Weather-driven cycling: developing a predictive model for urban bicycle usage based on five key weather factors

Recei ed: 29 No embe 2024
Re ised: 22 Janua y 2025
Accep ed: 6 Feb ua y 2025
Published: 11 Feb ua y 2025
Ci a ion: Falah, N.; Falah, N.;
Solis-Guzman, J. Wea he -D i en
Cycling: De eloping a P edic i e
Model o U ban Bicycle Usage Based
on Fi e Key Wea he Fac o s. U ban
Sci. 2025,9, 41. h ps://doi.o g/
10.3390/u bansci9020041
Copy igh : © 2025 by he au ho s.
Licensee MDPI, Basel, Swi ze land.
This a icle is an open access a icle
dis ibu ed unde he e ms and
condi ions o he C ea i e Commons
A ibu ion (CC BY) license
(h ps://c ea i ecommons.o g/
licenses/by/4.0/).
A icle
Wea he -D i en Cycling: De eloping a P edic i e Model o
U ban Bicycle Usage Based on Fi e Key Wea he Fac o s
Nahid Falah 1, Nadia Falah 2and Jaime Solis-Guzman 2,*
1Landk eis Ha bu g, Schloßpla z 6, 21423 Winsen (Luhe), Ge many; [email p o ec ed]
2
A DiTec Resea ch G oup, Depa men o A chi ec u al Cons uc ions II, Highe Technical School o Building
Enginee ing, Uni e sidad de Se illa, A . Reina Me cedes 4-a, 41012 Se ille, Spain; [email p o ec ed]
*Co espondence: [email p o ec ed]
Abs ac : Wea he condi ions signi ican ly in luence u ban cycling, shaping bo h i s e-
quency and in ensi y. This s udy de elops a p edic i e model o e alua e he impac o i e
key me eo ological ac o s, namely empe a u e, humidi y, p ecipi a ion, wind speed, and
dayligh du a ion, on u ban cycling ends. Using non-linea eg ession analysis, he e-
sea ch examines cycling da a om 2017 o 2019 in Hambu g, Ge many, compa ing p edic ed
alues o 2019 wi h ac ual da a o assess model accu acy. The s a is ical analyses e eal
s ong co ela ions be ween wea he pa ame e s and cycling ac i i y, highligh ing each ac-
o ’s unique in luence. The model achie ed high accu acy, wi h R
2
alues o 0.942 and 0.924
o 2017 and 2019, espec i ely. To u he alida e i s obus ness, he model is applied o
da a om 2021 and 2023—yea s no included in i s ini ial
de elopmen —yielding
R
2
alues
o 0.893 and 0.919. These esul s unde sco e he model’s eliabili y and adap abili y ac oss
di e en ime ames. This s udy no only con i ms he c i ical in luence o wea he on
u ban cycling pa e ns, bu also p o ides a scalable amewo k o b oade u ban planning
applica ions. Beyond he immedia e indings, his esea ch p oposes expanding he model
o inco po a e u ban ac o s, such as land use, popula ion densi y, and socioeconomic
condi ions, o e ing a comp ehensi e ool o u ban planne s and policymake s o enhance
sus ainable anspo a ion sys ems.
Keywo ds: wea he pa ame e s; p edic i e model; u ban cycling pa e ns; me eo ological
impac ; non-linea eg ession
1. In oduc ion
O e he pas cen u y, he global popula ion has quad upled, while he numbe o
passenge kilome e s using anspo a ion has inc eased a s agge ing 100- old [
1
–
3
]. The
apid expansion o u baniza ion and he g owing numbe o p i a e ehicles ha e exace -
ba ed he ela ed social and en i onmen al challenges, pa icula ly in la ge-scale ci ies [
4
,
5
].
Issues such as noise pollu ion, a ic conges ion, poo ai quali y, and g eenhouse gas
emissions now demand u gen solu ions [
3
,
5
–
8
]. In esponse o hese challenges, al e na-
i e modes o anspo a ion, such as public ansi , cycling, and walking, ha e gained
inc easing a en ion [9–11].
Among hese al e na i es, u ban cycling has eme ged as a sus ainable and p ac ical
op ion. As ci ies ace moun ing en i onmen al conce ns and he need o e icien mobili y
solu ions, cycling o e s signi ican bene i s [
4
,
5
,
12
]. I imp o es ai quali y, educes a ic
conges ion and noise, lowe s anspo a ion cos s, and p omo es physical wellbeing in
u ban a eas [
5
,
11
,
13
–
18
]. Resea ch has consis en ly shown ha cycling gene a es subs an-
ially lowe emissions compa ed o mo o ized anspo [
19
–
21
], making i a co ne s one o
U ban Sci. 2025,9, 41 h ps://doi.o g/10.3390/u bansci9020041
U ban Sci. 2025,9, 41 2 o 30
sus ainable u ban de elopmen [
22
]. Cycling is pa icula ly e ec i e o sho commu es
o up o 5 km [
23
,
24
], o e ing a iable al e na i e o ca s in c owded u ban a eas [
25
]. I
also se es as a c ucial mode o anspo o indi iduals wi hou access o p i a e ehicles,
including adolescen s, child en, and he elde ly [
5
,
11
,
26
–
28
]. Beyond i s en i onmen al
ad an ages, cycling is cos e ec i e, as , ene gy e icien , and enjoyable, making i an
inc easingly popula choice o daily u ban a el [29–31].
