720 nm).
All NDRE images we e esampled o a pixel esolu ion o 1 me e using he . esample
unc ion. This ensu es uni o m esolu ion, ma ching he lowes esolu ion o images collec ed
a a ligh al i ude o 46 me e s. A 1-me e pixel esolu ion equi es less s o age space and
ligh ime, making i mo e p ac ical. Addi ionally, he images we e c opped o he ield ex en
using he mask laye unc ion o minimize he in luence o bo de s, ails, and ees.
Delinea ion o managemen zones
The delinea ion o managemen zones was conduc ed using a spa ial gene alized uzzy
C-means algo i hm wi hin he R so wa e en i onmen (Ve sion 4.2.2; R Co e Team 2020).
This analysis u ilized he "geocmeans," "ggub ," " u u e," " map," and " e a" packages. This
app oach is an unsupe ised machine lea ning algo i hm ha g oups da a poin s based on
sha ed a ibu es. All poin s wi hin a clus e a e simila o each o he , whe eas he clus e s
hemsel es a e dis inc . Unlike K-means clus e ing, uzzy C-means employs so clus e ing,
in ol ing p obabilis ic clus e membe ships. This me hod can e ec i ely handle complex and
o e lapping clus e s. The spa ial gene alized e sion o he algo i hm con e ges mo e apidly
and is less sensi i e o noise due o i s inco po a ion o spa ial in o ma ion (Zhao e al. 2013).
Fo each ield, wo imes amps o NDRE images we e selec ed. These imes amps
ep esen a comp omise among iming jus be o e he ha es da e, image quali y, and he
combina ion o ege a i e and ep oduc i e s ages o g ass and clo e . Two imes amps we e
chosen o balance empo al he e ogenei y wi hou he isk o edundancy. Fo Powe Weg, he
chosen da es we e June 12, 2020 (29 days be o e he second ha es , all species in
ep oduc i e s ages) and Sep embe 10, 2020 (di ec ly be o e he hi d ha es , g ass in he
ege a i e s age, clo e species in he ep oduc i e s age). Da a o he i s ha es we e no
Scien i ic publica ions wi hin he con ex o his wo k
53
collec ed, due o ime delay in ial s a . Fo Kiesschach , May 14, 2021 was selec ed as i was
di ec ly be o e he i s ha es , wi h all species excep whi e clo e in he ep oduc i e s age.
Addi ionally, Augus 26, 2021 was chosen: 11 days be o e he second ha es , wi h g ass and
whi e clo e in he ege a i e s age, and ed clo e and alsike clo e in he lowe ing s age.
On B eme S aße, he imes amps 29 days a e he i s ha es we e selec ed (June 16,
2022), wi h ed clo e lowe ing and he o he species in he ege a i e s age. The second da e
was July 28, 2022, di ec ly be o e he second ha es , whe e he g ass plan s we e in
elonga ion, ed clo e in he ege a i e s age, and bo h alsike clo e and whi e clo e we e
lowe ing.
The numbe o clus e s (k) was es ima ed using he elbow me hod wi h he Calinski-
Ha abasz index, pa i ion en opy, and he silhoue e index o uzzy C-means. Calinski-
Ha abasz index and silhoue e index measu e he cohesion o clus e , whe e highe alues
indica e dense clus e s wi h be e di e en ia ion o o he clus e s. Pa i ion en opy indica es
he o ganisa ion o unce ain y, whe e highe alues a e a sign o highe o ganisa ion, and
lowe alues ep esen a uzzie esul (B i o Da Sil a e al., 2020). A signi ican di e ence
be ween k = 2 and k = 3 was obse ed o all ields, wi h a la e slope o highe alues o k
(Fig S 1, Supplemen a y Ma e ial); hus, k = 2 was chosen o all ields. Addi ional, k = 3 was
es ed, o e alua e he nex highe numbe o clus e s. O he pa ame e s - m (1.5), alpha (1.3),
be a (0.5), and window size (5 x 5) - we e de e mined by compu ing he silhoue e index and
explained ine ia o he combina ions o m (1 – 2), alpha (0.5 – 2), be a (0.1 – 1) and window
size (3 x 3, 5 x 5, 7 x 7).
Addi ionally, a each ield he clus e ing was conduc ed o each imes amp sepa a ely, o
obse ing empo al di e ence wi hin he clo e -g ass zone.
The clus e assignmen o each plo was ex ac ed om he clus e image by majo i y
decision and will be e e ed o as "Clo e g ass zone" (CGZ) om now on, whe e CGZ 1 is
associa ed wi h lowe and CGZ 2 wi h highe yield po en ial. In he case o h ee CGZs, he
low-yield clus e is called CGZ 1, he high-yield clus e CGZ 3, while CGZ 2 has a posi ion in
be ween bo h. The e m “Managemen Zone” (MZ) is used in discussion abou sub- ields in
gene al and includes CGZ.
Mean compa ison o zones
The goal is o delinea e he ields in o zones wi h signi ican ly di e en ag icul u al
p ope ies. Fo his, mean compa isons be ween he zones we e conduc ed. Gi en he
a ia ion in soil condi ions and wea he ac oss he ields and yea s, he CGZs we e compa ed
sepa a ely o each ield. Fo his analysis, he R so wa e en i onmen (Ve sion 4.2.2; R Co e
Team 2020) was used, inco po a ing he "dply ," " idy ," " s a ix," "pu ," and "b oom"
packages. The assump ions o no mali y and homogenei y o a iance we e es ed using he
Shapi o-Wilk es (α = 0.05; Roys on, 1995) and Le ene's es (α = 0.05; Fox, 2015),
espec i ely. While mos pa ame e s exhibi ed homogenei y o a iance, no all did. The e o e,
he Welch es (α = 0.05; Welch, 1947) was selec ed o compa e he means o he wo CGZs,
as his es does no need homogenei y o a iance and only wo g oups we e compa ed. Fo
he case o h ee CGZs, he K uskal-Wallis (α = 0.05) es was chosen wi h Dunn-Bon e oni-
Tes s (α = 0.05) as pos hoc es . Bo h es s lack he assump ion o no mali y and homogenei y
o a iance (Hollande , 1999).
Scien i ic publica ions wi hin he con ex o his wo k
54
Resul s
NDRE maps and managemen zones
The mean NDRE o Powe Weg was 0.9 on June 12, 2020 and 0.8 on Augus 27, 2020,
wi h a coe icien o a ia ion (CV) o 6.9 and 14.0 espec i ely. The eas e n im shows lowe
alues a bo h da es (Fig 15). The i s obse a ion o Kiesschach showed a mean NDRE o
0.49 (CV: 10.6) on May 14, 2021, which dec eased o 0.28 (CV: 18.8) a he second
obse a ion on Augus 26, 2021. On May 14, 2021, he wes e n pa shows lowe alues, while
on Augus 26, 2021, his pa e n e e ses. The mean NDRE alues o B eme S aße a e lowe
a 0.08 (CV: 60.6) on he i s da e (June 16, 2022) han on he second da e (July 28, 2022),
which shows an NDRE o 0.26 (CV: 26.1). On bo h da es, lowe NDRE alues we e obse ed
in he no heas .
Fig 15. NDRE-Maps (no malized di e ence ed edge) o clo e -g ass o he di e en ields and selec ed da es.
The clus e ing algo i hm e ealed wo CGZs pe ield. A he Powe Weg ield, 80.3 % and
42 o 48 plo s belong o CGZ 1. CGZ 2 can be ound a he eas e n im and in some smalle
spo s, as shown in Fig 16. In he Kiesschach ield, 63.7 % and 35 plo s a e pa o CGZ 1,
mos ly loca ed in he eas e n pa o he ield. CGZ 2 can be ound along he slope and in he
lowe a eas in he wes e n pa o he ield. The B eme S aße ield shows a mo e agmen ed
pa e n in he sou hwes and sou heas ; mos o CGZ 2 can be ound in he cen e and
no hwes , while CGZ 1 is loca ed in he no heas . In his ield, 45.8 % o he obse ed a ea
and 30 plo s belong o CGZ 2.
The delinea ion in o h ee CGZs did no show addi ional di e ences in he soil and
ag onomic pa ame e s, he second CGZ akes an in e media e posi ion be ween CGZ 1 and
3, wi h no signi ican di e ences o one o bo h CGZs. The e o e he esul s a e no shown
u he .
Scien i ic publica ions wi hin he con ex o his wo k
55
Fig 16. Maps o he delinea ed o clo e -g ass zones (CGZ) o each ield. Ligh blue is CGZ 1, da k blue is
CGZ 2. Rec angles show he plo s, illing indica es he zone.
Clo e -g ass g ow h
The clo e -g ass yield o he Powe Weg ield a ied be ween 1285.9 and 1562.7 g m-2,
(Fig 17) bu did no di e be ween he CGZs on any da e. The hi d ha es showed a highe
clo e a e in CGZ 2 o 82.5 % compa ed o 68.1 % in CGZ 1 (Fig 18).
Fig 17. Yield o clo e , g ass, and weeds a e compa ed be ween he wo clo e -g ass zones (CGZ) o di e en
ields and imes amps. Do s ep esen he indi idual alues o he clo e -g ass yield. The "Ha es " column
desc ibes he sampling be o e ha es ing. The "Sample" column desc ibes he sampling be ween ha es s. The
"Yea " column is he annual sum o yield. Red signs indica e Welch-Tes esul s: "ns" (p > 0.05), "*" (p < 0.05), "*"
(p < 0.005), "**" (p < 0.0005), "****" (p < 0.00005).
The annual yield o Kiesschach ield was 4386.8 g m-2. On he i s ha es da e (May 14,
2021), CGZ 2 o he Kiesschach ield achie ed 320 g m-2 highe o al biomass (p < 0.002,
Welch es ) in compa ison o 1061.2 g m-2 in CGZ 1. The annual g ass yield was also highe
in CGZ 2, bu no he clo e biomass. The annual yield o B eme S aße was lowe han he
o he ields wi h 1320.5 g m-2. CGZ 2 shows highe o al biomass on he i s (CGZ 2: 84.1,
CGZ 1: 63.7 g m-2, p = 0.029, Welch es ) and second sample da e (CGZ 2: 69.1, CGZ 1:
51.1 g m-2, p = 0.046, Welch es ) a e he i s ha es , due o highe clo e yield (p = 0.017
Scien i ic publica ions wi hin he con ex o his wo k
62
e al., 2025). Ano he possibili y is si e-speci ic seeding (Munna e al., 2022). Addi ionally, c op
a ie y o species could be selec ed based on he MZs. Fo example, summe spel is known
o i s lowe soil quali y equi emen s compa ed o win e whea . A mix u e o species, also
known as in e c opping, can balance ou lowe soil e ili y, wi h suppo ing c ops like peas.
Legumes can ans e biologically- ixed N o he pa ne c op (Zhao e al. 2022) o compensa e
o yield gaps i he main c op pe o ms poo ly (Munz e al., 2023). An addi ional app oach
could be o manage he low yield a eas less in ensi e, wi h lowe e ilisa ion, weeding and
illage o p omo e biodi e si y.
The e is an ongoing deba e on whe he lowe -pe o ming a eas should ecei e mo e
suppo h ough seeding o e iliza ion ( he "Robin Hood" app oach) o i high-po en ial zones
should ecei e mo e esou ces o maximize hei po en ial ( he "King" app oach). Fo si e-
speci ic seeding o maize, he "King" app oach is mo e sui able (Munna e al., 2022).
Meanwhile, Co i e al. (2023) used si e-speci ic manu e applica ion o e en ou SOM. In
con as , Mo a i e al. (2018) achie ed highe yields wi h he "Robin Hood" app oach in du um
whea , hough his ca ies he isk o highe losses due o leaching.
The delinea ion o ields based on he p e-c op has been done mul iple imes (B eunig e
al., 2020; Gi z and Ma ila, 2024; Oldoni e al., 2025; Ouazaa e al., 2022), bu o ou
knowledge, i has no been es ed in p ac ice. Fu u e esea ch should es si e-speci ic
applica ions (e.g. si e-speci ic seeding o e iliza ion) in on- a m expe imen s. Addi ionally, he
"King" and "Robin Hood" app oaches should be e alua ed o e iliza ion and a iable seeding.
Conclusions
This s udy demons a es he use o UAV-based NDRE image y o he delinea ion o
clo e -g ass ields and he subsequen ce eal c ops. The delinea ed CGZs we e di e en bo h
in he clo e -g ass pe iod, as well as in subsequen ce eal pe iod. Fo all h ee ields, high
p oduc i i y zones and low p oduc i i y zones we e ound. The di e ences a e mos ly due o
spa ial he e ogenous soil condi ions. Despi e hese esul s, he ields a e e y di e en o each
o he ega ding e ili y, soil ex u e and elie , u he in luencing he quali y o clo e -g ass and
ce eal. Image acquisi ion wi h UAVs o e s a lexible and accu a e base o delinea ion o sub
ields, he NDRE was sui ed o his ask. UAVs addi ionally o e s he oppo uni y o use he
da a o o he applica ions, like si e-speci ic weeding o plan moni o ing. Con a y o he
hypo hesis, highe clo e yield did no esul in highe SNC, di e ences in he subsequen c op
a e he e o e due o soil di e ences, elie and o he e ec s like imp o ed SOM. Adap ed
managemen decisions based on he MZs should be e alua ed in on- a m ials in u u e
esea ch.

Scien i ic publica ions wi hin he con ex o his wo k
63
Supplemen a y Ma e ial
Table S 1: Sampling da es ield da a and UAC campaigns. “X” ma ks i he pa ame e s was collec ed on his da e.
Field
Da e
Time
C op
Bio-
mass
Plan
heigh
Plan
Num-
be
Yield
P opo -
ion
VS
WC
SP
AD
Soil
Sampl-
ing
UAV
Ima-
ges
Powe
Weg
2020-
06-12
1.1
sample
Clo e -
g ass
X
X
X
Powe
Weg
2020-
07-10
2.
ha es
Clo e -
g ass
X
X
X
Powe
Weg
2020-
07-21
2.1
sample
Clo e -
g ass
X
X
X
Powe
Weg
2020-
07-31
2.2
sample
Clo e -
g ass
X
X
X
Powe
Weg
2020-
08-11
2.3
sample
Clo e -
g ass
X
X
Powe
Weg
2020-
08-27
2.4
sample
Clo e -
g ass
X
X
X
Powe
Weg
2020-
09-10
3.
ha es
Clo e -
g ass
X
X
X
Powe
Weg
2021-
04-09
Soil
sampling
Summe
spel
X
Powe
Weg
2021-
04-29
1.
sample
Summe
spel
X
X
Weed
co e
X
Powe
Weg
2021-
05-20
2.
sample
Summe
spel
X
X
Weed
co e
X
Powe
Weg
2021-
06-03
3.
sample
Summe
spel
X
X
Weed
co e
X
X
Powe
Weg
2021-
06-09
Soil
sampling
Summe
spel
X
Powe
Weg
2021-
06-24
4.
sample
Summe
spel
X + Ea
leng h
X
Weed
co e
X
X
Powe
Weg
2021-
08-11
Ha es
Summe
spel
X
X + Ea
leng h
X
Powe
Weg
2021-
08-31
Soil
sampling
Summe
spel
X
Kies-
schach
2021-
05-14
1.
ha es
Clo e -
g ass
X
X
X
X
X
Kies-
schach
2021-
06-10
Soil
sampling
Clo e -
g ass
X
Kies-
schach
2021-
06-14
1.1
sample
Clo e -
g ass
X
X
X
X
Kies-
schach
2021-
07-08
2.
ha es
Clo e -
g ass
X
X
X
X
Kies-
schach
2021-
07-29
2.1
sample
Clo e -
g ass
X
X
X
X
X
Kies-
schach
2021-
08-11
2.2
sample
Clo e -
g ass
X
X
X
X
X
Kies-
schach
2021-
08-26
2.3
sample
Clo e -
g ass
X
X
X
X
X
Kies-
schach
2021-
09-02
Soil
sampling
Clo e -
g ass
X
Kies-
schach
2021-
09-07
3.
ha es
Clo e -
g ass
X
X
X
X
X
X
Kies-
schach
2022-
04-25
1.
sample
Summe
spel
X
X
Weed
co e
X
X
Kies-
schach
2022-
05-12
2.
sample
Summe
spel
X
X
Weed
co e
X
X
Kies-
schach
2022-
06-10
3.
sample
Summe
spel
X
X
Weed
co e
X
X
Kies-
schach
2022-
06-21
Soil
sampling
Summe
spel
X
Kies-
schach
2022-
07-14
4.
sample
Summe
spel
X
X
Weed
co e
X
X
Kies-
schach
2022-
07-25
Ha es
Summe
spel
X
X + Ea
leng h
X
Weed
co e
Scien i ic publica ions wi hin he con ex o his wo k
64
Field
Da e
Time
C op
Bio-
mass
Plan
heigh
Plan
Num-
be
Yield
P opo -
ion
VS
WC
SP
AD
Soil
Sampl-
ing
UAV
Ima-
ges
Kies-
schach
2022-
09-01
Soil
sampling
Summe
spel
X
B eme
S aße
2022-
05-03
0.1
sample
Clo e -
g ass
X
X
X
X
X
B eme
S aße
2022-
05-18
1.
ha es
Clo e -
g ass
X
X
X
X
B eme
S aße
2022-
06-16
1.1
sample
Clo e -
g ass
X
X
X
X
X
B eme
S aße
2022-
06-29
1.2
sample
Clo e -
g ass
X
X
X
X
X
B eme
S aße
2022-
07-12
1.3
sample
Clo e -
g ass
X
X
X
X
X
B eme
S aße
2022-
07-28
2.
ha es
Clo e -
g ass
X
X
X
X
X
B eme
S aße
2022-
08-17
2.1
sample
Clo e -
g ass
X
X
X
X
X
B eme
S aße
2022-
08-31
2.2
sample
Clo e -
g ass
X
B eme
S aße
2022-
09-15
2.3
sample
Clo e -
g ass
X
X
X
X
X
B eme
S aße
2022-
09-29
3.
ha es
Clo e -
g ass
X
X
X
X
X
X
B eme
S aße
2022-
11-08
1.
sample
Win e
whea
X
X
X
B eme
S aße
2023-
03-30
2.
sample
Win e
whea
X
X
Weed
co e
X
X
B eme
S aße
2023-
04-03
Soil
sampling
Win e
whea
X
B eme
S aße
2023-
05-17
3.
sample
Win e
whea
X
X
Weed
co e
X
X
B eme
S aße
2023-
05-22
4.
sample
Win e
whea
X
X
Weed
co e
X
X
B eme
S aße
2023-
06-18
Ha es
Win e
whea
X
X + Ea
leng h
X
Weed
co e
X
X
Scien i ic publica ions wi hin he con ex o his wo k
65
Fig S 1. Indexes o inding bes numbe o clus e s (k) o uzzy C-means. A highe alue o Calinski-Ha abasz
index and he silhoue e index indica es a ma ching wi h he own clus e . A high pa i ion en opy indica es a uzzy
esul .
Fig S 2. Maps o he delinea ed o clo e -g ass zones (CGZ) o he single obse ed da es and o each ield.
Ligh blue is CGZ 1, da k blue is CGZ 2.
Scien i ic publica ions wi hin he con ex o his wo k
66
2.3 E ec s o mixed in e c opping on he ag onomic pa ame e s o
wo o ganically g own mal ing ba ley cul i a s (Ho deum ulga e) in
No hwes Ge many
Tobias Reu e 1, The ese B inkmeye 1, Johann Sch eibe 1, Valen in F eese1, Die e T au z1,
Insa Kühling2
1Osnab ück Uni e si y o Applied Sciences, Am K ümpel 31, 49090 Osnab ück, Ge many
2Kiel Uni e si y, He mann-Rodewald-S . 9, 24118 Kiel, Ge many
Abs ac
The ma ke o o ganic mal ing ba ley (Ho deum
ulga e L.) is g owing. One goal o p oducing mal ing
ba ley is o a ain a de ined p o ein con en . This is
challenging in o ganic a ming because nu ien
up ake is unp edic able. Since mal ing ba ley is quali y
sensi i e o high amoun s o a ailable ni ogen (N)
du ing he g ain illing phase, a i icial compe i ion o
N om a second c op wi hin a mixed in e c opping
sys em could help o limi N up ake du ing hese la e
g owing s ages. The objec i e o his s udy was o
e alua e he e ec s o in e c opping on he b ewing
quali y pa ame e s o mal ing ba ley wi hin he
amewo k o o ganic a ming. In a ield ial, wi h wo
sp ing ba ley cul i a s (Odilia, Ma he) as sole c ops
and in in e c opping ea men s o 4 in e media e
mixing a ios wi h camelina (Camelina sa i a), linseed
(Linum usi a issimum) and pea (Pisum sa i um), he
e ec s on yield and mal ing quali y we e obse ed.
The esul s om h ee g owing seasons on a s udy si e
in no h-wes e n Ge many showed an opposi e p o ein
esponse o mix u es wi h linseed and pea. Howe e ,
sole s ands o ba ley mos ly pe o med as well as he
mix u es. Fo mix u es wi h camelina, nei he yield no
quali y aspec s we e a ec ed. Compa ing he wo
cul i a s, he well-es ablished Ma he, om adi ional
con en ional selec ion, showed he be e o e all
pe o mance wi h a 13 % highe g ain yield, a 3 %
highe hec oli e weigh and 5 % highe p opo ion o
size ac ion >2.5 mm han Odilia, which was eleased
la e om an o ganic b eeding p og amme. Howe e ,
he yield componen analysis showed a consis en
yield de e mina ion o Odilia ega dless o
in e c opping pa ne o mixing a ios which migh
indica e i s sui abili y in polycul u es. The obse ed
a e age land equi alen a ios (LER) o mix u es wi h
linseed (1.04) and pea (1.13) showed he po en ial o
inc ease land-use e iciency bu we e lowe compa ed
o he mean LER ound in a ecen me a-analysis.
PICTURE CREDIT: The ese B inkmeye
(2017)
Ci a ion
Reu e , T., B inkmeye , T., Sch eibe , J.,
F eese, V., T au z, D., & Kühling, I. (2022).
E ec s o mixed in e c opping on he
ag onomic pa ame e s o wo o ganically
g own mal ing ba ley cul i a s (Ho deum
ulga e) in No hwes Ge many. Eu opean
Jou nal o Ag onomy, 134 (Janua y),
126470.
h ps://doi.o g/10.1016/j.eja.2022.126470
Keywo d
mixed c opping, c op mix u es, yield,
componen s, di e si y, oilseeds, o ganic,
ag icul u e, sus ainabili y, Eu ope
Au ho con ibu ions
T.R.: in es iga ion, o mal analysis, w i ing
– o iginal d a / e iew & edi ing. T.B.:
in es iga ion, me hodology. J.S:
in es iga ion. V.F.: in es iga ion. D.T.:
concep ualisa ion, esou ces, supe ision,
w i ing – e iew & edi ing. I.K.:
concep ualisa ion, me hodology,
alida ion, o mal analysis, isualisa ion,
w i ing – o iginal d a / e iew & edi ing,
supe ision.
Acknowledgemen s
This s udy was suppo ed by Osnab ück
Uni e si y o Applied Sciences.
Recei ed 31 Augus 2021; Recei ed in
e ised o m 18 Janua y 2022; Accep ed
21 Janua y 2022 1161-0301/© 2022
Else ie B.V. All igh s ese ed.
Recei ed 31 Augus 2021; Recei ed in e ised
o m 18 Janua y 2022; Accep ed 21 Janua y
2022 1161-0301/© 2022 Else ie B.V. All igh s
ese ed.
Gene al discussion
67
Chap e 3 Gene al discussion
Pic u e 3: Summe spel (T i icum aes i um subsp. Spel a) ial, 19 h July 2022.

Gene al discussion
68
3.1 In eg a ion o P ecision Fa ming and o ganic ag icul u e:
po en ial and challenges
The global ag icul u al sec o is con on ed wi h he dual challenge o sus aining ood
p oduc ion o an inc easing wo ld popula ion (Wille e al., 2019) while coping wi h a decline
in pe capi a ag icul u al land a ailabili y (FAO, 2024). This challenge is u he compounded
by signi ican losses in plan and animal biodi e si y (Geige e al., 2010; Wille e al., 2019).
Bo h P ecision Fa ming (PF) and o ganic ag icul u e o e possible solu ions, as o ganic
a ming enhances soil e ili y and p omo es biodi e si y, by p ohibi ing syn he ic e ilize s and
pes icides (Sande s e al., 2025). PF op imizes esou ce applica ion based on ield-speci ic
condi ions, he eby inc easing esou ce use e iciency (Amma e al., 2024). The in eg a ion o
PF in o o ganic ag icul u e can help o close he yield gap be ween o ganic and con en ional
a ming and u he inc ease he sus ainabili y.
Adop ion o P ecision Fa ming in o ganic sys ems
Resea ch ha di ec ly combines PF and o ganic a ming me hods is limi ed. Fo ins ance,
Loewen (2023) has es ed a iable a e seeding in o ganic con ex s, his app oach educed
he weed p essu e and inc eased ne e u ns. Whe eas si e-speci ic mechanical weeding has
been e alua ed p ima ily unde con en ional a ming condi ions (Be ge e al., 2024; Niemeye
e al., 2024). Addi ionally, Reumaux (2024) es ima ed wi hin- ield a ia ion o he si e-speci ic
applica ion o biogas diges a e e iliza ion in o ganic whea p oduc ion in Sweden. This hesis
is among he ew s udies ha in eg a e bo h app oaches o PF and o ganic a ming, he eby
subs an ially enhancing ou unde s anding o sus ainable ag icul u e
This hesis esea ch used comme cial machine y, such as Unmanned ae ial ehicles
(UAV), hoes and sowing machines. The concep s p esen ed can be applied in p ac ical
ag icul u e wi h comme cial machine y, al hough modi ica ions a e needed o allow he hoe o
au oma ically adjus based on he weed map. Cu en ly, no da a is a ailable ha speci ically
examines PF adop ion on o ganic a ms in Ge many. Howe e , su eys conduc ed in Ge many
indica e ha app oxima ely 70 % o all a ms employ PF echnologies. Among hese, 95 % use
au oma ic s ee ing sys ems, 40 % ely on yield mapping, 45 % implemen a iable a e
e iliza ion and 33 % p ac ice a iable a e seeding (Sonn ag e al., 2022). In small-scale
a ming egions such as Ba a ia, only 21 % o a ms use digi al ield eco ds, 17 % employ
au oma ic s ee ing sys ems and 14 % use sa elli e-de i ed maps (Gab iel and Gando e ,
2023). Al hough hese su eys include o ganic a ms, hey did no analyse he speci ic e ec s
o o ganic managemen p ac ices on PF adop ion. Fac o s ha posi i ely in luence adop ion
a es include highe educa ion le els and la ge a m sizes (Sonn ag e al., 2022). Gi en ha
o ganic a ms a e ypically smalle (abou 51 ha) compa ed o con en ional a ms
(app oxima ely 66 ha; BMEL, 2023b), i can be expec ed ha PF adop ion in o ganic sys ems
may be close o he 21 % obse ed in Ba a ian small-scale a ms.
The adop ion a es o 21 % aligns wi h su eys on o ganic a ms in he Czech Republic,
Hunga y, Poland and Slo akia, whe e 15 o 26 % o a ms use P ecision Fa ming echnology.
The highes usage was ound in he Czech Republic (25 %) and Poland (26 %). The p ima y
mo i a ion o adop ing P ecision Fa ming echnology was cos sa ings, while he main
obs acle was he lack o inancial suppo (Pe o ic e al., 2025). Fo PF echnology o be
success ully adop ed, i mus deli e clea economic o ecological bene i s, be eliable and
use - iendly, while a me s mus also possess he necessa y knowledge o i s applica ion and
e alua ion (Moh and Kühl, 2021; Sonn ag e al., 2022).
Gene al discussion
69
The esul s p esen ed in his hesis indica e ha PF and Decision Suppo Sys ems (DSS)
bene i o ganic a ming by imp o ing esou ce u iliza ion—such as op imizing soil nu ien
managemen and p ese ing wild plan communi ies. Table 9 p o ides an o e iew o he
hypo heses es ed ac oss he h ee pee - e iewed pape s. O he eigh hypo heses, six we e
accep ed due o he ield ial esul s.
Table 9: Tes ed hypo hesises in he h ee pee - e iewed pape .
✓
: accep ed;

: ejec ed. SSMW: si e-speci ic
mechanical weed managemen . RWC: Rela i e Weed Co e . NDRE: No malized Di e ence Red Edge.
Hypo hesis
Resul
Si e-speci ic mechanical weeding o maize (Zea mays)
Hypo hesis 1: SSMW does no in luence maize yield compa ed o uni o m weeding.
✓
Hypo hesis 2: SSMW does no in luence weed biomass compa ed o uni o m weeding.
✓
Hypo hesis 3: SSMW based on RWC esul s in less ea ed a ea han uni o m weeding
and SSMW based on weed co e .
✓
Delinea ion o managemen zones in clo e -g ass o si e-speci ic managemen o subsequen c ops
Hypo hesis 4: High p oduc i i y zones du ing he clo e -g ass pe iod a e also highly
p oduc i e o he subsequen c op.
✓
Hypo hesis 5: NDRE images a e use ul o delinea ing managemen zones in clo e -
g ass ields.
✓
Hypo hesis 6: Highe clo e yields lead o highe ixed ni ogen (N) and, he e o e,
highe soil ni ogen con en (SNC) du ing bo h he clo e -g ass pe iod and he
subsequen c op.

E ec s o mixed in e c opping on he ag onomic pa ame e s o wo o ganically g own mal ing ba ley cul i a s
(Ho deum ulga e)
Hypo hesis 7: In e c opping ba ley wi h a non-legume pa ne can limi p o ein con en
and enhance mal ing quali y.
✓
Hypo hesis 8: O ganic a ie ies in e c opped wi h pa ne s ha e be e yield and quali y
pe o mance.

The si e-speci ic mechanical weeding o maize (Zea mays), examined in sec ion 2.1,
achie ed a educ ion in he ea ed a ea by 58 – 83 %, pa icula ly in he Rela i e Weed Co e
(RWC) ea men plo s (suppo ing hypo hesis 3). These sa ings we e ealized wi hou any
yield losses (suppo ing hypo hesis 1) o inc eases in weed biomass (suppo ing hypo hesis
2). This educ ion in he ea ed a ea can dec ease he isk o soil e osion and c op damage,
while also spa ing wild plan s by applying weeding only when a ha m ul h eshold is exceeded.
In sec ion 2.2, he delinea ion o managemen zones in clo e –g ass ields was
success ully ca ied ou using UAV-based NDRE images (suppo ing hypo hesis 5). Ac oss all
ields, highly p oduc i e sub- ields we e iden i ied du ing bo h he clo e –g ass pe iod and he
subsequen ce eal p oduc ion pe iod (suppo ing hypo hesis 4). Con a y o hypo hesis 6,
hese a eas did no exhibi an inc ease in soil ni ogen con en , e en hough highe clo e
yields we e obse ed.
Sec ion 2.3 desc ibes he in e c opping o mal ing ba ley (Ho deum ulga e). In e c opping
wi h a non-legume pa ne e ec i ely limi ed he p o ein con en (suppo ing hypo hesis 7),
whe eas in e c opping wi h pea (Pisum sa i um) inc eased p o ein con en . In con as o he
assump ion o hypo hesis 8, he con en ional a ie y ‘Ma he’ achie ed highe yields han he
o ganic a ie y ‘Odelia’ unde in e c opping condi ions.
Gene al discussion
70
In he ollowing discussion, he esul s o he h ee pape s a e examined in a b oade
con ex wi h espec o hei po en ial applica ions in sus ainable ag icul u al p ac ices.