The adop ion o cycling, howe e , is in luenced by a complex in e play o ac o s,
anging om na u al and buil en i onmen al condi ions o empo al and con ex ual
dynamics. In Table 1, he indica o s iden i ied in p e ious esea ch a e summa ized,
p o iding a comp ehensi e iew o he ac o s in luencing cycling pa e ns. This s udy
ocuses on one key aspec : wea he condi ions. By de eloping a p edic i e model ha
inco po a es i e c i ical wea he ac o s, we aim o deepen ou unde s anding o how
wea he d i es u ban cycling beha io and o p o ide ac ionable insigh s o os e ing
cycling- iendly ci ies.
Table 1. Key indica o s in luencing cycling pa e ns and ele an s udies.
Indica o s Resea ch
Topog aphy [15,32–34]
Cycling in as uc u e [34–43]
Social and cul u al cha ac e is ics [36,44,45]
Ai pollu ion [4,12,35,46–49]
Economic condi ions and popula ion densi y [5,24,50,51]
Di e en pe iods o ime (day and mon h) [34,47,50,52,53]
S uc u e o he en i onmen al condi ions [54–56]
Wea he pa ame e s [12,32,35,40,46–49,57–62]
Table 1ou lines key ac o s in luencing u ban cycling ac oss en i onmen al, in as-
uc u al, and social dimensions, o ming a basis o analyzing he challenges in ega d o
p omo ing cycling in he ollowing sec ions.
Among a ious de e minan s, wea he condi ions ha e been iden i ied as key ac o s
in luencing cyclis s’ com o , sa e y, and o e all willingness o ide (Table 2). Resea ch
indica es ha wea he impac s cycling beha io mo e signi ican ly han ac o s such as o-
pog aphy, in as uc u e, land use composi ion, and calenda e en s. Un a o able wea he
condi ions o en lead o a no able educ ion in cycling ips compa ed o o he modes o
anspo , p ima ily because cyclis s a e di ec ly exposed o wea he pa ame e s, making
hem mo e ulne able o hei e ec s [
12
,
46
]. Nume ous s udies ha e in es iga ed he in lu-
ence o wea he pa ame e s on a el beha io . These s udies consis en ly demons a e ha
ad e se wea he condi ions discou age people om cycling, whe eas mode a e wea he
condi ions end o encou age i , making cycling a mo e appealing mode o anspo in mod-
e a e wea he condi ions. Empi ical e idence highligh s ha inclemen wea he condi ions
supp ess cycling ips o a g ea e ex en han o he a el modes, as cyclis s a e di ec ly
exposed o en i onmen al elemen s, ende ing hem mo e suscep ible o hei impac s.
Table 2summa izes he mos inno a i e esea ch examining he ela ionship be ween
clima ic ac o s and cycling beha io . These s udies unde sco e he c i ical ole ha wea he
plays in shaping u ban cycling ends and p o ide a ounda ion o de eloping p edic i e
models o p omo e cycling in di e se wea he condi ions.
U ban Sci. 2025,9, 41 3 o 30
Table 2. Key s udies on he impac o wea he on u ban cycling beha io .
Main Focus S udies
Insigh s in o how wea he a ec s cycling in No h Ame ica,
emphasizing immedia e and lagged e ec s. [55,63]
Mode a e empe a u es encou age highe pa icipa ion
in cycling. [64]
Real- ime wea he in o ma ion impac s cyclis s;
p ecipi a ion has immedia e e ec s on choices. [12,65]
Long- e m e ec s o seasonal wea he a ia ions on u ban
cycling pa icipa ion. [12,66]
Seasonal luc ua ions in u ban cycling ends; impo ance
o lexible policies and in as uc u e. [67]
Highe empe a u es and less p ecipi a ion linked o highe
cycling a es; wind and cold educe cycling. [68]
P edic i e models show ising empe a u es inc ease
cycling, pa icula ly du ing colde mon hs. [62,69]
Agen -based model showing impo ance o spa ially dense
wea he measu emen s compa ed o single s a ions. [11,70]
Table 2summa izes he s a is ical ela ionship be ween wea he condi ions and cycling
equency. The da a e eal signi ican a ia ions based on empe a u e and p ecipi a ion
le els, emphasizing he impo ance o wea he - esilien in as uc u e o cycling. This
analysis se es as a basis o he subsequen me hodological amewo k p esen ed in
his pape .
1.1. Li e a u e Re iew
The backg ound esea ch unde sco es he need o mo e p ecise wea he - ela ed
s udies on cycling. While p e ious esea ch has es ablished a ounda ional unde s anding
o he impac o wea he on cycling, i o en lacks he spa ial and empo al esolu ion
necessa y o de ailed u ban planning. Achie ing he es ablishmen o a sus ainable and
p ac ical mode o anspo , pa icula ly cycling, equi es con inuous moni o ing o cycling
demand h oughou he yea and analyzing how di e en wea he condi ions a ec he
numbe o cyclis s.
Recen s udies ha e unde sco ed he c i ical ole o cycling in ad ancing sus ainable u -
ban mobili y, wi h di e se app oaches o unde s anding i s in eg a ion in o u ban sys ems.
Hong e al. [
4
] demons a ed ha ai pollu ion con inues o nega i ely impac bike-sha ing
usage, e en du ing he COVID-19 pandemic, emphasizing he need o cleane u ban en i-
onmen s o p omo e cycling adop ion. Simila ly, Chen e al. [
11
] explo ed he non-linea
e ec s o spa ial and empo al ac o s on dockless bike-sha ing sys ems, highligh ing he
impo ance o adop ing localized app oaches in u ban cycling policies. Lan in e al. [
62
]
p o ided an upda ed analy ical model showcasing he non-linea e ec s o wea he condi-
ions on bicycle a ic, unde lining he complexi y o en i onmen al in luences on cycling
beha io . Valen ini e al. [
3
] a gued ha despi e cycling’s po en ial o deca boniza ion, i
emains unde u ilized due o gaps in go e nance and echnological in eg a ion, such as he
limi ed deploymen o e-bikes and bike-sha ing sys ems. We schmölle e al. [
24
] iden i ied
g ass oo s mo emen s as key d i e s in ins i u ionalizing cycling policies, demons a ing
hei in luence on enhancing cycling in as uc u e in Ge man ci ies. Gaio and Cugu-
ullo [
45
] wa ned o he po en ial ma ginaliza ion o cyclis s in u ban mobili y sys ems
domina ed by au onomous ehicles, ad oca ing o p oac i e policies o main ain cycling’s
U ban Sci. 2025,9, 41 4 o 30
ole in sus ainable anspo a ion. Finally, he wo k by Lan in e al. [
62
] highligh ed he
necessi y o ada -s yle amewo ks o e alua e he mul i- ace ed impac o cycling on
sus ainabili y. Collec i ely, hese s udies emphasize he need o in eg a ed app oaches ha
add ess social, en i onmen al, and policy dimensions, o ully ha ness cycling’s po en ial
in ega d o sus ainable u ban ansi ions.
These s udies p o ide a ounda ion o unde s anding he mul i-dimensional chal-
lenges o u ban cycling, which his pape seeks o add ess by de eloping a mo e in eg a ed
and da a-d i en amewo k o enhancing u ban cycling p ac ices. The e o e, his s udy
ocuses on de eloping a p edic i e model o moni o u ban bicycle usage based on key
wea he ac o s, examining hei in luence on cycling pa e ns in he ci y o Hambu g,
Ge many. In Ge many, app oxima ely 25 million bicycle ips we e made daily in 2002.
O e he pas 15 yea s, his numbe has inc eased o 28 million ips pe day. Fu he mo e,
he a e age dis ance a eled by bicycle inc eased by 20%, ising om 3.2 km in 2002 o
3.8 km in 2017 [71].
In Hambu g, he p opo ion o bicycle use g ew om sligh ly less han 10% o 15%
be ween 2002 and 2017 [
72
]. Addi ionally, pe capi a bicycle owne ship in Hambu g has
isen signi ican ly o e he pas 15 yea s [
73
]. Despi e hese ad ancemen s, he pe cen age
o cyclis s in Hambu g emains lowe han in o he ci ies wi h compa able wea he pa e ns,
such as Ams e dam (32%), Copenhagen (29%), Mon eal (18.2%), U ech (51%), An we p
(28.9%), Malmo (30%), and Hangzhou (30%) [
71
,
74
]. These igu es highligh he u gen
need o Hambu g o e alua e he ele an condi ions and implemen s a egies o inc ease
cycling a es, pa icula ly when compa ed o simila ly si ua ed ci ies. In a ecen s udy
ocusing on Hambu g, an agen -based model e ealed ha empe a u e discom o occu s
33% o he ime, while wind and p ecipi a ion discom o occu 5% o he ime [
69
,
70
].
These indings ein o ce he p o ound impac o wea he condi ions on u ban cycling
ends, which signi ican ly in luences he equency and du a ion o cycling ac i i ies in
Hambu g [
49
]. The subsequen sec ions will p o ide a comp ehensi e li e a u e e iew
in ega d o he speci ic wea he ac o s conside ed in his esea ch. Clima e change
exace ba es wea he condi ions, p esen ing bo h challenges and oppo uni ies o u ban
cycling. Un a o able wea he condi ions o en lead o ip escheduling, e ou ing, o
e en cancella ion, as no ed by Sabi [
5
,
61
]. P edic ed changes in wea he pa e ns, such
as wa me win e s and mo e ex eme wea he e en s, a e expec ed o signi ican ly al e
cycling pa e ns.
Al hough he e has been subs an ial esea ch on wea he pa ame e s and hei e ec s
on bicycle ips, mos s udies ha e analyzed hese ac o s indi idually, such as empe a u e,
p ecipi a ion, wind, sunligh , dayligh hou s, o seasonal a ia ions (Tables 1and 2). Each
s udy aims o unde s and speci ic aspec s o he ela ionship be ween wea he and cycling
globally, pa icula ly in e ms o how wea he condi ions impac cycling a es. Fo example,
Nosal and Mi anda-Mo eno [
75
] analyzed u ban bicycle acili ies in No h Ame ican
ci ies and highligh ed ha empe a u e and humidi y signi ican ly in luence cycling, o en
wi h non-linea ela ionships. They ound ha p ecipi a ion nega i ely a ec s cycling
lows, wi h g ea e in ensi y exace ba ing he impac . High empe a u es, hea y ain all,
and s ong winds we e shown o subs an ially educe cycling olumes. Simila ly, s udies
conduc ed in he Ne he lands e ealed ha ec ea ional cycling is mo e sensi i e o wea he
condi ions han u ili a ian cycling. A wea he model by K. Thomas e al. [
76
], demons a ed
ha empe a u e, p ecipi a ion, and wind signi ican ly in luence cycling demand, wi h
ec ea ional demand being mo e wea he sensi i e.