Si e-speci ic mechanical weed managemen
The applica ion o si e-speci ic mechanical weed managemen , as desc ibed in sec ion
2.1, showed signi ican educed managed a ea o a ound -58 % o -83 %, his esul s a ied
be ween he yea s. These educ ions can enhance plan biodi e si y (Sowiński, 2023),
dec ease soil e osion isk (Sei z e al., 2019) and educe machine y wea by minimizing illage
ime (G y e al., 2019). Al hough ini ial cos s in mapping weeds and acqui ing specialized
machine y may be highe , he me hod can sa e money in he long e m h ough up o a 97 %
educ ion in uel consump ion and equipmen wea (G y e al., 2019).
Fo a able c ops such as maize and ce eals—which a e ypically hoed wo o h ee imes
pe g owing season— he bene i s o si e-speci ic weeding a e signi ican . Vege ables like
le uce (Lac uca sa i a), cabbage (B assica ole acea) o ca o s (Daucus ca o a), which equi e
mo e in ensi e egula ion due o lowe compe i i e abili y (Ge ha ds e al., 2025), may see
mul iplica i e economic and ag onomic bene i s. Recen ad ances, including senso -guided
in e - ow hoes, ha e success ully educed weed co e wi hin c op ows. Howe e , c op ow
iden i ica ion gene ally equi es slowe ope a ing speeds (app oxima ely 1 km h-1; Ge ha ds e
al., 2025). This limi a ion could be o e come by pa ially aising he hoe in a eas wi h low weed
co e and whe e egula ion is unnecessa y. Ano he oppo uni y is he deploymen o
au onomous weeding obo s—whe e labou cos s a e less c i ical compa ed o manned
ac o s (G iepen og and S ein, 2024). Field obo s also o e he po en ial o seed c ops in
uni o m pa e ns, ensu ing consis en spacing be ween plan s. This uni o mi y op imizes
nu ien and ligh accessibili y while acili a ing bo h in e - and in a- ow weeding, as he obo s
can ope a e in any di ec ion (Wegene e al., 2019).
Despi e po en ial u u e ad ancemen s in mechanical weeding echniques, his app oach
will always ca y he isk o causing plan inju ies (Machleb e al., 2020) and educing he
abundance o wild plan s (Sowiński, 2023). The esea ch p esen ed he e o e s aluable
insigh s in o DSS o si e-speci ic mechanical weeding, he eby mi iga ing some o hese
d awbacks. In pa icula , he Rela i e Weed Co e (RWC) app oach is ecommended because
i esul ed in he smalles weeded a ea. The RWC me ic akes in o accoun bo h weed and
c op co e , enabling an e ec i e es ima ion o c op compe i i eness.
Managemen zones and in e c opping in o ganic ields
Th ee o ganic ields we e success ully delinea ed in o managemen zones, based on
NDRE-Maps o clo e -g ass (Sec ion 2.2). Bo h he clo e -g ass and he subsequen ce eal
c ops exhibi ed signi ican di e ences in yield and quali y pa ame e s ac oss managemen
zones. This zona ion pe mi s mo e a ge ed managemen app oaches. Fe ilize can be
applied based on he yield po en ial, as hey a e limi ed in o ganic a ming and a e a eason
o he yield gap o o ganic ag icul u e (Dö ing and Neuho , 2021). As Reumaux (2024)
showed, si e-speci ic biogas diges a e can imp o e he yield and p o ein con en o o ganic
whea .
Fu he mo e, managemen zones can se e as he basis o selec ing c op a ie ies and
in e c opping s a egies. In e c opping wi h legumes, o example, can enhance p o ein con en
(Sec ion 2.3) and mi iga e he e ec s o nu ien limi a ions in low-yield a eas. In e c opping
has been epo ed o o e g ea e yield s abili y han sole c opping, mi iga ing bo h spa ial and
empo al a iabili y (Munz e al., 2023; Raseduzzaman and Jensen, 2017). This app oach is
Gene al discussion
71
pa icula ly bene icial in uns able ield zones—o en ound in dep essions wi h a highe isk o
wa e logging—whe e c ops exhibi lowe yields du ing high-p ecipi a ion yea s and imp o ed
yields du ing d y pe iods. In e se, a eas wi h low wa e -holding capaci y bene i wi h highe
p ecipi a ion and expe ience low yields unde d y condi ions (Maes ini and Basso, 2018a;
McEn ee e al., 2020). Addi ionally, win e c ops, which gene ally possess be e -de eloped
oo sys ems, a e less sensi i e o d y summe condi ions compa ed o summe c ops
(He nández-ochoa e al., 2025). Resea ch in he midwes e n USA indica es ha app oxima ely
18 % o ields a e uns able, wi h unp edic able annual a iabili y. In such cases, in e c opping
can s abilize yields as di e en c op species complemen one ano he by compensa ing o
losses due o hei a ying equi emen s (Raseduzzaman and Jensen, 2017). This dynamic
adjus men o c op composi ion has also been e med “ecological p ecision a ming” (Jensen
e al., 2015). Munz e al. (2023) ecommends adjus ing he legume sowing a io based on soil
condi ions o imp o e hei compe i i eness agains ce eals. Inc eased legume p opo ions a e
bene icial in low ni ogen a eas.
Ano he app oach in ol es iden i ying a eas wi h high weed p essu e. The weed
ecogni ion could be conduc ed wi h UAV-images desc ibed in sec ion 2.1. In hese zones,
c op mix u es and species ha exhibi apid ea ly g ow h and p o ide dense g ound co e can
indi ec ly supp ess weed popula ions. In con as , egions wi h lowe weed p essu e migh
bene i om sowing species and mix u es op imized o highe yields.
3.2 Va ia ion o ag icul u al pa ame e s be ween yea s and si es
In all ials, signi ican annual di e ences we e obse ed. In pape 1 (Sec ion 2.1) on si e-
speci ic mechanical weeding, a maize yield o 1879 g m-2 was eco ded in 2021 compa ed o
533 g m-2 in 2022. In he con ol ea men , weed biomass was wi h 13.6 g m-2 in 2021 lowe
as in 2022 wi h 58.8 g m-2. The educ ion in ea ed a ea o he si e-speci ic ea men s a ied
widely om 0 % o 100 %, esul ing in a e age sa ings o 58 % in 2021 and 83 % in 2022.
Fo he delinea ion o managemen zones epo ed in pape 2 (Sec ion 2.2), ials we e
conduc ed ac oss di e en si es, whe e seasonal di e ences we e obse ed in addi ion o
in e - ield a ia ions. The clo e –g ass yield a ied conside ably: 1320.5 g m-2 a Powe Weg
(2020), 4386.6 g m-2 a Kiesschach (2021) and 983.7 g m-2 a B eme S aße (2022). Simila ly,
ce eal yields di e ed among si es, wi h yields o 158.8 g m-2 a Powe Weg (2021), 141.2 g m-2
a Kiesschach (2022) and 287.6 g m-2 a B eme S aße (2023).
In pape 3 (Sec ion 2.3), an in e c opping ial e ealed ha ba ley yields in sole s ands
a ied o e he yea s: 142.8 g m-2; 113.5 g m-2; and 50.0 g m-2 in 2017, 2018 and 2019,
espec i ely. The p o ein con en exhibi ed an annual pa e n, wi h alues o 9.6 % in 2017,
10.4 % in 2018 and 8.5 % in 2019.
Annual di e ences ha e been documen ed in o he s udies on si e-speci ic applica ions as
well. Co i e al. (2023) epo ed a ia ions in yield and ni ogen use e iciency in ials o si e-
speci ic manu e applica ion. The simula ions o si e-speci ic he bicide applica ion simila ly
showed a ying e iciencies be ween yea s (Maillo e al., 2023). Tempo al and spa ial
a iabili y was an icipa ed; he e o e, all ials we e epea ed o e wo o h ee yea s.
Ne e heless, his inhe en a iabili y makes i challenging o accu a ely analyse he e ec s o
PF echnology.
Tempo al a ia ion
Tempo al a ia ion in hese ials is p ima ily a ibu ed o wea he di e ences. Especially
d ough s ess can a y dynamically om yea o yea (Johnen e al., 2014). Du ing he g owing
Summa y
78
Summa y
The global ag icul u e is acing he challenge o p o iding enough ood o a g owing
popula ion while add essing declining biodi e si y. O ganic a ming bene i s he en i onmen
bu ypically yields 20 % less han con en ional me hods. P ecision Fa ming (PF) can help o
close his gap by ailo ing inpu s like e ilize s o a ia ions wi hin ields.
This disse a ion p esen s a se ies o ials— epo ed in h ee pee - e iewed pape s—
aimed a inc easing he p oduc i i y and sus ainabili y o o ganic a ming. Conduc ed nea
Osnab ück in no hwes e n Ge many, hese ield s udies employed ei he andomized block
designs o on- a m ials o e wo o h ee yea s. In hese ials, PF echnologies such as
Unmanned ae ial ehicles (UAVs) and Decision Suppo Sys ems (DSS) we e in eg a ed wi h
con en ional o ganic managemen p ac ices including mechanical weeding, c op o a ion and
in e c opping.
In pape 1 (Sec ion 2.1), he applica ion o si e-speci ic mechanical weeding based on UAV
image y educed he managed a ea by 58 % o 83 % wi hou comp omising yield o in luencing
weed biomass (Suppo ing hypo hesis 1 & 2). Two decision-suppo s a egies we e compa ed
agains a uni o m weeding app oach. Among hese, he Rela i e Weed Co e eme ged as a
p omising me ic (Suppo ing hypo hesis 3), since i conside s bo h maize and weed co e , he
c op compe i i eness can be es ima ed and he hoed a ea was minimized. This app oach no
only educes inpu use bu can also enhance plan biodi e si y, dec ease soil e osion and
educe machine y wea , o e ing long- e m cos sa ings despi e highe ini ial mapping and
equipmen cos s.
Pape 2 (Sec ion 2.2) ocuses on delinea ing managemen zones wi hin clo e –g ass ields
using Vege a ion Indices. Th ee o ganic ields we e success ully pa i ioned in o wo dis inc
managemen zones each. The used NDRE (No malized Di e ence Red Edge)-Maps and uzzy
C-means clus e ing algo i hms we e app op ia ed o his ask (Suppo ing hypo hesis 5).
Signi ican di e ences in clo e –g ass pa ame e s and in he yield o subsequen ce eal c ops
we e obse ed be ween zones (Suppo ing hypo hesis 4). These di e ences a e la gely
a ibu able o a ia ions in soil p ope ies and opog aphy. This zona ion suppo s a ge ed
managemen s a egies, such as si e-speci ic e iliza ion, which is pa icula ly impo an in
o ganic sys ems whe e e ilize a ailabili y is limi ed. In con as o hypo hesis 6 no signi ican
di e ence in soil ni ogen con en be ween he zones we e obse ed, despi e di e ences in
clo e p opo ion and biomass.
The inal pape (Sec ion 2.3) compa es he pe o mance o mal ing ba ley (Ho deum
ulga e) in e c opped wi h a ious pa ne s. T ials e ealed ha in e c opping ba ley wi h peas
(Pisum sa i um) imp o ed p o ein con en and land-use e iciency, whe eas in e c opping wi h
linseed (Linum usi a issimum) educed p o ein le els (Suppo ing hypo hesis 7); ne e heless,
sole s ands o ba ley gene ally pe o med simila ly o mixed sys ems. These esul s in o m he
selec ion o op imal c op mix u es acco ding o di e en managemen zones and unde sco e
he impo ance o in e c opping in s abilizing yields in a eas wi h a iable e ili y. The o ganic
ba ley a ie y did no show bene i s in ega d o yield and quali y compa ed o he con en ional
a ie y, he e o e he hypo hesis 8 can be ejec ed.

Summa y
79
Subs an ial empo al a iabili y was obse ed ac oss all ials— o example, maize yields
anged om 533 o 1879 g m-2, clo e –g ass yields a ied be ween 984 g m-2 and 4387 g m-2
and ce eal yields anged om 50 o 143 g m-2. Such a iabili y, d i en la gely by wea he
condi ions, along wi h di e ences be ween si es, highligh s he impo ance o mul i-yea and
mul i-si e ials o obus e alua ion o PF echnologies. Finally, he in eg a ion o on- a m ials
wi h a co-design app oach in ol ing s akeholde s is c ucial o enhancing he accep ance and
p ac ical implemen a ion o inno a i e managemen p ac ices.
Recen ad ances in da a analy ics and ield obo ics o e u he po en ial o ans o ming
ag icul u al p ac ices owa d biodi e si y-based sys ems. Fu u e app oaches may in ol e
subdi iding ields in o smalle managemen uni s, which a e in eg a ed wi h landscape
elemen s o p omo e biodi e si y.
This hesis is among he ew s udies ha in eg a e PF- echnologies wi h o ganic a ming
p ac ices, he eby signi ican ly enhancing ou unde s anding o sus ainable ag icul u e. The
esul s indica e ha he combined use o PF and DSS bene i s o ganic a ming by imp o ing
esou ce u iliza ion. This is achie ed by applying a ge ed, si e-speci ic weeding in a eas whe e
i is necessa y and by delinea ing ields in o managemen zones o apply e iliza ion and
seeding based on sub- ield condi ions.
Zusammen assung (Ge man summa y)
80
Zusammen assung (Ge man summa y)
Die Landwi scha s eh o de He aus o de ung gleichzei ig ü E näh ungssiche hei
und ein in ak es Ökosys em zu so gen. De ökologische Landbau gil als umwel e ägliche ,
e n e jedoch im Schni 20 % wenige als de kon en ionelle Landbau. P ecision Fa ming (PF)
bie e die Möglichkei die E agslücke du ch e izien e e Ressou cen e eilung basie en au
Ungleichhei en im Feld zu e inge n.
In diese Disse a ion we den d ei pee - e iewed A ikel o ges ell , die das Ziel e olgen,
die P oduk i i ä und Nachhal igkei des ökologischen Landbaus zu e besse n. PF-
Anwendungen wie de Einsa z on Unmanned ae ial ehicles (UAVs) und
En scheidungsun e s ü zungssys emen (englisch Decision Suppo Sys em; DSS) wu den mi
klassischen Me hoden des ökologischen Landbaus, wie mechanische Beik au egulie ung,
Leguminosenanbau und Mischkul u anbau, kombinie . Dazu wu den im Raum Osnab ück in
No dwes deu schland e schiedene Ve suche du chge üh , die en wede als andomisie e
Block e suche ode als On- a m-Ve suche ealisie wu den. Jede Ve such wu de in zwei
ode d ei Ve suchsjah en wiede hol .
A ikel 1 (Abschni 2.1) be ass sich mi de eil lächenspezi ischen mechanischen
Beik au egulie ung. Die E kennung de Beik äu e e olg e anhand on UAV-Au nahmen. Mi
diesem Ve ah en konn e die bea bei e e Fläche um 58 % bis 83 % eduzie we den, ohne
dass E ags e lus e au a en ode die Beik au biomasse beein luss wu de, was Hypo hese
1 & 2 beleg . Von den un e such en En scheidungss a egien üh e de ela i e
Unk au bedeckungsg ad (englisch: Rela i e Weed Co e ), de sowohl den Beik au - als auch
den Maisbedeckungsg ad be ücksich ig , zu de ge ings en bea bei e en Fläche (Bes ä igung
Hypo hese 3). Die Reduzie ung de bea bei e en Fläche du ch diese Me hode bie e das
Po enzial, Wildp lanzen zu schü zen, E osion zu e hinde n und den Ve schleiß on
Maschinen zu e inge n.
A ikel 2 (Abschni 2.2) besch eib , wie mi hil e des d ohnenbasie en Vege a ionsindexes
(No malized Di e ence Red Edge), Kleeg as lächen in zwei Managemen zonen un e eil
we den konn en. Diese Ein eilung wu de e olg eich an d ei Flächen du chge üh , was
Hypo hese 5 bes ä ig . Diese Zonen un e schieden sich signi ikan im Kleeg ase ag und -
zusammense zung. In den anschließenden Ge eidekul u en konn en zudem Di e enzen in
E ag und Quali ä zwischen den Zonen es ges ell we den (Bes ä igung Hypo hese 4). Diese
Un e schiede sind haup sächlich au Va ia ionen in Bodenbescha enhei und Topog a ie
zu ückzu üh en. Die so e mi el en Managemen zonen können als G undlage ü
eil lächenspezi ische Managemen maßnahmen dienen, e wa eine geziel en Düngung,
welche ü den ökologischen Landbau besonde s ele an is , da Düngemi el nu beg enz
o handen sind. En gegen de Hypo hese 6, gab es jedoch keine Un e schiede zwischen den
Zonen in Bezug au mine alischen S icks o gehal im Boden.
Im le z en A ikel (Abschni 2.3) wu den B auge s e (Ho deum ulga e) im
Gemengeanbau mi e schiedenen Mischungspa ne n (E bse, Pisum sa i um; Leindo e ,
Linum usi a issimum; und Öllein, Camelina sa i a) un e such . De Gemengeanbau mi E bse
(Pisum sa i um) e höh e den P o eingehal und die Landnu zungse izienz, wäh end die
Mischung mi Leindo e (Linum usi a issimum) den P o eingehal e inge e (Bes ä igung de
Hypo hese 7). Die Reinsaa on B auge s e (Ho deum ulga e) e ziel e hinsich lich des
E ages und de Quali ä ähnliche E gebnisse wie die Mischkul u en. Diese E kenn nisse
können G undlage ü die Kul u auswahl in den un e schiedlichen Managemen zonen sein.
Insbesonde e in e agsschwachen Be eichen können Gemenge mi E bse (Pisum sa i um)
Zusammen assung (Ge man summa y)
81
zu eine Ve besse ung de E agss abili ä bei agen, da in Gemengeanbau, die
Pa ne kul u en sich e gänzen und E agsaus älle kompensie en können. Dies p ädes inie
den Gemengeanbau ü Be eichen mi schwankende E agspo en ial. Die ökologische So e
zeig en keine Vo eil gegenübe de kon en ionellen So e, wede in Bezug au E ag noch
Quali ä , was zu Ablehnung on Hypo hese 8 üh e.
In allen Ve suchen wu den e hebliche Schwankungen zwischen den Jah en es ges ell :
So a iie e de Maise ag zwischen 533 und 1879 g m-2, de Kleeg ase ag zwischen 984
und 4387 g m-2 und de Ge eidee ag zwischen 50 und 143 g m-2. Diese Va ia ionen sind
haup sächlich au We e un e schiede zu ückzu üh en – ein Be und, de auch in ande en
Un e suchungen bes ä ig wu de. Da übe hinaus wu den g oße Un e schiede zwischen den
Flächen es ges ell , was eben alls in wei e en Feld e suchen es ges ell wu de. Dies
un e s eich die Bedeu ung on meh jäh igen Ve suchen an e schiedenen S ando en ü
eine obus e E alua ion on PF-Technologien. Fü die Akzep anz und Implemen ie ung on
PF und inno a i en Anbausys emen is es wich ig S akeholde pa izipa i e in die Planung und
Du ch üh ung on On- a m-Ve suchen mi einzubeziehen.
Die ak uellen En wicklungen in de Da enanalyse und Feld obo ik ebnen den Weg zu eine
biodi e si ä sbasie en Landwi scha . Zukün ige Ag a sys eme könn en so ges al e we den,
dass g oße Flächen in kleine e Bewi scha ungseinhei en mi ähnlichem E agspo enzial
un e eil und en sp echend ih en spezi ischen Bedingungen bewi scha e we den. Dabei wi d
nich nu das einzelne Feld be ach e , sonde n die gesam e Ag a landscha , welche du ch
Landscha selemen e wie Hecken und Bäume mi einande e bunden we den. Ein solches
Sys em weis komplexe Wechselwi kungen au und s ell hohe An o de ungen an das
Managemen . DSS können dabei hel en, diese Zusammenhänge zu e s ehen und undie e
En scheidungen zu e en.
Die hie o ges ell e Disse a ion gehö zu den wenigen A bei en, die PF-Technologien
im ökologischen Landbau anwenden. Die du chge üh en Ve suche zeigen, dass diese
Ansä ze dazu bei agen, die Landwi scha nachhal ige zu ges al en und Ressou cen
e izien e zu nu zen. So konn e eil lächenspezi ische Beik au egulie ung die behandel e
Fläche signi ikan eduzie en, wäh end die Ein eilung de Flächen in Managemen zonen eine
undie e G undlage ü s ando spezi ische Düngung und Aussaa bie e .
Re e ences
82
Re e ences
Adeux, G., Vie en, E., Ca lesi, S., Bà be i, P., Munie -Jolain, N., Co deau, S., 2019. Mi iga ing
c op yield losses h ough weed di e si y. Na . Sus ain. 2, 1018–1026.
h ps://doi.o g/h ps://doi.o g/10.1038/s41893-019-0415-y
Aksakal, E.L., Ba ik, K., Angin, I., Sa i, S., Islam, K.R., 2019. Spa io- empo al a iabili y in
physical p ope ies o di e en ex u ed soils unde simila managemen and semi-a id
clima ic condi ions. Ca ena 172, 528–546. h ps://doi.o g/10.1016/j.ca ena.2018.09.017
Ali, A., S eibig, J.C., And easen, C., 2013. Yield loss p edic ion models based on ea ly
es ima ion o weed p essu e. C op P o . 53, 125–131.
h ps://doi.o g/10.1016/j.c op o.2013.06.010
Ali, A., S eibig, J.C., Ch is ensen, S., And easen, C., 2015. Image-based h esholds o weeds
in maize ields. Weed Res. 55, 26–33. h ps://doi.o g/10.1111/w e.12109
Allmendinge , A., Spae h, M., Saile, M., Pe eina os, G.G., Ge ha ds, R., 2024. Ag onomic and
Technical E alua ion o He bicide Spo Sp aying in Maize Based on High-Resolu ion
Ae ial Weed Maps—An On-Fa m T ial. Plan s 13. h ps://doi.o g/10.3390/plan s13152164
Amma , E.E., Aziz, S.A., Zou, X., Elmas y, S.A., Ghosh, S., Khala , B.M., EL-She shaby, N.A.,
Tou ky, G.F., AL-Fa ga, A., Khan, A.N., Abdelha eez, M.M., Younis, F.E., 2024. An in-
dep h e iew on he concep o digi al a ming. En i on. De . Sus ain.
h ps://doi.o g/h ps://doi.o g/10.1007/s10668-024-05161-9
A bei sg uppe Boden (Soil wo king G oup), 2024. Bodenkundliche Ka ie anlei ung KA6, 6.
Edi ion. ed. Schweize ba Science Publishe s, S u ga , Ge many.
Azimi, S., Kau , T., Gandhi, T.K., 2021. A deep lea ning app oach o measu e s ess le el in
plan s due o Ni ogen de iciency. Measu emen 173, 108650.
h ps://doi.o g/10.1016/j.measu emen .2020.108650
Bai, Z., Caspa i, T., Ruipe ez, M., Ba jes, N.H., Mäde , P., Bünemann, E.K., Goede, R. De,
B ussaa d, L., Xu, M., Fe ei a, C.S.S., Rein am, E., Fan, H., Miheličh, R., Gla an, M.,
Tó h, Z., 2018. E ec s o ag icul u al managemen p ac ices on soil quali y: A e iew o
long- e m expe imen s o Eu ope and China. Ag ic. Ecosys . En i on. 265, 1–7.
h ps://doi.o g/10.1016/j.agee.2018.05.028
Balasund am, S.K., Shamshi i, R.R., S idha a, S., Rizan, N., 2023. The Role o Digi al
Ag icul u e in Mi iga ing Clima e Change and Ensu ing Food Secu i y: An O e iew.
Sus ainabili y 15, 5325. h ps://doi.o g/10.3390/SU15065325
Ball, B.C., C aw o d, C.E., 2009. Mechanical weeding e ec s on soil s uc u e unde ield
ca o s (Daucus ca o a L.) and beans (Vicia aba L.). Soil Use Manag. 25, 303–310.
h ps://doi.o g/10.1111/j.1475-2743.2009.00226.x
Banniza, S., Vandenbe g, A., 2003. The in luence o plan inju y on de elopmen o
Mycosphae ella pinodes in ield pea. Can. J. Plan Pa hol. 25, 304–311.
h ps://doi.o g/10.1080/07060660309507083
Bà be i, P., Bu gio, G., Dinelli, G., Moonen, A.C., O o, S., Vazzana, C., Zanin, G., 2010.
Func ional biodi e si y in he ag icul u al landscape: Rela ionships be ween weeds and
a h opod auna. Weed Res. 50, 388–401. h ps://doi.o g/10.1111/j.1365-
3180.2010.00798.x
Ba e h, G., Lussem, U., Menne, J., Hollbe g, J., Schellbe g, J., 2019. Po en ial o non-
calib a ed UAV-based RGB image y o o age moni o ing: Case s udy a he engen long-
e m g assland expe imen ( ge), Ge many. In . A ch. Pho og amm. Remo e Sens. Spa .
In . Sci. - ISPRS A ch. 42, 203–206. h ps://doi.o g/10.5194/isp s-a chi es-XLII-2-W13-
203-2019
Re e ences
83
Ba asso, C., K üge , R., El ne , A., Co d, A.F., 2024. Mapping indica o species o sege al
lo a o esul -based paymen s in a able land using UAV image y and deep lea ning.
Ecol. Indic. 169, 112780. h ps://doi.o g/10.1016/j.ecolind.2024.112780
Ba e o, A., Ispizua Yama i, F.R., Va elmann, M., Paulus, S., Mahlein, A.K., 2023. Disease
Incidence and Se e i y o Ce cospo a Lea Spo in Suga Bee Assessed by Mul ispec al
Unmanned Ae ial Images and Machine Lea ning. Plan Dis. 107, 188–200.
h ps://doi.o g/10.1094/PDIS-12-21-2734-RE
Ba zin, R., Pa hak, R., Lo i, H., Va co, J., Bo a, G.C., 2020. Use o UAS mul ispec al image y
a di e en physiological s ages o yield p edic ion and inpu esou ce op imiza ion in
co n. Remo e Sens. 12. h ps://doi.o g/10.3390/RS12152392
Basso, B., An le, J., 2020. Digi al ag icul u e o design sus ainable ag icul u al sys ems. Na .
Sus ain. 3, 254–256. h ps://doi.o g/10.1038/s41893-020-0510-0
Bedoussac, L., Jou ne , É.-P., Hauggaa d-Nielsen, H., Naudin, C., Co e-Hellou, G., P ieu ,
L., Jensen, E.S., Jus es, E., 2014. Eco- unc ional In ensi ica ion by Ce eal-G ain Legume
In e c opping in O ganic Fa ming Sys ems o Inc eased Yields, Reduced Weeds and
Imp o ed G ain P o ein Concen a ion, in: O ganic Fa ming, P o o ype o Sus ainable
Ag icul u es. Sp inge Ne he lands, Do d ech , pp. 47–63. h ps://doi.o g/10.1007/978-94-
007-7927-3_3
Bedoussac, L., Jou ne , E.P., Hauggaa d-Nielsen, H., Naudin, C., Co e-Hellou, G., Jensen,
E.S., P ieu , L., Jus es, E., 2015. Ecological p inciples unde lying he inc ease o
p oduc i i y achie ed by ce eal-g ain legume in e c ops in o ganic a ming. A e iew.
Ag on. Sus ain. De . h ps://doi.o g/10.1007/s13593-014-0277-7
Behe a, S.K., Ma hu , R.K., Shukla, A.K., Su esh, K., P akash, C., 2018. Spa ial a iabili y o
soil p ope ies and delinea ion o soil managemen zones o oil palm plan a ions g own in
a ho and humid opical egion o sou he n India. Ca ena 165, 251–259.
h ps://doi.o g/10.1016/j.ca ena.2018.02.008
Beikü ne , M., Kühling, I., Ve ga a-He nandez, M.E., B oll, G., T au z, D., 2024. Impac o
mechanical weed con ol on soil N dynamics, soil mois u e, and c op yield in an o ganic
c opping sequence. Nu . Cycl. Ag oecosys ems. h ps://doi.o g/10.1007/s10705-024-
10370-9
Beillouin, D., Ben-A i, T., Malézieux, E., Seu e , V., Makowski, D., 2021. Posi i e bu a iable
e ec s o c op di e si ica ion on biodi e si y and ecosys em se ices. Glob. Chang. Biol.
27, 4697–4710. h ps://doi.o g/10.1111/gcb.15747
Be deni, D., Tu ne , A., G ayson, R.P., Llanos, J., Holden, J., Fi bank, L.G., Lappage, M.G.,
Hun , S.P.F., Chapman, P.J., Hodson, M.E., Helgason, T., Wa , P.J., Leake, J.R., 2021.
Soil quali y egene a ion by g ass-clo e leys in a able o a ions compa ed o pe manen
g assland: E ec s on whea yield and esilience o d ough and looding. Soil Tillage Res.
212. h ps://doi.o g/10.1016/j.s ill.2021.105037
Be ge, T.W., U dal, F., To p, T., And easen, C., 2024. A Senso -Based Decision Model o
P ecision Weed Ha owing. Ag onomy 14, 1–14.
h ps://doi.o g/10.3390/ag onomy14010088
Be y, P., Syl es e -B adley, R., Philipps, L., Ha ch, D., Cu le, S., Rayns, F., Gosling, P., 2002.
Is he p oduc i i y o o ganic a ms es ic ed by he supply o a ailable ni ogen? Soil Use
Manag. 18, 248–255. h ps://doi.o g/10.1079/SUM2002129
Blank, L., Rozenbe g, G., Ga ni, R., 2023. Spa ial and empo al aspec s o weeds dis ibu ion
wi hin ag icul u al ields – A e iew. C op P o . 172.
h ps://doi.o g/10.1016/j.c op o.2023.106300
BMEL, 2023a. Bio-S a egie 2030 Na ionale S a egie ü 30 P ozen ökologische Land- und
Lebensmi elwi scha bis 2030. Bundesminis e ium ü E näh ung und Landwi scha .

Re e ences
84
BMEL, 2023b. Visualisie ung de S uk u da en zum ökologischen Landbau in Deu schland.
Bundesminis e ium ü E näh ung und Landwi scha [WWW Documen ]. URL
h ps://bmel-s a is ik.de/landwi scha /oekologische -landbau
BMEL, 2010. Ge eideeinhei enschlüssel Übe a bei ung. Bundesminis e ium ü E näh ung
und Landwi scha .
Boenecke, E., Ueck, E., Ruehlmann, J., G uendling, R., F anko, U., 2018. De e mining he
wi hin- ield yield a iabili y om seasonally changing soil condi ions. P ecis. Ag ic. 19,
750–769. h ps://doi.o g/10.1007/s11119-017-9556-z
Bölenius, E., S enbe g, B., A idsson, J., 2017. Wi hin ield ce eal yield a iabili y as a ec ed
by soil physical p ope ies and wea he a ia ions – A case s udy in eas cen al Sweden.
Geode ma Reg. 11, 96–103. h ps://doi.o g/10.1016/j.geod s.2017.11.001
BÖLW, 2024. B anchen epo 2024. Bund Ökologische Lebensmi elwi scha e.V.