The main wea he pa ame e s ha no ably a ec cycling olumes include ai empe -
a u e (in deg ees Celsius), p ecipi a ion (in millime e s), sunligh du a ion (in hou s pe
day), humidi y, and wind speed (in kilome e s pe hou ) [
57
,
60
,
63
,
77
,
78
]. Despi e hese
U ban Sci. 2025,9, 41 5 o 30
indings, mos p io s udies ocus on single pa ame e s o analyze hem in isola ion. Ad-
d essing his gap, he p esen s udy in eg a es all i e main wea he pa ame e s, e alua ing
hei independen and in e dependen e ec s on cycling olumes. Fo he i s ime, his
esea ch aims o de elop a o mula o p edic ing cycling ends based on hese pa ame e s,
p o iding ac ionable insigh s o u ban planning and sus ainable anspo de elopmen .
Table 3below summa izes key s udies ela ed o hese i e wea he pa ame e s,
highligh ing hei ole in shaping u ban cycling ends.
Table 3. Summa y o s udies on key wea he pa ame e s in luencing u ban cycling.
Wea he Pa ame e Main Findings Key S udies
Tempe a u e
Rising empe a u es lead o an inc ease in he numbe o cyclis s. [63,66,79]
A 0.56
◦
C empe a u e inc ease in he Uni ed S a es esul s in a 3% ise
in he numbe o cyclis s. [80]
A 1 deg ee empe a u e inc ease in Vancou e leads o a 1.65%
inc ease in he numbe o cyclis s. [57]
A 1 deg ee empe a u e inc ease in Auckland esul s in a 3.2% inc ease
in he numbe o cyclis s. [63]
The ideal cycling empe a u e is 25–28 ◦C. [81]
Mos cycling ips occu wi hin he empe a u e ange om 26.7
◦
C o
31.7 ◦C, wi h empe a u es abo e 32 ◦C educing he numbe
o cyclis s.
[66]
Cyclis s a e sensi i e o empe a u es below 15 ◦C. [63,82]
High o low empe a u es can de e cycling, pa icula ly among
ec ea ional cyclis s. [40,49,53,83]
The “ideal” cycling empe a u e ange is be ween 17 ◦C and 33 ◦C. [35]
P ecipi a ion
Rain all is linked o a dec ease in cycling numbe s. [84]
Daily ain all o app oxima ely 8 mm esul s in a 50% educ ion in he
bicycle olume compa ed o ain- ee days. [81,85]
Fo e e y 1 mm o ain all, he e is a 10.6% dec ease in he numbe
o cyclis s. [63]
When daily ain all anges om 0.2 o 2 mm, he cyclis coun
dec eases by 8–19%. [85]
Rain all and empe a u e a iables signi ican ly educe cycling olume
in ci y a ea. [86]
Wind
An inc ease in wind speed has a ela i e, ad e se e ec on he numbe
o cyclis s and winds exceeding 5 km/h lead o a 17% dec ease in
bicycle ips; also, o e e y 1.6 km/h inc ease in wind speed, he e is a
5% dec ease in he numbe o cyclis s.
[47,66,87]
Va ious wind speeds ha e di e ing impac s on cycling. [76,88,89]
Wind speeds anging om 25 o 52 km/h esul in an 11% o 23%
educ ion in he cyclis coun . [85]
Humidi y
An inc ease in humidi y has been associa ed wi h a dec ease in cycling.
[41,47,90]
A one pe cen inc ease in humidi y esul s in a 0.08 pe cen dec ease in
bicycle a ic in Vancou e . [57]
Dayligh
The e is a di ec ela ionship be ween he hou s in he day and he
numbe o cyclis s. [91,92]
Dayligh e ec appea s o be s a is ically insigni ican o e y low. [78,93]

U ban Sci. 2025,9, 41 6 o 30
Table 3demons a es how empe a u e, p ecipi a ion, wind, humidi y, and dayligh
hou s impac cyclis numbe s, p o iding a comp ehensi e o e iew o he en i onmen al
ac o s a ec ing u ban cycling pa e ns.
1.1.1. Tempe a u e
Tempe a u e is one o he mos signi ican wea he pa ame e s in luencing cycling
ac i i y [
35
]. Mode a e empe a u es a e gene ally a o able, while ex eme empe a u es,
ei he ho o cold, end o de e cyclis s [
68
]. A s udy by Schmi e al. [
70
] ound a non-linea
ela ionship be ween empe a u e and cycling a es, indica ing ha highe empe a u es
and lowe p ecipi a ion le els co ela e wi h inc eased cycling ac i i y. This highligh s he
impo ance o empe a u e as a c i ical ac o in shaping u ban cycling ends.
1.1.2. P ecipi a ion
P ecipi a ion, including ain and snow, is among he mos in luen ial wea he a iables
a ec ing cycling decisions, p ima ily due o sa e y conce ns and discom o [
47
,
66
]. In-
c eased p ecipi a ion ele a es he likelihood o acciden s and makes cycling less enjoyable.