Boo h, J.C., Mccall, D.S., Sulli an, D., Askew, S.A., Koche sbe ge , K., 2021. In es iga ing
a ge ed sp ing dead spo managemen ia ae ial mapping and p ecision-guided ungicide
applica ions. C op Sci. 3134–3144. h ps://doi.o g/10.1002/csc2.20623
B eiman, L., 2001. Random Fo es s. Mach. Lea n. 45, 5–32.
h ps://doi.o g/h ps://doi.o g/10.1023/A:1010933404324
B eunig, F.M., Gal ão, L.S., Dalagnol, R., Dau e, C.E., Pa aga, A., San i, A.L., Della Flo a,
D.P., Chen, S., 2020. Delinea ion o managemen zones in ag icul u al ields using co e –
c op biomass es ima es om Plane Scope da a. In . J. Appl. Ea h Obs. Geoin . 85.
h ps://doi.o g/10.1016/j.jag.2019.102004
B i o Da Sil a, L.E., Mel on, N.M., Wunsch, D.C., 2020. Inc emen al Clus e Validi y Indices o
Online Lea ning o Ha d Pa i ions: Ex ensions and Compa a i e S udy. IEEE Access 8,
22025–22047. h ps://doi.o g/10.1109/ACCESS.2020.2969849
B ooke , R.W., Benne , A.E., Cong, W.-F.F., Daniell, T.J., Geo ge, T.S., Halle , P.D., Hawes,
C., Ianne a, P.P.M.M., Jones, H.G., Ka ley, A.J., Li, L., McKenzie, B.M., Pakeman, R.J.,
Pa e son, E., Schöb, C., Shen, J., Squi e, G., Wa son, C.A., Zhang, C., Zhang, F., Zhang,
J., Whi e, P.J., 2015. Imp o ing in e c opping: A syn hesis o esea ch in ag onomy, plan
physiology and ecology. New Phy ol. 206, 107–117. h ps://doi.o g/10.1111/nph.13132
B ooke , R.W., Pakeman, R.J., Adam, E., Ban ield-Zanin, J.A., Be elsen, I., Bickle , C., Fog-
Pe e sen, J., Geo ge, D., New on, A.C., Rubiales, D., Ta ole i, S., Villegas-Fe nández,
Á.M., Ka ley, A.J., 2024. Posi i e e ec s o in e c op yields in a ms om ac oss Eu ope
depend on ain all, c op composi ion, and managemen . Ag on. Sus ain. De . 44.
h ps://doi.o g/10.1007/s13593-024-00968-2
BSA, 2021. Besch eibende So enlis e Ge eide, Mais Öl- und Fase p lanzen Leguminosen
Rüben Zwischen üch e. Fede al Plan Va ie y O ice (Bundesso enam ), Hanno e .
Buladaco II, M.S., Tandugon, H.M.F., Bunquin, M.A.B., Sanchez, P.B., Bugia, S.A.C., Yales,
N.A.P., Casacop, S.M., 2024. Mapping and Assessmen o Wi hin-Field Spa ial Va iabili y
o Soil pH, Elec ical Conduc i i y, and Pa icle Size Dis ibu ion o Delinea e Managemen
Zones. Ecol. Eng. En i on. Technol. 25, 75–86.
Buman, T., 2013. Oppo uni y now: In eg a e conse a ion wi h p ecision ag icul u e. J. Soil
Wa e Conse . 68, 96–98. h ps://doi.o g/10.2489/jswc.68.4.96A
Ca o , M., Godino , O., Le Cad e, E., 2022. Biodi e si y-based c opping sys ems: A long- e m
pe spec i e is necessa y. Sci. To al En i on. 838, 156022.
h ps://doi.o g/10.1016/j.sci o en .2022.156022
Cas aldi, F., Pelosi, F., Pascucci, S., Casa, R., 2017. Assessing he po en ial o images om
unmanned ae ial ehicles (UAV) o suppo he bicide pa ch sp aying in maize. P ecis.
Ag ic. 18, 76–94. h ps://doi.o g/10.1007/s11119-016-9468-3
Re e ences
85
Chambe s, J.M., F eeny, A., Heibe ge , R.M., 1992. Analysis o a iance; Designed
Expe imen s, in: Chambe s, J.M., Has ie, T.J. (Eds.), S a is ical Models in S. Wadswo h
& B ooks, Cole.
Chan e, G.R., 2020. Decision Suppo Sys ems o Weed Managemen . Sp inge -Ve lag,
Cham,. h ps://doi.o g/10.1007/978-3-030-44402-0
Cheng, E., Zhang, B., Peng, D., Zhong, L., Yu, L., Liu, Y., Xiao, C., Li, C., Li, X., Chen, Y., Ye,
H., Wang, H., Yu, R., Hu, J., Yang, S., 2022. Whea yield es ima ion using emo e sensing
da a based on machine lea ning app oaches. F on . Plan Sci. 13, 1–16.
h ps://doi.o g/10.3389/ pls.2022.1090970
Cimpoiaşua, M.O., Ku as, O., P idmo e, T., Mooney, S.J., 2020. Po en ial o geoelec ical
me hods o moni o oo zone p ocesses and s uc u e : A e iew. Geode ma 365.
h ps://doi.o g/h ps://doi.o g/10.1016/j.geode ma.2020.114232
Co i, M., Ca alli, D., P icca, N., Fe è, C., Comolli, R., Ma ino, P., Da ide, G., El, A., 2023.
Si e ‑ speci ic ecommenda ions o ca le manu e ni ogen and u ea o silage maize. Nu .
Cycl. Ag oecosys ems 127, 155–169. h ps://doi.o g/10.1007/s10705-023-10302-z
Cowden, R.J., Shah, A.N., Lehmann, L.M., Kiæ , L.P., Hen iksen, C.B., Ghaley, B.B., 2020.
Ni ogen e ilize e ec s on pea–ba ley in e c op p oduc i i y compa ed o sole c ops in
Denma k. Sus ain. 12, 1–17. h ps://doi.o g/10.3390/su12229335
da Sil a, E.R.O., Pe ei a, M.G., de Ba os, M.M., dos San os, L.M.M., Gomes, J.H.G., 2022.
Soil O ganic Ma e F ac ions and Mul i a ia e Analysis in he De ini ion o Pas u e
Managemen Zones. Eng. Ag ic. 42. h ps://doi.o g/10.1590/1809-4430-
ENG.AGRIC.V42N6E20220099/2022
Damian, J.M., De Cas o Pias, O.H., Che ubin, M.R., Da Fonseca, A.Z., Fo na i, E.Z., San i,
A.L., 2020. Applying he NDVI om sa elli e images in delimi ing managemen zones o
annual c ops. Sci. Ag ic. 77, 1–11. h ps://doi.o g/10.1590/1678-992x-2018-0055
Da is, A.S., Hill, J.D., Chase, C.A., Johanns, A.M., Liebman, M., 2012. Inc easing C opping
Sys em Di e si y Balances P oduc i i y , P o i abili y and En i onmen al Heal h. PLoS
One 7, 1–8. h ps://doi.o g/10.1371/jou nal.pone.0047149
de la C uz, V.Y. V., Tan iani, Cheng, W., Tawa aya, K., 2023. Yield gap be ween o ganic and
con en ional a ming sys ems ac oss clima e ypes and sub- ypes: A me a-analysis.
Ag ic. Sys . 211. h ps://doi.o g/10.1016/j.agsy.2023.103732
de La a, A., Mieno, T., Luck, J.D., Pun el, L.A., 2023. P edic ing si e ‑ speci ic economic op imal
ni ogen a e using machine lea ning me hods and on ‑ a m p ecision expe imen a ion.
P ecis. Ag ic. 24, 1792–1812. h ps://doi.o g/10.1007/s11119-023-10018-8
De Pon i, T., Rijk, B., Van I e sum, M.K., 2012. The c op yield gap be ween o ganic and
con en ional ag icul u e. Ag ic. Sys . 108, 1–9.
h ps://doi.o g/10.1016/j.agsy.2011.12.004
Den ika, P., Ozie -La on aine, H., Pene , L., 2021. Weeds as pa hogen hos s and disease isk
o c ops in he wake o a educed use o he bicides: E idence om yam (Diosco ea ala a)
ields and colle o ichum pa hogens in he opics. J. Fungi 7.
h ps://doi.o g/10.3390/jo 7040283
Dessu eaul -Romp é, J., Zeba h, B.J., Geo gallas, A., Bu on, D.L., G an , C.A., D u y, C.F.,
2010. Tempe a u e dependence o soil ni ogen mine aliza ion a e: Compa ison o
ma hema ical models, e e ence empe a u es and o igin o he soils. Geode ma 157, 97–
108. h ps://doi.o g/10.1016/j.geode ma.2010.04.001
Re e ences
86
DESTATIS, 2024. Landwi scha liche Be iebe insgesam und Be iebe mi ökologischem
Landbau nach Bundeslände n. S a is isches Bundesam [WWW Documen ]. URL
h ps://www.des a is.de/DE/Themen/B anchen-Un e nehmen/Landwi scha -
Fo s wi scha -Fische ei/Landwi scha liche-Be iebe/Tabellen/oekologische -landbau-
bundeslaende .h ml (accessed 3.14.25).
Diacono, M., Cas ignanò, A., Vi i, C., S ellacci, A.M., Ma ino, L., Cocozza, C., De Benede o,
D., T occoli, A., Rubino, P., Ven ella, D., 2014. An app oach o assessing he e ec s o
si e-speci ic e iliza ion on c op g ow h and yield o du um whea in o ganic ag icul u e.
P ecis. Ag ic. 15, 479–498. h ps://doi.o g/10.1007/s11119-014-9347-8
DMK, 2023. Flächene äge on Kö ne mais und Silomais in Deu schland. Deu sches
Maiskomi ee e.V. [WWW Documen ]. URL
h ps://www.maiskomi ee.de/Fak en/S a is ik/Deu schland/Flächene äge (accessed
4.15.23).
Dö ing, T.F., Neuho , D., 2021. Uppe limi s o sus ainable o ganic whea yields. Sci. Rep. 11,
1–11. h ps://doi.o g/10.1038/s41598-021-91940-7
Du u, M., The ond, O., Ma in, G., Ma in-Clouai e, R., Magne, M.A., Jus es, E., Jou ne , E.P.,
Aube o , J.N., Sa a y, S., Be gez, J.E., Sa hou, J.P., 2015. How o implemen
biodi e si y-based ag icul u e o enhance ecosys em se ices: a e iew. Ag on. Sus ain.
De . 35, 1259–1281. h ps://doi.o g/10.1007/s13593-015-0306-1
DWD, 2024. Index o
/clima e_en i onmen /CDC/obse a ions_ge many/clima e/daily/mo e_p ecip/his o ical.
Deu sche We e diens [WWW Documen ]. URL
h ps://openda a.dwd.de/clima e_en i onmen /CDC/obse a ions_ge many/clima e/daily
/mo e_p ecip/his o ical/
DWD, 2020. Clima e da ase - a chi e da a s a ion 342 Belm. Deu sche We e diens [WWW
Documen ].
Eli-Chukwu, N.C., 2019. Applica ions o A i icial In elligence in Sma Ag icul u e: A Re iew.
Eng. Technol. Appl. Sci. Res. 9, 4377–4383. h ps://doi.o g/10.1007/978-981-16-8248-
3_11
Elsalahy, H.H., Belling a h-Kimu a, S.D., Roß, C.L., Kau z, T., Dö ing, T.F., 2020. C op
Resilience o D ough Wi h and Wi hou Response Di e si y. F on . Plan Sci. 11.
h ps://doi.o g/10.3389/ pls.2020.00721
Engels, A.M., Gaise , T., Ewe , F., G ahmann, K., 2025. Simula ing Soil Mois u e Dynamics
in a Di e si ied C opping Sys em Unde He e ogeneous Soil Condi ions. Ag onomy 15.
h ps://doi.o g/10.3390/ag onomy15020407
Esposi o, M., Wes b ook, A.S., Maggio, A., Ci illo, V., DiTommaso, A., 2023. Neu al weed
communi ies: The in e sec ion be ween c op p oduc i i y, biodi e si y, and weed
ecosys em se ices. Weed Sci. 1, 273–299. h ps://doi.o g/10.1017/wsc.2023.27
Eu opean Union, 2020. Fa m o Fo k S a egy. h ps://doi.o g/h ps://doi.o g/10.2875/653604
Eu os a , 2025. O ganic c op a ea by ag icul u al p oduc ion me hods and c ops [WWW
Documen ]. h ps://doi.o g/h ps://doi.o g/10.2908/ORG_CROPAR
FAO, 2024. Land s a is ics 2001–2022 - Global, egional and coun y ends. Food and
Ag icul u e O ganiza ion o he Uni ed Na ions, Rome.
h ps://doi.o g/h ps://doi.o g/10.4060/cd1484en
FAO, 2022. In B ie o The S a e o Food and Ag icul u e 2022. Le e aging au oma ion in
ag icul u e o ans o ming ag i ood sys ems. Rome,. Food and Ag icul u e O ganiza ion
o he Uni ed Na ions, Rome. h ps://doi.o g/10.4060/cc2459en
Re e ences
87
FAO, 2015. Wo ld e e ence base o soil esou ces 2014 in e na ional soil classi ica ion
sys em o naming soils and c ea ing legends o soil maps - Upda e 2015, Wo ld Soil
Resou ces Repo s No. 106. Food and Ag icul u e O ganiza ion o he Uni ed Na ions,
Rome.
Fe nández-Delgado, M., Ce nadas, E., Ba o, S., Amo im, D., 2014. Do we need hund eds o
classi ie s o sol e eal wo ld classi ica ion p oblems? J. Mach. Lea n. Res. 15, 3133–
3181.
Fe nández-Quin anilla, C., Peña, J.M., Andúja , D., Do ado, J., Ribei o, A., López-G anados,
F., 2018. Is he cu en s a e o he a o weed moni o ing sui able o si e-speci ic weed
managemen in a able c ops? Weed Res. 58, 259–272.
h ps://doi.o g/10.1111/w e.12307
Fo es, R., Millán, S., P ie o, M.H., Campillo, C., 2015. A me hodology based on appa en
elec ical conduc i i y and guided soil samples o imp o e i iga ion zoning. P ecis. Ag ic.
16, 441–454. h ps://doi.o g/10.1007/s11119-015-9388-7
Fox, J., 2015. Applied Reg ession Analysis and Gene alized Linea Models, 2. ed. Sage
publica ions.
F ancksen, R.M., Tu nbull, S., Rhyme , C.M., Hi on, M., Bu e, C., Klaus, V.H., Newell-P ice,
P., S ewa , G., Whi ingham, M.J., 2022. The E ec s o Ni ogen Fe ilisa ion on Plan
Species Richness in Eu opean Pe manen G asslands: A Sys ema ic Re iew and Me a-
Analysis. Ag onomy 12. h ps://doi.o g/10.3390/ag onomy12122928
F anzluebbe s, A.J., Zen ella, R., Ka le, A., 2025. Soil-p o ile e ili y is al e ed by soil ex u e
and land use ac oss physiog aphic egions in he sou heas e n Uni ed S a es. Ag on. J.
1–21. h ps://doi.o g/10.1002/agj2.70041
Fussell, J., Rundquis , D., Ha ing on, J.A., 1986. On de ining emo e sensing. Pho og amm.
Eng. Remo e Sens. 52, 1507–1511.
Gab iel, A., Gando e , M., 2023. Adop ion o digi al echnologies in ag icul u e—an in en o y
in a eu opean small-scale a ming egion. P ecis. Ag ic. 24, 68–91.
h ps://doi.o g/10.1007/s11119-022-09931-1
Gao, C., Ji, X., He, Q., Gong, Z., Sun, H., Wen, T., Guo, W., 2023. Moni o ing o Whea
Fusa ium Head Bligh on Spec al and Tex u al Analysis o UAV Mul ispec al Image y.
Ag ic. 13, 1–16. h ps://doi.o g/10.3390/ag icul u e13020293
Geige , F., Beng sson, J., Be endse, F., Weisse , W.W., Emme son, M., Mo ales, M.B.,
Ce yngie , P., Lii a, J., Tscha n ke, T., Winq is , C., Egge s, S., Bomma co, R., Pä , T.,
B e agnolle, V., Plan egenes , M., Clemen , L.W., Dennis, C., Palme , C., Oña e, J.J.,
Gue e o, I., Haw o, V., Aa ik, T., Thies, C., Floh e, A., Hänke, S., Fische , C., Goedha ,
P.W., Inchaus i, P., 2010. Pe sis en nega i e e ec s o pes icides on biodi e si y and
biological con ol po en ial on Eu opean a mland. Basic Appl. Ecol. 11, 97–105.
h ps://doi.o g/10.1016/j.baae.2009.12.001
Geologische Diens NRW, 2003. Geologie im Wese - und Osnab ücke Be gland, 11 h ed.