Resea ch conduc ed in Canadian ci ies ound ha an inc ease in ainy and eezing days
signi ican ly educed he annual numbe o cyclis s [
94
]. O he s udies [
77
,
78
,
85
,
95
,
96
] u -
he emphasize he non-linea impac o ain on bicycle olume, iden i ying p ecipi a ion
as a p ima y de e en o cycling beha io .
1.1.3. Wind Speed
Wind speed plays a c ucial ole in cycling ac i i y, as i di ec ly a ec s he physical
e o equi ed o cycling. S ong headwinds a e especially de imen al, while ailwinds
ha e a less p onounced nega i e impac [
80
]. A s udy by Böcke e al. [
92
] demons a ed
ha he a e age daily wind speed nega i ely impac s cycling du a ion wi h 90% con idence,
al hough i does no signi ican ly a ec he numbe o cyclis s. Addi ionally, a machine
lea ning app oach by Ma ocks [
97
] showed he easibili y o in eg a ing wind speed
in o p edic i e models o bicycle usage, unde lining he impo ance o his pa ame e in
o ecas ing cycling demand.
1.1.4. Humidi y
The e ec o humidi y on cycling beha io a ies depending on he geog aphic con-
ex [
76
]. While less s udied han empe a u e o p ecipi a ion, high humidi y can lead o
discom o and excessi e pe spi a ion, educing he appeal o cycling [
41
,
90
]. These s udies
sugges ha humidi y, al hough o en o e looked, is a no able ac o in cycling beha io ,
pa icula ly in humid clima es.
1.1.5. Dayligh
The dayligh du a ion signi ican ly in luences cycling pa e ns. Longe dayligh
hou s no only encou age mo e equen ides, bu also imp o e isibili y and sa e y,
os e ing highe bicycle usage [
78
,
91
]. Resea ch on bike-sha ing sys ems has demons a ed
ha dayligh hou s play a c ucial ole in shaping usage pa e ns [
63
,
92
]. These indings
unde line he necessi y o inco po a ing he dayligh du a ion in o p edic i e models
o be e unde s and and op imize u ban cycling beha io . This s udy p esen s a no el
in eg a ion o i e key wea he pa ame e s, namely empe a u e, p ecipi a ion, humidi y,
wind speed, and sunligh du a ion, in o a single non-linea p edic i e model. While p io
esea ch o en ocuses on one o wo wea he a iables, ou app oach add esses hei
combined in luence on bicycle usage. Addi ionally, we conduc a de ailed co ela ion
analysis o e alua e po en ial in e dependencies among hese a iables, ensu ing he
obus ness o ou me hodology.
U ban Sci. 2025,9, 41 7 o 30
The i e wea he pa ame e s a e selec ed based on hei p o en impac on cycling
beha io in p io s udies. This esea ch in eg a es hese pa ame e s o de elop a obus
p edic i e model o moni o ing u ban cycling ends.
This s udy in es iga es bo h he independen and combined e ec s o wea he pa am-
e e s on cycling beha io . By employing ad anced da a analy ics and his o ical da ase s,
a p edic i e model is de eloped o moni o and analyze luc ua ions in bicycle a ic in
a ying wea he condi ions. The model’s accu acy is alida ed using his o ical da a and
simula ions ac oss selec ed yea s, ensu ing i s obus ness. Ul ima ely, he indings p o ide
ac ionable insigh s o add ess wea he - ela ed challenges, op imize u ban in as uc u e,
and p omo e cycling as a sus ainable mode o anspo a ion, wi h a ocus on Hambu g’s
wea he pa e ns.
2. Ma e ials and Me hods
This s udy u ilizes an in eg a ed me hodology o analyze he ela ionship be-
ween wea he pa ame e s and cycling beha io , building on es ablished esea ch ame-
wo ks [
11
,
12
,
62
]. P io s udies ha e explo ed he impac o wea he condi ions on u ban
cycling using a ious analy ical app oaches. Schmi e al. [
70
] highligh ed he non-linea
ela ionship be ween empe a u e and cycling a es, emphasizing he ole o mode a e
wea he condi ions in p omo ing cycling. Lan in e al. [
62
] demons a ed he signi i-
can in luence o empe a u e and humidi y on cycling beha io in ci ies like Vancou e ,
while s udies on p ecipi a ion [
94
] ha e consis en ly shown i s nega i e impac on cycling
olumes, pa icula ly du ing ainy condi ions.
Wind speed, a c i ical pa ame e , has been analyzed in s udies such as ha o Böcke
e al. [
92
], which ocused on he ad e se e ec s on cycling du a ion, pa icula ly in Ge man
ci ies. Ma ocks [
97
] demons a ed he in eg a ion o wind speed in o p edic i e machine
lea ning models, emphasizing i s impo ance in o ecas ing cycling demand [
11
]. These
s udies p o ided he ounda ion o including wind speed as a a iable in ou model.
Dayligh du a ion has also been iden i ied as a signi ican ac o in shaping cycling
pa e ns, whe e longe dayligh hou s we e posi i ely associa ed wi h inc eased cycling
ac i i y [12,24].