Geologische Diens NRW, K e eld.
Ge ha ds, R., Andúja Sanchez, D., Hamouz, P., Pe eina os, G.G., Ch is ensen, S.,
Fe nandez-Quin anilla, C., 2022. Ad ances in si e-speci ic weed managemen in
ag icul u e—A e iew. Weed Res. 62, 123–133. h ps://doi.o g/10.1111/w e.12526
Ge ha ds, R., Gu jah , C., Weis, M., Kelle , M., Söke eld, M., Möh ing, J., Piepho, H.P., 2012.
Using p ecision a ming echnology o quan i y yield e ec s a ibu ed o weed compe i ion
and he bicide applica ion. Weed Res. 52, 6–15. h ps://doi.o g/10.1111/j.1365-
3180.2011.00893.x
Re e ences
94
Mon i, M., Pellicanò, A., San onoce o, C., P ei i, G., P is e i, A., 2016. Yield componen s and
ni ogen use in ce eal-pea in e c ops in Medi e anean en i onmen . F. C op. Res. 196,
379–388. h ps://doi.o g/10.1016/j. c .2016.07.017
Moo e, K.J., Mose , L.E., K. P. Vogel, S. S. Walle , B. E. Johnson, J. F. Pede sen, P., 1991.
Desc ibing and Quan i ying G ow h S ages o Pe ennial Fo age G asses. Ag on. J. 1077,
1073–1077.
Mo an, M.S., Inoue, Y., Ba nes, E.M., 1997. Oppo uni ies and limi a ions o image-based
emo e sensing in p ecision c op managemen . Remo e Sens. En i on. 61, 319–346.
h ps://doi.o g/10.1016/S0034-4257(97)00045-X
Mo a i, F., Zanella, V., Sa o i, L., Visioli, G., Be zaghi, P., Mosca, G., 2018. Op imising du um
whea cul i a ion in No h I aly: unde s anding he e ec s o si e-speci ic e iliza ion on
yield and p o ein con en . P ecis. Ag ic. 19, 257–277. h ps://doi.o g/10.1007/s11119-017-
9515-8
Munna , M.A., Haesae , G., Mouazen, A.M., 2022. Si e-speci ic seeding o maize p oduc ion
using managemen zone maps delinea ed wi h mul i-senso s da a usion scheme. Soil
Tillage Res. 220. h ps://doi.o g/10.1016/j.s ill.2022.105377
Munna , M.A., Haesae , G., Van Mei enne, M., Mouazen, A.M., 2020a. Si e-speci ic seeding
using mul i-senso and da a usion echniques: A e iew, 1s ed, Ad ances in Ag onomy.
Else ie Inc. h ps://doi.o g/10.1016/bs.ag on.2019.08.001
Munna , M.A., Haesae , G., Van Mei enne, M., Mouazen, A.M., 2020b. Map-based si e-
speci ic seeding o consump ion po a o p oduc ion using high- esolu ion soil and c op
da a usion. Compu . Elec on. Ag ic. 178. h ps://doi.o g/10.1016/j.compag.2020.105752
Munz, S., Zachmann, J., Raj, I., Raj, N., Jens, D., E ik, H., Jensen, S., Ca lsson, G., 2023.
Yield s abili y and weed d y ma e in esponse o ield-scale soil a iabili y in pea-oa
in e c opping. Plan Soil 506, 391–310. h ps://doi.o g/10.1007/s11104-023-06316-9
Nah s ed , K., Reu e , T., T au z, D., Waske, B., 2024. Classi ying S and Composi ions in
Clo e G ass Based on High-Resolu ion Mul ispec al UAV Images. Remo e Sens. 16, 1–
18. h ps://doi.o g/h ps://doi.o g/10.3390/ s16142684
Neupane, J., Guo, W., 2019. Ag onomic basis and s a egies o p ecision wa e managemen :
A e iew. Ag onomy 9. h ps://doi.o g/10.3390/ag onomy9020087
Ngouajio, M., Lemieux, C., Le oux, G.D., 1999. P edic ion o co n (Zea mays) yield loss om
ea ly obse a ions o he ela i e lea a ea and he ela i e lea co e o weeds. Weed Sci.
47, 297–304. h ps://doi.o g/10.1017/s0043174500091803
Niemeye , M., Renz, M., Puk op, M., Hagemann, D., Zu heide, T., Di Ma co, D., Hö e lin, M.,
S a k, P., Rahe, F., Igelb ink, M., Jenz, M., Ja me , T., T au z, D., S iene, S., He zbe g,
J., 2024. Cogni i e Weeding: An App oach o Single-Plan Speci ic Weed Regula ion. KI
- Küns liche In elligenz. h ps://doi.o g/10.1007/S13218-023-00825-6
Nikolić, N., Rizzo, D., Ma accini, E., Go o , A.A., Ma i i, P., Saule , P., Pe siche i, A., Masin,
R., 2021. Si e-and ime-speci ic ea ly weed con ol is able o educe he bicide use in
maize-a case s udy. I al. J. Ag on. 16. h ps://doi.o g/10.4081/ija.2021.1780
Obe son, A., Ja osch, K.A., F ossa d, E., Hammelehle, A., Fliessbach, A., Mäde , P., Maye ,
J., 2024. Highe han expec ed: Ni ogen lows, budge s, and use e iciencies o e 35
yea s o o ganic and con en ional c opping. Ag ic. Ecosys . En i on. 362.
h ps://doi.o g/10.1016/j.agee.2023.108802
Odone, A., Popo ic, O., Tho up-K is ensen, K., 2024. Deep oo s: implica ions o ni ogen
up ake and d ough ole ance among win e whea cul i a s. Plan Soil 500, 13–32.
h ps://doi.o g/10.1007/s11104-023-06255-5

Re e ences
95
Oldoni, H., Magalhães, P.S.G., Oli ei a, A.L.G., Lima, J.P., Figuei edo, G.K.D.A., Mo o, E.,
Ama al, L.R., 2025. Managemen zones delinea ion: a p oposal o o e come he c op-
pas u e o a ion challenge. P ecis. Ag ic. 26, 1–29. h ps://doi.o g/10.1007/s11119-024-
10214-0
Oli ei a, R.A., Näsi, R., Niemeläinen, O., Nyholm, L., Alhonoja, K., Kai osoja, J., Jauhiainen,
L., Viljanen, N., Nezami, S., Ma kelin, L., Hakala, T., Honka aa a, E., 2020. Machine
lea ning es ima o s o he quan i y and quali y o g ass swa ds used o silage p oduc ion
using d one-based imaging spec ome y and pho og amme y. Remo e Sens. En i on.
246, 111830. h ps://doi.o g/10.1016/j. se.2020.111830
Oli e , M.A., Bishop, T.F.A., Ma chan , B.P., 2013. P ecision ag icul u e o sus ainabili y and
en i onmen al p o ec ion, P ecision Ag icul u e o Sus ainabili y and En i onmen al
P o ec ion. h ps://doi.o g/10.4324/9780203128329
Ouazaa, S., Ja amillo-Ba ios, C.I., Chaali, N., Amaya, Y.M.Q., Ca ajal, J.E.C., Ramos, O.M.,
2022. Towa ds si e speci ic managemen zones delinea ion in o a ional c opping sys em:
Applica ion o mul i a ia e spa ial clus e ing model based on soil p ope ies. Geode ma
Reg. 30. h ps://doi.o g/10.1016/j.geod s.2022.e00564
Pahmeye , C., Kuhn, T., B i z, W., 2021. ‘F uch olge’: A c op o a ion decision suppo sys em
o op imizing c opping choices wi h big da a and spa ially explici modeling. Compu .
Elec on. Ag ic. 181. h ps://doi.o g/10.1016/j.compag.2020.105948
Palka, M., Manschadi, A.M., 2024. On- a m e alua ion o a c op o ecas -based app oach o
season-speci ic ni ogen applica ion in win e whea . P ecis. Ag ic. 25, 2394–2420.
h ps://doi.o g/h ps://doi.o g/10.1007/s11119-024-10175-4 On- a m
Pang, X., Le ey, J., 2000. O ganic Fa ming: Challenge o Timing Ni ogen A ailabili y o C op
Ni ogen Requi emen s. Soil Sci. Soc. Am. J. 64, 247–253.
h ps://doi.o g/10.2136/sssaj2000.641247x
Pannacci, E., Tei, F., 2014. E ec s o mechanical and chemical me hods on weed con ol,
weed seed ain and c op yield in maize, sun lowe and soyabean. C op P o . 64, 51–59.
h ps://doi.o g/10.1016/j.c op o.2014.06.001
Papageo giou, E.I., Ma kinos, A.T., Gem os, T.A., 2011. Fuzzy cogni i e map based app oach
o p edic ing yield in co on c op p oduc ion as a basis o decision suppo sys em in
p ecision ag icul u e applica ion. Appl. So Compu . J. 11, 3643–3657.
h ps://doi.o g/10.1016/j.asoc.2011.01.036
Pä zold, S., Hbi kou, C., Dicke, D., Ge ha ds, R., Welp, G., 2020. Linking weed pa e ns wi h
soil p ope ies: a long- e m case s udy. P ecis. Ag ic. 21, 569–588.
h ps://doi.o g/10.1007/s11119-019-09682-6
Paul, C., Ba kowski, B., Dönmez, C., Don, A., Maye , S., S e ens, M., Weigl, S., Wiesmeie ,
M., Wol , A., Helming, K., 2023. Ca bon a ming: A e soil ca bon ce i ica es a sui able
ool o clima e change mi iga ion? J. En i on. Manage. 330.
h ps://doi.o g/10.1016/j.jen man.2022.117142
Paulsen, H.M., 2008. Misch uch anbausys eme mi Ölp lanzen im ökologischen Landbau 2.
E agss uk u des Misch uch anbaus on Lein (Linum us i a issi um L.) mi
Somme weizen, Ha e ode Leindo e . Landbau o sch. - TI Ag ic. Fo . Res. 4 4, 307–
314.
Paulsen, H.M., 2007. Misch uch anbausys eme mi Ölp lanzen im ökologischen Landbau : 1.
E agss uk u des Misch uch anbaus on Leguminosen ode Somme weizen mi
Leindo e (Camelina sa i a L. C an z). Ag ic. Fo . Res. 57, 107–117.
Paulsen, H.M., Schochow, M., 2007. Anbau on Mischkul u en mi Ölp lanzen zu
Ve besse ung de Flächenp oduk i i ä im Ökologischen Landbau - Näh s o au nahme,
Unk au un e d ückung, Schade ege be all und P oduk quali ä en.
Bundes o schungsans al ü Landwi scha (FAL), Wes e au.
Re e ences
96
Paulsen, H.M., Seling, S., 2007. Quali ä on Ge eide aus Misch uch anbausys emen mi
Ölp lanzen im ökologischen Landbau. Landbau o sch. Völken ode Sonde h. 309, 68–80.
Ped egosa, F., Va oquaux, G., G am o , A., Michel, V., Thi ion, B., G isel, O., Blondel, M.,
P e enho e , P., Weiss, R., Dubou g, V., Vande plas, J., Passos, A., Cou napeau, D.,
B uche , M., Pe o , M., Duchesnay, É., 2011. Sciki -lea n: Machine lea ning in Py hon. J.
Mach. Lea n. Res. 12, 2825–2830.
Pei, H., Sun, Y., Huang, H., Zhang, W., Sheng, J., Zhang, Z., 2022. Weed De ec ion in Maize
Fields by UAV Images Based on C op. Ag icul u e 12, 975.
h ps://doi.o g/h ps://doi.o g/10.3390/ag icul u e12070975
Peña, J.M., To es-Sánchez, J., de Cas o, A.I., Kelly, M., López-G anados, F., 2013. Weed
Mapping in Ea ly-Season Maize Fields Using Objec -Based Analysis o Unmanned Ae ial
Vehicle (UAV) Images. PLoS One 8, 1–11. h ps://doi.o g/10.1371/jou nal.pone.0077151
Pe al a, N.R., Cos a, J.L., Balza ini, M., Cas o F anco, M., Có doba, M., Bullock, D., 2015.
Delinea ion o managemen zones o imp o e ni ogen managemen o whea . Compu .
Elec on. Ag ic. 110, 103–113. h ps://doi.o g/10.1016/j.compag.2014.10.017
Pe o ic, B., Konone s, Y., Csambalik, L., 2025. Adop ion o d one, senso , and obo ic
echnologies in o ganic a ming sys ems o Viseg ad coun ies. Heliyon 11.
h ps://doi.o g/10.1016/j.heliyon.2024.e41408
Pö ke , M., Kiehl, K., Ja me , T., T au z, D., 2023. Con olu ional Neu al Ne wo k Maps Plan
Communi ies in Semi-Na u al G asslands Using Mul ispec al Unmanned Ae ial Vehicle
Image y. Remo e Sens. 15. h ps://doi.o g/10.3390/ s15071945
R Co e Team, 2020. R: A Language and En i onmen o S a is ical Compu ing R. Founda ion
o S a is ical Compu ing Ve sion 4.0.2.
Raa z, L., Bacchi, N., Pi ho e , K., Glemni z, M., Mülle , M.E.H., Joshi, J., Sche be , C., 2019.
How much do we eally lose ?— Yield losses in he p oximi y o na u al landscape
elemen s in ag icul u al landscapes. Ecol. E ol. 9, 7838–7848.
h ps://doi.o g/10.1002/ece3.5370
Rajcan, I., Swan on, C.J., 2001. Unde s anding maize-weed compe i ion: Resou ce
compe i ion, ligh quali y and he whole plan . F. C op. Res. 71, 139–150.
h ps://doi.o g/10.1016/S0378-4290(01)00159-9
Rajmis, S., Ka pinski, I., Pohl, J.P., He mann, M., Kehlenbeck, H., 2022. Economic po en ial
o si e-speci ic pes icide applica ion scena ios wi h di ec injec ion and au oma ic
applica ion assis an in no he n Ge many. P ecis. Ag ic. 23, 2063–2088.
h ps://doi.o g/10.1007/s11119-022-09888-1
Raseduzzaman, M., Jensen, E.S., 2017. Does in e c opping enhance yield s abili y in a able
c op p oduc ion? A me a-analysis. Eu . J. Ag on. 91, 25–33.
h ps://doi.o g/10.1016/j.eja.2017.09.009
Rasmussen, J., Azim, S., Boldsen, S.K., Ni schke, T., Jensen, S.M., Nielsen, J., Ch is ensen,
S., 2021. The challenge o ep oducing emo e sensing da a om sa elli es and
unmanned ae ial ehicles (UAVs) in he con ex o managemen zones and p ecision
ag icul u e. P ecis. Ag ic. 22, 834–851. h ps://doi.o g/10.1007/s11119-020-09759-7
Rasmussen, Jespe , G iepen og, H.W., Nielsen, J., Hen iksen, C.B., 2012. Au oma ed
in elligen o o ine cul i a ion and punch plan ing o imp o e he selec i i y o mechanical
in a- ow weed con ol. Weed Res. 52, 327–337. h ps://doi.o g/10.1111/j.1365-
3180.2012.00922.x
Rasmussen, Jim, Søegaa d, K., Pi ho e -Walzl, K., E iksen, J., 2012. N2- ixa ion and esidual
N e ec o ou legume species and ou companion g ass species. Eu . J. Ag on. 36, 66–
74. h ps://doi.o g/10.1016/j.eja.2011.09.003
Re e ences
97
Raymond Hun , E., Daugh y, C.S.T., Ei el, J.U.H., Long, D.S., 2011. Remo e sensing lea
chlo ophyll con en using a isible band index. Ag on. J. 103, 1090–1099.
h ps://doi.o g/10.2134/ag onj2010.0395
Redwi z, C. Von, Ande , S., S ehlow, B., Ulbe , L., Bensch, J., Fo s e , R., Scha ke, M., 2025.