Building on insigh s om di e se geog aphic con ex s, including Ge many, Canada,
and New Zealand, his s udy inco po a es empe a u e, p ecipi a ion, wind speed, humid-
i y, and dayligh du a ion in o a obus p edic i e model. By le e aging ad anced da a
analy ics and p io esea ch, he model cap u es he nuanced ela ionships be ween hese
clima ic ac o s and cycling beha io . I e alua es bo h independen and in e dependen
e ec s on cycling ends and alida es i s applicabili y using his o ical da ase s om Ham-
bu g. Addi ionally, he use o p edic i e models and machine lea ning echniques p o ides
ac ionable insigh s o enhancing u ban cycling in as uc u e ac oss a ied en i onmen al
con ex s. Fo ins ance, in Be lin, p edic i e models es ima e an annual inc ease o 1–4%
in cycling a ic due o ising empe a u es, wi h he mos signi ican g ow h an icipa ed
du ing he win e mon hs [69].
This s udy u ilizes da a om wo base yea s (2017 and 2019) and applies a non-linea
eg ession model o p edic cycling ac i i y. S a is ical analyses and da a isualiza ions
alida e he model’s eliabili y and accu acy, wi h addi ional es ing conduc ed o subse-
quen yea s (2021 and 2023). The me hodology is designed o ensu e obus p edic ions
and o es ablish he model’s lexibili y o b oade u ban applica ions. The me hods and
models de eloped o his esea ch ocus no only on he i e key clima ic ac o s, namely
empe a u e, p ecipi a ion, wind speed, humidi y, and dayligh du a ion, bu also allow
o he po en ial in eg a ion o o he a iables, such as ai quali y, a ic condi ions, and
socioeconomic ac o s. This adap abili y enhances he model’s u ili y o u ban planne s
U ban Sci. 2025,9, 41 8 o 30
and policymake s seeking o p omo e sus ainable anspo a ion solu ions, ailo ed o
di e se u ban con ex s.
This s udy adop s a s ep-by-s ep me hodology, which includes he ollowing
componen s
:
•
Co ela ion analysis and model design: The iden i ica ion o s ong co ela ions be-
ween clima ic ac o s and cycling ends using da a om he base yea s 2017 and 2019,
ollowed by he design o a p edic i e model;
•
E alua ion o wea he impac s on cycling olume: The assessmen o a ious wea he
condi ions, namely empe a u e, p ecipi a ion, wind speed, humidi y, and dayligh
du a ion, on cycling olumes using SPSS;
•
S a is ical analysis and model alida ion: The analysis o cycling ends using s a is ical
me hods such as Pea son co ela ion coe icien s, ANOVA es s, and R-squa e alues.
The accu acy and ele ance o he model a e e alua ed using da a om he yea 2019;
•
De elopmen and es ing o p edic i e models: The c ea ion o p edic i e models o
bicycle ip olumes based on wea he da a, wi h u he alida ion pe o med using
independen da a om 2021 and 2023. These yea s a e in en ionally excluded om
he model design p ocess o ensu e unbiased accu acy es ing.
By le e aging his o ical wea he and cycling da a, his esea ch aims o de elop a
obus , adap able model o p edic ing u ban cycling ends. The in eg a ion o ad anced
analy ical me hods ensu es he model’s p ecision and applicabili y, p o iding ac ionable
insigh s o suppo he p omo ion o cycling as a sus ainable u ban anspo a ion solu ion.
2.1. Case S udy Backg ound
In his s udy, he ci y o Hambu g is selec ed as a case s udy. Loca ed a la i ude
53
◦
33
′
3.9096
′ ′
N and longi ude 9
◦
59
′
37.2552
′ ′
E, Hambu g had a popula ion o app oxi-
ma ely 1.8 million esiden s in 2017, he i s yea o he examina ion (Whe e Is Hambu g,
Ge many on Map La Long Coo dina es; [
72
]). Below is a b ie o e iew o Hambu g’s key
cha ac e is ics and u ban s uc u e.
2.1.1. Modal Spli in Hambu g
Be ween 2002 and 2017, Hambu g wi nessed no able shi s in i s modal spli . The p o-
po ion o bicycle use inc eased om nea ly 10% o 15%, while public anspo usage ose
modes ly om less han 20% o 22%. Pedes ian a el emained s able a app oxima ely
27%. In con as , he sha e o p i a e ehicle use dec eased signi ican ly, om 47% o 36%.
These ends, e lec an inc easing p e e ence among Hambu g esiden s o bicycles and
public anspo o e p i a e ehicles du ing his pe iod [72].
2.1.2. Ele a ion and Topog aphy in Hambu g
Acco ding o he opog aphic map o Hambu g (Figu e 1), he ci y has a minimum
ele a ion o
−
3 m, a maximum ele a ion o 150 m, and an a e age ele a ion o 23 m.
This ela i ely la opog aphy makes Hambu g highly sui able o cycling [
98
]. Gi en
ha opog aphy is a c i ical ac o in luencing cycling beha io , Hambu g o e s a o able
condi ions o p omo ing bicycle usage.
Figu e 1depic s he opog aphy o Hambu g, highligh ing i s ela i ely la e ain,
which p o ides a o able condi ions o p omo ing cycling beha io in he ci y.
2.1.3. Cu en S a e o Hambu g’s Bicycle In as uc u e
Hambu g’s Velo ou e ne wo k consis s o 14 ci ywide cycling ou es, spanning a o al
o app oxima ely 280 km, o which a ound 80 km ha e been comple ed o da e. As pa
o i s ision o become a bicycle- iendly ci y, he Hambu g Sena e has announced plans
o de elop addi ional cycling pa hs and op imize exis ing ou es. These ini ia i es aim o
U ban Sci. 2025,9, 41 9 o 30
enhance he accessibili y and usabili y o cycling in as uc u e h oughou he ci y. Fo a
isual o e iew o Hambu g’s Velo ou e ne wo k, see Figu e 2.
U banSci.2025,9,xFORPEERREVIEW8o 29

2.1.CaseS udyBackg ound
In hiss udy, heci yo Hambu gisselec edasacases udy.Loca eda la i ude
53°33′3.9096′′Nandlongi ude9°59′37.2552′′E,Hambu ghadapopula iono app oxi-
ma ely1.8million esiden sin2017, he i s yea o  heexamina ion(Whe eIsHambu g,
Ge manyonMapLa LongCoo dina es;[72]).Belowisab ie o e iewo Hambu g’s
keycha ac e is icsandu bans uc u e.
2.1.1.ModalSpli inHambu g
Be ween2002and2017,Hambu gwi nessedno ableshi sini smodalspli .The
p opo iono bicycleuseinc eased omnea ly10% o15%,whilepublic anspo usage
osemodes ly omless han20% o22%.Pedes ian a el emaineds ablea app oxi-
ma ely27%.Incon as , hesha eo p i a e ehicleusedec easedsigni ican ly, om47%
o36%.These ends, e lec aninc easingp e e enceamongHambu g esiden s o bicy-
clesandpublic anspo o e p i a e ehiclesdu ing hispe iod[72].
2.1.2.Ele a ionandTopog aphyinHambu g
Acco ding o he opog aphicmapo Hambu g(Figu e1), heci yhasaminimum
ele a iono −3m,amaximumele a iono 150m,andana e ageele a iono 23m.This
ela i ely la  opog aphymakesHambu ghighlysui able o cycling[98].Gi en ha  o-
pog aphyisac i ical ac o in luencingcyclingbeha io ,Hambu goffe s a o ablecon-
di ions o p omo ingbicycleusage.

Figu e1.Topog aphyo Hambu g[98].
Figu e1depic s he opog aphyo Hambu g,highligh ingi s ela i ely la  e ain,
whichp o ides a o ablecondi ions o p omo ingcyclingbeha io in heci y.
2.1.3.Cu en S a eo Hambu g’sBicycleIn as uc u e
Hambu g’sVelo ou ene wo kconsis so 14ci ywidecycling ou es,spanninga o-
alo app oxima ely280km,o whicha ound80kmha ebeencomple ed oda e.Aspa 
o i s ision obecomeabicycle- iendlyci y, heHambu gSena ehasannouncedplans
Figu e 1. Topog aphy o Hambu g [98].
U banSci.2025,9,xFORPEERREVIEW9o 29

ode elopaddi ionalcyclingpa hsandop imizeexis ing ou es.Theseini ia i esaim o
enhance heaccessibili yandusabili yo cyclingin as uc u e h oughou  heci y.Fo a
isualo e iewo Hambu g’sVelo ou ene wo k,seeFigu e2.

Figu e2.Thecu en s a eo Hambu g’sbicyclein as uc u e[98].
Figu e2illus a es hecu en s a eo Hambu g’sbicyclein as uc u e,highligh ing
key ou esand hei dis ibu ionac oss heci y’sdis ic s.
Toensu e obus nessand eliabili y, heda awe ecollec ed omHambu g’scycling
moni o ingne wo k,whichincludesau oma icg oundcoun e sac oss heci y.Theda-
ase spans ou yea s(2017,2019,2021,and2023),p o iding48mon hlyobse a ions.
Thisapp oachensu essufficien samplesize,add essingconce nsabou da a eliabili y.
Fu u ewo kcouldexplo ehighe  equencyda ase s o enhancedg anula i y.
2.1.4.A e ageWea he Condi ionsinHambu g
Hambu g’sclima eischa ac e izedbycom o ablesumme s,wi hpa lycloudy
skies,andlong,cold,ando enwindy,win e s.Annually, empe a u es ypically ange
om−1°C o23°C, a elyd oppingbelow−9°Co exceeding29°C.Humidi yle els
emain ela i elys able h oughou  heyea , luc ua ingbe ween67%and90%.Thewind-
ies pe iod ea u esa e agewindspeedsexceeding11.2milespe hou [99,100].
Fo acompa a i eanalysis,Figu e3illus a es heclima ic ac o s(meanai  empe -
a u e,meanhumidi y,p ecipi a ion,dayligh du a ion,andmeanwindspeed)ac oss he
yea s2017,2019,2021,and2023,alongside he olumeo bicycle ipsdu ing hesame
imepe iods.
Figu e 2. The cu en s a e o Hambu g’s bicycle in as uc u e [98].
Figu e 2illus a es he cu en s a e o Hambu g’s bicycle in as uc u e, highligh ing
key ou es and hei dis ibu ion ac oss he ci y’s dis ic s.
To ensu e obus ness and eliabili y, he da a we e collec ed om Hambu g’s cycling
moni o ing ne wo k, which includes au oma ic g ound coun e s ac oss he ci y. The da ase
spans ou yea s (2017, 2019, 2021, and 2023), p o iding 48 mon hly obse a ions. This
U ban Sci. 2025,9, 41 16 o 30
•Humidi y has he weakes nega i e co ela ion ( = −0.744, p< 0.01), indica ing
i s lesse , bu s ill measu able, impac on cycling;
•
P ecipi a ion shows a mode a e nega i e co ela ion ( =
−
0.515, p< 0.05), educ-
ing cycling ac i i y;
•
Wind speed exhibi s a no iceable nega i e co ela ion ( =
−
0.696, p< 0.05),
e lec ing he challenges posed by windy condi ions.