Enhancing A able Weed Di e si y by Reduced He bicide Use ? J. C op Heal.
h ps://doi.o g/10.1007/s10343-025-01127-7
Reiss, E.R., D inkwa e , L.E., 2018. Cul i a mix u es: a me a-analysis o he e ec o
in aspeci ic di e si y on c op yield. Ecol. Appl. 28, 62–77.
h ps://doi.o g/10.1002/eap.1629
Reumaux, R., 2024. Cons ain s and oppo uni ies o o ganic c op p oduc ion in a eas o high
ag icul u al p oduc i i y. Ac a Uni e si a is Ag icul u ae Sueciae. Swedish Uni e si y o
Ag icul u al Sciences. h ps://doi.o g/10.54612/a.1gli1g6sk
Reu e , T., Mo ales, J.C.S., Tieben, C., Nah s ed , K., K aa z, F., Meemken, H., Hünke , G.,
Lingemann, K., B oll, G., Ja me , T., He zbe g, J., T au z, D., 2023. E alua ion o a
decision suppo sys em o he ecommenda ion o pas u e ha es da e and o m, in:
In o ma ik in De Land-, Fo s - Und E näh ungswi scha -- Re e a e De 43. GIL-
Jah es agung. Ch is a Ho mann; An hony S ein; A no Ruckelshausen; Henning Mülle ;
Thilo S eckel; Helga Flo o, Bonn, pp. 489–494.
Richa dson, K., S e en, W., Luch , W., Bend sen, J., Co nell, S.E., Donges, J.F., D üke, M.,
Fe ze , I., Bala, G., on Bloh, W., Feulne , G., Fiedle , S., Ge en, D., Gleeson, T.,
Ho mann, M., Huiskamp, W., Kummu, M., Mohan, C., Nogués-B a o, D., Pe i, S.,
Po kka, M., Rahms o , S., Schapho , S., Thonicke, K., Tobian, A., Vi kki, V., Wang-
E landsson, L., Webe , L., Rocks öm, J., 2023. Ea h beyond six o nine plane a y
bounda ies. Sci. Ad . 9, 1–16. h ps://doi.o g/10.1126/sciad .adh2458
Richne , N., Holde egge , R., Linde , H.P., Wal e , T., 2015. Re iewing change in he a able
lo a o Eu ope: A me a-analysis. Weed Res. 55, 1–13. h ps://doi.o g/10.1111/w e.12123
Riesinge , P., He zon, I., 2010. Symbio ic ni ogen ixa ion in o ganically managed ed clo e -
g ass leys unde a ming condi ions. Ac a Ag ic. Scand. Sec . B Soil Plan Sci. 60, 517–
528. h ps://doi.o g/10.1080/09064710903233870
Rod iguez Mi anda, D.A., de Oli ei a Ala i, F., Oldoni, H., Bazzi, C.L., do Ama al, L.R.,
G aziano Magalhães, P.S., 2021. Delinea ion o managemen zones in in eg a ed c op–
li es ock sys ems. Ag on. J. 113, 5271–5286. h ps://doi.o g/10.1002/agj2.20912
Roilo, S., Engle , J.O., Václa ík, T., Co d, A.F., 2023. Landscape-le el he e ogenei y o ag i-
en i onmen measu es imp o es habi a sui abili y o a mland bi ds. Ecol. Appl. 33, 1–
15. h ps://doi.o g/10.1002/eap.2720
Roilo, S., Ho mees e , T.R., F auendo , M., Widén, A., Co d, A.F., 2024. The un apped
po en ial o came a aps o a mland biodi e si y moni o ing: cu en p ac ice and
ou s anding ag oecological ques ions. Remo e Sens. Ecol. Conse . 1–12.
h ps://doi.o g/10.1002/ se2.426
Roys on, P., 1995. Rema k AS R94 : A Rema k on Algo i hm AS 181: The W- es o No mali y.
R. S a . Soc. 44, 547–551. h ps://doi.o g/h ps://doi.o g/10.2307/2986146
Sad as, V.O., Cal iño, P.A., 2001. Quan i ica ion o g ain yield esponse o soil dep h in
soybean, maize, sun lowe , and whea . Ag on. J. 93, 577–583.
h ps://doi.o g/10.2134/ag onj2001.933577x
Saikai, Y., Pa el, V., Mi chell, P.D., 2020. Machine lea ning o op imizing complex si e-speci ic
managemen . Compu . Elec on. Ag ic. 174, 105381.
h ps://doi.o g/10.1016/j.compag.2020.105381
Re e ences
98
Saile, M., Spae h, M., Ge ha ds, R., 2022. E alua ing Senso -Based Mechanical Weeding
Combined wi h P e-and Pos -Eme gence He bicides o In eg a ed Weed Managemen
in Ce eals. Ag onomy 12. h ps://doi.o g/10.3390/ag onomy12061465
Samuel, A.L., 2000. Some s udies in machine lea ning using he game o checke s. IBM J.
Res. De . 44, 206–226. h ps://doi.o g/10.1147/ d.441.0206
Sande s, J., B inkmann, J., Chmeliko a, L., Ebe sede , F., F eibaue , A., Go wald, F., Haub,
A., Hauschild, M., Hoppe, J., Hülsbe gen, K.J., Jung, R., Kusche, D., Le in, K., Ma ch,
S., Schmid ke, K., S ein-Bachinge , K., T eu, H., Weckenb ock, P., Wiesinge , K.,
Ga inge , A., Heß, J., 2025. Bene i s o o ganic ag icul u e o en i onmen and animal
wel a e in empe a e clima es. O g. Ag ic. h ps://doi.o g/10.1007/s13165-025-00493-w
Sapko a, A., Ve di, A., Scudie o, E., Mon aza , A., 2024. Assessing he e ec i eness o
sa elli e and UAV-based emo e sensing o delinea ing al al a managemen zones unde
he e ogeneous oo zone soil salini y. Sma Ag ic. Technol. 9.
h ps://doi.o g/10.1016/j.a ech.2024.100583
Schaack, D., Rampold, C., 2021. AMI Ma k Bilanz Öko-Landbau 2021 (in Ge man). Bonn.
Schaan, L.N., Finch, E.A., Wa enbe g, A.C., Boe ne , V.S., Belling a h-Kimu a, S.D., Bonn,
A., Pe’e , G., 2025. Mapping and p io i ising landscape ea u e es o a ion in ag icul u al
landscapes: A case s udy in B andenbu g, Ge many. Land use policy 154.
h ps://doi.o g/10.1016/j.landusepol.2025.107531
Schulp, C.J.E., Ve bu g, P.H., 2009. E ec o land use his o y and si e ac o s on spa ial
a ia ion o soil o ganic ca bon ac oss a physiog aphic egion. Ag ic. Ecosys . En i on.
133, 86–97. h ps://doi.o g/10.1016/j.agee.2009.05.005
Schus e , J., Hagn, L., Mi e maye , M., Hülsbe gen, K.J., 2024. A e e ec s o his o ical
g assland on soil o ganic ca bon con en and plan g ow h in c oplands in sou he n
Ge many de e mined using sa elli e da a. Sci. To al En i on. 947.
h ps://doi.o g/10.1016/j.sci o en .2024.174507
Sei z, S., Goebes, P., Pue a, V.L., Pe ei a, E.I.P., Wi we , R., Six, J., an de Heijden, M.G.A.,
Schol en, T., 2019. Conse a ion illage and o ganic a ming educe soil e osion. Ag on.
Sus ain. De . 39. h ps://doi.o g/10.1007/s13593-018-0545-z
Sel o s, L., We s, P., G een, T., 2018. Looking beyond he jug: Non-chemical weed seedbank
managemen . C op. Soils 51, 28–53. h ps://doi.o g/10.2134/cs2018.51.0504
Shams, M.Y., Gamel, S.A., Talaa , F.M., 2024. Enhancing c op ecommenda ion sys ems wi h
explainable a i icial in elligence: a s udy on ag icul u al decision-making. Neu al Compu .
Appl. 36, 5695–5714. h ps://doi.o g/10.1007/s00521-023-09391-2
Sha ma, L.K., Bali, S.K., 2018. A e iew o me hods o imp o e ni ogen use e iciency in
ag icul u e. Sus ain. 10, 1–23. h ps://doi.o g/10.3390/su10010051
Shukla, M.K., Sha ma, P., 2023. Fuzzy K-Means and P incipal Componen Analysis o
Classi ying Soil P ope ies o E icien Fa m Managemen and Main aining Soil Heal h.
Sus ain. 15. h ps://doi.o g/10.3390/su151713144
Si ami, C., G oss, N., Baillod, A.B., Be and, C., Ca ié, R., Hass, A., Henckel, L., Migue , P.,
Vuillo , C., Alignie , A., Gi a d, J., Ba á y, P., Clough, Y., Violle, C., Gi al , D., Bo a, G.,
Badenhausse , I., Le eb e, G., Gau e, B., Viala e, A., Cala ayud, F., Gil-Tena, A.,
Tischendo , L., Mi chell, S., Lindsay, K., Geo ges, R., Hilai e, S., Recasens, J., Solé-
Senan, X.O., Robleño, I., Bosch, J., Ba ien os, J.A., Rica e, A., Ma cos-Ga cia, M.Á.,
Miñano, J., Ma he e , R., Gibon, A., Baud y, J., Balen , G., Poulin, B., Bu el, F.,
Tscha n ke, T., B e agnolle, V., Si iwa dena, G., Ouin, A., B o ons, L., Ma in, J.L., Fah ig,
L., 2019. Inc easing c op he e ogenei y enhances mul i ophic di e si y ac oss
ag icul u al egions. P oc. Na l. Acad. Sci. U. S. A. 116, 16442–16447.
h ps://doi.o g/10.1073/pnas.1906419116
Re e ences
99
Skendži, S., Zo ko, M., Leši, V., Ži ko ić, I.P., Lemic, D., 2023. De ec ion and E alua ion o
En i onmen al S ess in Win e Whea Using Remo e and P oximal Sensing Me hods and
Vege a ion Indices—A Re iew. Di e s. 15, 481. h ps://doi.o g/h ps://doi.o g/10.3390/
d15040481
Skie ucha, W., Wilczek, A., Szypłowska, A., Sławiński, C., Lamo ski, K., 2012. A TDR-based
soil mois u e moni o ing sys em wi h simul aneous measu emen o soil empe a u e and
elec ical conduc i i y. Senso s (Swi ze land) 12, 13545–13566.
h ps://doi.o g/10.3390/s121013545
Smi h, O.M., Cohen, A.L., Riese , C.J., Da is, A.G., Taylo , J.M., Adesanya, A.W., Jones, M.S.,
Meie , A.R., Reganold, J.P., O pe , R.J., No h ield, T.D., C owde , D.W., 2019. O ganic
Fa ming P o ides Reliable En i onmen al Bene i s bu Inc eases Va iabili y in C op
Yields: A Global. F on . Sus ain. Food Sys . 3, 1–10.
h ps://doi.o g/10.3389/ su s.2019.00082
Smolka, M., Puchbe ge -Enengl, D., Bipoun, M., Klasa, A., Kiczkajlo, M., Śmiechowski, W.,
Sowiński, P., K u zle , C., Keplinge , F., Vellekoop, M.J., 2017. A mobile lab-on-a-chip
de ice o on-si e soil nu ien analysis. P ecis. Ag ic. 18, 152–168.
h ps://doi.o g/10.1007/s11119-016-9452-y
Sol ani, N., Dille, A.J., Bu ke, I.C., E e man, W.J., VanGessel, M.J., Da is, V.M., Sikkema,
P.H., 2016. Po en ial co n yield losses due o weeds in No h Ame ica. Weed Technol.
30, 979–984. h ps://doi.o g/h ps://doi.o g/10.1614/WT-D-16-00046.1
Song, H., He, Y., 2005. C op Nu i ion Diagnosis Expe Sys em Based on A i icial Neu al
Ne wo ks, in: P oceedings o he Thi d In e na ional Con e ence on In o ma ion
Technology and Applica ions (ICITA’05).
Sonn ag, W.I., Wien ich, N., Se e in, M., Schulze Schwe ing, D., 2022. P ecision Fa ming –
Nullnumme ode Nu zb inge ? Be ich e übe Landwi scha .
Sonobe, R., Yamaya, Y., Tani, H., Wang, X., Kobayashi, N., Mochizuki, K., 2018. C op
classi ica ion om Sen inel-2-de i ed ege a ion indices using ensemble lea ning. J. Appl.
Remo e Sens. 12, 1. h ps://doi.o g/10.1117/1.j s.12.026019
Sowiński, J., 2023. E ec o mechanical illage ea men s in ensi y on weed in es a ion and
yield o quinoa (Chenopodium quinoa Willd.). C op P o . 172.
Spae h, M., Söke eld, M., Schwade e , P., Gaue , M.E., S u m, D.J., Dela ée, C.C., Ge ha ds,
R., 2024. Sma sp aye a echnology o si e-speci ic he bicide applica ion. C op P o .
177. h ps://doi.o g/10.1016/j.c op o.2023.106564
Spe anza, E.A., Mendonça, D., Manoel, C., Vaz, P., Ribei o, L., Rabelo, L.M., And , L., Cas o,
D., Chagas, S., Schelp, M.X., Vecchi, L., 2023. Delinea ing Managemen Zones wi h
Di e en Yield Po en ials in Soybean – Co n and Soybean – Co on P oduc ion Sys ems.
Ag iEnginee ing 5, 1481–1497.
h ps://doi.o g/h ps://doi.o g/10.3390/ag ienginee ing5030092
S ein-Bachinge , K., P eißel, S., Kühne, S., Reckling, M., 2022. Mo e di e se bu less in ensi e
a ming enhances biodi e si y. T ends Ecol. E ol. 37, 395–396.
h ps://doi.o g/10.1016/j. ee.2022.01.008
S einmann, H.H., 2002. Impac o ha owing on he ni ogen dynamics o plan s and soil. Soil
Tillage Res. 65, 53–59. h ps://doi.o g/10.1016/S0167-1987(01)00278-1
Su cli e, L.M.E., Schellenbe g, J., Meye , S., Leuschne , C., 2024. Close o he edge: Spa ial
a ia ion in plan di e si y, biomass and lo al esou ces in con en ional and ag i-
en i onmen ce eal ields. J. Appl. Ecol. 2075–2086. h ps://doi.o g/10.1111/1365-
2664.14737

Re e ences
100
Su e , M., Connolly, J., Finn, J.A., Loges, R., Ki wan, L., Sebas ià, M.T., Lüsche , A., 2015.
Ni ogen yield ad an age om g ass-legume mix u es is obus o e a wide ange o
legume p opo ions and en i onmen al condi ions. Glob. Chang. Biol. 21, 2424–2438.
h ps://doi.o g/10.1111/gcb.12880
Swan on, C.J., Mahoney, K.J., Chandle , K., Gulden, R.H., 2008. In eg a ed Weed
Managemen : Knowledge-Based Weed Managemen Sys ems. Weed Sci. 56, 168–172.
h ps://doi.o g/10.1614/ws-07-126.1
Thomsen, I.K., Schjønning, P., Olesen, J.E., Ch is ensen, B.T., 2003. C and N u no e in
s uc u ally in ac soils o di e en ex u e. Soil Biol. Biochem. 35, 765–774.
h ps://doi.o g/10.1016/S0038-0717(03)00093-2
Tilman, D., 2020. Bene i s o in ensi e ag icul u al in e c opping. Na . Plan s 6, 604–605.
h ps://doi.o g/10.1038/s41477-020-0677-4
Topog aphic-map.com, 2025. Topog a ische Ka e Osnab ück [WWW Documen ]. URL
h ps://de-de. opog aphic-map.com/map-
h66m /Osnab ück/?cen e =52.21198%2C8.06741
To es-Sánchez, J., López-G anados, F., De Cas o, A.I., Peña-Ba agán, J.M., 2013.
Con igu a ion and Speci ica ions o an Unmanned Ae ial Vehicle (UAV) o Ea ly Si e
Speci ic Weed Managemen . PLoS One 8, 1–15.
h ps://doi.o g/10.1371/jou nal.pone.0058210
Tscha n ke, T., G ass, I., Wange , T.C., Wes phal, C., Ba á y, P., 2021. Beyond o ganic
a ming – ha nessing biodi e si y- iendly landscapes. T ends Ecol. E ol. 36, 919–930.
h ps://doi.o g/10.1016/j. ee.2021.06.010
Tuck, S.L., Winq is , C., Mo a, F., Ahns öm, J., Tu nbull, L.A., Beng sson, J., 2014. Land-use
in ensi y and he e ec s o o ganic a ming on biodi e si y: A hie a chical me a-analysis.
J. Appl. Ecol. 51, 746–755. h ps://doi.o g/10.1111/1365-2664.12219
Tucke , C.J., 1979. Red and pho og aphic in a ed linea combina ions o moni o ing
ege a ion. Remo e Sens. En i on. 8, 127–150. h ps://doi.o g/10.1016/0034-
4257(79)90013-0
Tuomis o, H.L., Hodge, I.D., Rio dan, P., Macdonald, D.W., 2012. Does o ganic a ming
educe en i onmen al impac s? - A me a-analysis o Eu opean esea ch. J. En i on.
Manage. 112, 309–320. h ps://doi.o g/10.1016/j.jen man.2012.08.018
Tu sun, N., Da a, A., Sakinmaz, M.S., Kan a ci, Z., Kneze ic, S.Z., Chauhan, B.S., 2016. The
c i ical pe iod o weed con ol in h ee co n (Zea mays L.) ypes. C op P o . 90, 59–65.
h ps://doi.o g/10.1016/j.c op o.2016.08.019
Umwel s i ung Michael O o, Deu sche Baue n e band e.V, 2023. F.R.A.N.Z. Zwischenbilanz
2023 – Ak uelle E kenn nisse aus dem F.R.A.N.Z.- P ojek . 54.
Usowicz, B., Lipiec, J., 2017. Spa ial a iabili y o soil p ope ies and ce eal yield in a cul i a ed
ield on sandy soil. Soil Tillage Res. 174, 241–250.
h ps://doi.o g/10.1016/j.s ill.2017.07.015
Vallen in, C., Dobe s, E.S., I ze o , S., Kleinschmi , B., Spengle , D., 2020. Delinea ion o
managemen zones wi h spa ial da a usion and belie heo y. P ecis. Ag ic. 21, 802–830.
h ps://doi.o g/10.1007/s11119-019-09696-0
Van Ee d, L.L., Chahal, I., Peng, Y., Aw ey, J.C., 2023. In luence o co e c ops a he ou
sphe es: A e iew o ecosys em se ices, po en ial ba ie s, and u u e di ec ions o No h
Ame ica. Sci. To al En i on. 858. h ps://doi.o g/10.1016/j.sci o en .2022.159990
Vande mee , J., 1989. The ecology o in e c opping. Camb idge Uni e si y P ess.
Re e ences
101
Vannoppen, A., Gobin, A., 2022. Es ima ing Yield om NDVI, Wea he Da a, and Soil Wa e
Deple ion o Suga Bee and Po a o in No he n Belgium. Wa e (Swi ze land) 14, 1–15.
h ps://doi.o g/10.3390/w14081188
Vasileiadis, V.P., O o, S., an Dijk, W., U ek, G., Lesko šek, R., Ve schwele, A., Fu lan, L.,
Sa in, M., 2015. On- a m e alua ion o in eg a ed weed managemen ools o maize
p oduc ion in h ee di e en ag o-en i onmen s in Eu ope: Ag onomic e icacy, he bicide
use educ ion, and economic sus ainabili y. Eu . J. Ag on. 63, 71–78.
h ps://doi.o g/10.1016/j.eja.2014.12.001
VDLUFA, 2007. Handbuch de landwi scha lichen Ve suchs-und Un e suchungsme hodik
(VDLUFA-Me hodenbuch). Ve band Deu sche Landwi scha liche Un e suchungs- und
Fo schungsans al en e. V.-Ve lag, Ge many.
Vidican, R., Mălinaș, A., Ran a, O., Moldo an, C., Ma ian, O., Ghețe, A., Ghișe, C.R., Popo ici,
F., Că unescu, G.M., 2023. Using Remo e Sensing Vege a ion Indices o he
Disc imina ion and Moni o ing o Ag icul u al C ops: A C i ical Re iew. Ag onomy 13, 1–
27. h ps://doi.o g/10.3390/ag onomy13123040
Vikuk, V., Spi kanede , A., Noack, P., Duemig, A., 2024. Sma Ag icul u al Technology
Valida ion o a senso -sys em o eal- ime measu emen o mine alized ni ogen in soils.
Sma Ag ic. Technol. 7, 100390. h ps://doi.o g/10.1016/j.a ech.2023.100390
Viljanen, N., Honka aa a, E., Näsi, R., Hakala, T., Niemeläinen, O., Kai osoja, J., 2018. A
no el machine lea ning me hod o es ima ing biomass o g ass swa ds using a
pho og amme ic canopy heigh model, images and ege a ion indices cap u ed by a
d one. Ag ic. 8. h ps://doi.o g/10.3390/ag icul u e8050070
Walke , E., Wooli e , R., Russo, L., Jagadamma, S., 2024. The con ex -dependen bene i s o
o ganic a ming on pollina o biodi e si y: A me a-analysis. J. Appl. Ecol. 62, 41–52.
h ps://doi.o g/10.1111/1365-2664.14826
Walsh, M.J., Squi es, C.C., Coleman, G.R.Y., Widde ick, M.J., McKie nan, A.B., Chauhan,
B.S., Pe essini, C., Guzzomi, A.L., 2020. Tillage based, si e-speci ic weed con ol o
conse a ion c opping sys ems. Weed Technol. 34, 704–710.
h ps://doi.o g/10.1017/we .2020.34
Wegene , J.K., U so, L.-M., on Hö s en, D., Hegewald, H., Minßen, T.-F., Scha enbe g, J.,
Gaus, C.-C., de Wi e, T., Niebe g, H., Ise meye , F., F e ichs, L., Backhaus, G.F., 2019.
Spo a ming – an al e na i e o u u e plan p oduc ion. J. ü Kul . 71, 69–69.
h ps://doi.o g/10.5073/J K.2019.04.01
Welch, B.L., 1947. The gene aliza ion o “s uden s” p oblem when se e al di e en popula ion
a iances a e in ol ed. Biome ika 34, 28–35. h ps://doi.o g/10.1093/biome /34.1-2.28
Wezel, A., Casag ande, M., Cele e, F., Vian, J.-F., Fe e , A., Peigné, J., 2014. Ag oecological
p ac ices o sus ainable ag icul u e. A e iew. Ag on. Sus ain. De . 34, 1–20.
h ps://doi.o g/10.1007/s13593-013-0180-7
Wie man, M.J., Dob ansky, M.K., 1993. A e iew o : “NEURAL NETWORKS AND FUZZY
SYSTEMS:” A Dynamic Sys ems App oach o Machine In elligence. In . J. Gen. Sys . 20,
37–41. h ps://doi.o g/10.1080/03081079208945045
Wiesmeie , M., U banski, L., Hobley, E., Lang, B., on Lü zow, M., Ma in-Spio a, E., an
Wesemael, B., Rabo , E., Ließ, M., Ga cia-F anco, N., Wollschläge , U., Vogel, H.J.,
Kögel-Knabne , I., 2019. Soil o ganic ca bon s o age as a key unc ion o soils - A e iew
o d i e s and indica o s a a ious scales. Geode ma 333, 149–162.
h ps://doi.o g/10.1016/j.geode ma.2018.07.026
Wiles, L.J., 2009. Beyond pa ch sp aying: Si e-speci ic weed managemen wi h se e al
he bicides. P ecis. Ag ic. 10, 277–290. h ps://doi.o g/10.1007/s11119-008-9097-6
Re e ences
102
Wille , H., T á níček, J., Meie , Cl., Schla e , B. (Eds. ., 2021. The Wo ld o O ganic
Ag icul u e. S a is ics and Eme ging T ends 2021. In . J. Sus ain. High. Educ.
h ps://doi.o g/10.1108/ijshe.2009.24910aae.004
Wille , W., Rocks öm, J., Loken, B., Sp ingmann, M., Lang, T., Ve meulen, S., Ga ne , T.,
Tilman, D., DeCle ck, F., Wood, A., Jonell, M., Cla k, M., Go don, L.J., Fanzo, J., Hawkes,
C., Zu ayk, R., Ri e a, J.A., De V ies, W., Majele Sibanda, L., A shin, A., Chaudha y, A.,
He e o, M., Agus ina, R., B anca, F., La ey, A., Fan, S., C ona, B., Fox, E., Bigne , V.,
T oell, M., Lindahl, T., Singh, S., Co nell, S.E., S ina h Reddy, K., Na ain, S., Nish a , S.,
Mu ay, C.J.L., 2019. Food in he An h opocene: he EAT–Lance Commission on heal hy
die s om sus ainable ood sys ems. Lance 393, 447–492.
h ps://doi.o g/10.1016/S0140-6736(18)31788-4
Williams, J.T., 1963. Chenopodium Album L. J. Ecol. 51, 711. h ps://doi.o g/10.2307/2257758
Woźniak, A., 2020. Mechanical and chemical weeding e ec s on he weed s uc u e in du um
whea . I al. J. Ag on. 15, 102–108. h ps://doi.o g/10.4081/ija.2020.1559
Wueppe , D., Hube , R., 2022. Compa ing e ec i eness and e u n on in es men o ac ion-
and esul s-based ag i-en i onmen al paymen s in Swi ze land. Am. J. Ag ic. Econ. 104,
1585–1604. h ps://doi.o g/10.1111/ajae.12284
Wulanning yas, H.S., Gong, Y., Li, P., Sakagami, N., Nishiwaki, J., Koma suzaki, M., 2021. A
co e c op and no- illage sys em o enhancing soil heal h by inc easing soil o ganic
ma e in soybean cul i a ion. Soil Tillage Res. 205, 104749.
h ps://doi.o g/10.1016/j.s ill.2020.104749
Xu, K., Su, Y., Liu, J., Hu, T., Jin, S., Ma, Q., Zhai, Q., Wang, R., Zhang, J., Li, Y., Liu, H., Guo,
Q., 2020. Es ima ion o deg aded g assland abo eg ound biomass using machine
lea ning me hods om e es ial lase scanning da a. Ecol. Indic. 108, 105747.
h ps://doi.o g/10.1016/j.ecolind.2019.105747
Yu, Y., S omph, T.-J., Makowski, D., an de We , W., 2015. Tempo al niche di e en ia ion
inc eases he land equi alen a io o annual in e c ops: A me a-analysis. F. C op. Res.
184, 133–144. h ps://doi.o g/10.1016/j. c .2015.09.010
Yu, Y., S omph, T.-J., Makowski, D., Zhang, L., an de We , W., 2016. A me a-analysis o
ela i e c op yields in ce eal/legume mix u es sugges s op ions o managemen . F. C op.
Res. 198, 269–279. h ps://doi.o g/10.1016/j. c .2016.08.001
Zeng, L., Chen, C., 2018. Using emo e sensing o es ima e o age biomass and nu ien
con en s a di e en g ow h s ages. Biomass and Bioene gy 115, 74–81.
h ps://doi.o g/10.1016/j.biombioe.2018.04.016
Zhao, F., Jiao, L., Liu, H., 2013. Ke nel gene alized uzzy c-means clus e ing wi h spa ial
in o ma ion o image segmen a ion. Digi . Signal P ocess. 23, 184–199.
h ps://doi.o g/10.1016/j.dsp.2012.09.016
Zhao, Y., Tian, Y., Li, X., Song, M., Fang, X., Jiang, Y., Xu, X., 2022. Ni ogen ixa ion and
ans e be ween legumes and ce eals unde a ious c opping egimes. Rhizosphe e 22.
h ps://doi.o g/10.1016/j. hisph.2022.100546
Zhu, H., Chen, X., Zhang, Y., 2013. Tempo al and spa ial a iabili y o ni ogen in ice-whea
o a ion in ield scale. En i on. Ea h Sci. 68, 585–590. h ps://doi.o g/10.1007/s12665-
012-1762-4
Zhu, X.X., Tuia, D., Mou, L., Xia, G.S., Zhang, L., Xu, F., F aundo e , F., 2017. Deep Lea ning
in Remo e Sensing: A Comp ehensi e Re iew and Lis o Resou ces. IEEE Geosci.
Remo e Sens. Mag. 5, 8–36. h ps://doi.o g/10.1109/MGRS.2017.2762307
Zingsheim, M.L., Dö ing, T.F., 2024. Wha weeding obo s need o know abou ecology. Ag ic.
Ecosys . En i on. 364, 108861. h ps://doi.o g/10.1016/j.agee.2023.108861
Acknowledgemen s
103
Acknowledgemen s
Th oughou my esea ch jou ney, including ield ials, analysis, and w i ing, I ecei ed
conside able assis ance and suppo . The e o e, I ex end my g a i ude o e e yone in ol ed.
Thank you o you eedback on my p esen a ions and w i en wo k. I am g a e ul o he
discussions ega ding me hodology and da a in e p e a ion, as well as he hou s spen in he
ield, especially, so ing clo e and g ass lea es. You dese e li e long cookies! You open
ea s o my complain s and emo ional suppo we e in aluable. Fu he mo e, I app ecia e he
enjoyable momen s sha ed du ing ieldwo k, o ice hou s, con e ences, and a home. Thank
you o you companionship!
Du ing his jou ney, I encoun e ed nume ous inspi ing and nice indi iduals. I wish o
exp ess my since e hanks o Hochschule Osnab ück Uni e si y o Applied Sciences o
p o iding he acili ies, unding, and suppo . Speci ically, I am g a e ul o he membe s o T ial
Fa m Waldho and he Sus ainable Ag icul u al Land Use wo king g oup: Ma ia, Ma eike,
Da id, Janis, Tim, Phyu Phyu, Tobias, Niko, Lisa, Hannes, Maike, and Maik. I g ea ly alued
my ime wi h his wo king g oup and hei unwa e ing suppo . A signi ican po ion o his hesis
was con ibu ed h ough he "Ag o-No dwes " p ojec , unded by he Fede al Minis y o Food
and Ag icul u e (BMEL). The expe iences wi hin his p ojec and he people I me he ein we e
en iching, and I lea ned ex ensi ely.
I ex end special hanks o my co-au ho s, Insa Kühling, Thomas Ja me , and Kons an in
Nah s ed . Insa Kühling has signi ican ly aided in my academic de elopmen since my
bachelo 's s udies. Thom and Kons an in p o ided new, “ emo e” pe spec i es, and ou
discussions on me hodology, da a in e p e a ion, and Led Zeppelin we e immensely
cons uc i e. F om my pe spec i e, Kons an in and I o med an excellen eam, om da a
collec ion in he ield o p esen ing he esul s.
I also app ecia e he help and p oo eading e o s o my iends: Ka h in, Ka ee Ka hi,
I mi, Le en , Juan-Ca los and Melanie. Mo e impo an ly, I alue hei emo ional suppo .
Thank you!
A special acknowledgmen goes o Gab iele B oll o supe ising my p omo ion. I ha e
lea ned a g ea deal om you in w i ing scien i ic ex s and app ecia e you expe ise and clea
guidance. My p o ound hanks ex end o Die e T au z o his unwa e ing suppo in esol ing
oubles and secu ing necessa y unding. I pa icula ly app ecia e you dedica ion o os e ing
a welcoming wo king cul u e. I belie e I had one o he bes supe iso s I could wish o .
Las ly, bu ce ainly no leas , I am deeply g a e ul o my amily o suppo ing me
h oughou my li e and p o iding me he oppo uni y o a end uni e si y and comple e my Ph.D.
Dankeschön!