3. Co a iance Analysis:
•
Tempe a u e and dayligh oge he s ongly in luence cycling olumes, explain-
ing a signi ican po ion o he a iance in he numbe o bicycle ips and
ein o cing hei impo ance o p edic i e modeling.
To u he e alua e he impac o indi idual clima ic ac o s on bicycle ip olumes
in 2017, an ANOVA es and R-squa e analysis a e conduc ed. In his analysis, bicycle
ips a e conside ed he dependen a iable, while he wea he pa ame e s se e as he
independen a iables. The esul s, summa ized in Table 10, p o ide insigh s in o he
ela i e impo ance o each pa ame e in in luencing cycling beha io .
Table 10. ANOVA es ca ied ou in SPSS.
Dependen Va iable: Bicycle T ip Volume 2017
Model Summa y ANOVA
R R Squa e Adjus ed R
Squa e
S d. E o o
Es ima e F Sig.
Independen a iable
Mean ai empe a u e (oC) 0.945 0.893 0.870 3,445,632.512
37.742
0.000
Mean humidi y (%) 0.721 0.520 0.413 7,314,097.367
4.875
0.037
Mean p ecipi a ion (mm) 0.575 0.331 0.182 8,635,688.334
2.225
0.164
Mean dayligh du a ion
(Minu e) 0.949 0.900 0.878 3,333,413.408
40.634
0.000
Mean wind speed (kph) 0.142 0.020 0.197 10,449,116.286
0.093
0.912
The key obse a ions om Table 10 a e as ollows:
4. R-Squa e Values:
•
Dayligh has he highes R-squa e alue (0.900), meaning i explains he la ges
p opo ion o a iance in bicycle ip olumes;
•
Wind speed has he lowes R-squa e alue (0.020), indica ing i s minimal in lu-
ence on cycling ends;
5. F-Ra e Analysis:
•
Dayligh has he highes F a e (40.634, p< 0.01), con i ming i s signi ican impac
on he amoun o bicycle ips;
•
Wind speed: has he lowes F a e (0.093, p= 0.912), indica ing a negligible e ec
on he dependen a iable;
6. Signi icance Le els:
•
Tempe a u e and dayligh bo h ha e p- alues less han 0.05, indica ing hei
s a is ically signi ican impac on cycling ac i i y;
•
P ecipi a ion, humidi y, and wind speed do no show any s a is ically signi ican
e ec s, wi h p- alues g ea e han 0.05.
3.2. Analysis o Co a iance
The analysis o co a iance, summa ized in Table 11, e alua es he indi idual and
combined e ec s o empe a u e and dayligh du a ion on he olume o bicycle ips. The
esul s highligh he ollowing key indings:

U ban Sci. 2025,9, 41 17 o 30
•Indi idual Impac s:
Table 11. Tes s o be ween-subjec e ec s (dependen a iable: bicycle ip olume 2017).
Sou ce Type III Sum o Squa es d Mean Squa e F Sig. Pa ial E a
Squa ed
Co ec ed Model 949,932,721,313,230.000 a3
316,644,240,437,743.300
47.714 0.000 0.947
In e cep 357,175,731,059.297 1 357,175,731,059.297 0.054 0.042 0.029
Tempe a u e
(◦C) 7,164,973,489,802.059 1 7,164,973,489,802.059 1.080 0.020 0.471
Dayligh (Minu e) 23,781,541,670,782.650 1
23,781,541,670,782.650
3.584 0.014 0.505
Tempe a u e * Dayligh 953,712,204.422 1 953,712,204.422 0.000 0.991 0.000
E o 53,090,519,077,594.910 8
To al 9,995,710,685,640,824.000 12
Co ec ed To al 1,003,023,240,390,824.900 11
aR2= 0.947 (adjus ed R2= 0.927).
Bo h he empe a u e and dayligh du a ion ha e s a is ically signi ican indi idual
e ec s on cycling olumes, wi h p- alues o 0.020 and 0.014, espec i ely;
The Pa ial E a Squa ed alues indica e ha dayligh du a ion (0.505) has a sligh ly
g ea e in luence on bicycle ips han empe a u e (0.471);
•Combined E ec s:
The in e ac ion e m ( empe a u e * dayligh ) has a p- alue g ea e han 0.05 (0.991),
indica ing ha he wo pa ame e s do no collec i ely in luence cycling olumes. This
sugges s ha hei impac s a e la gely independen .
The key akeaway poin s om Table 11 a e as ollows:
•
Dayligh ’s dominance: Dayligh du a ion has he g ea es impac on cycling olumes,
wi h he highes R-squa e alue (0.900) and Pa ial E a Squa ed (0.505);
•
Tempe a u e’s ole: Tempe a u e also signi ican ly in luences cycling olumes, al-
hough i s e ec (Pa ial E a Squa ed = 0.471) is sligh ly less han dayligh ’s;
•
Minimal impac o wind speed: Wind speed has a minimal e ec on cycling beha io ,
as shown by i s low R-squa e alue (0.020) and F a e;
•
Independen impac s: Tempe a u e and dayligh du a ion impac cycling indepen-
den ly, wi h no signi ican combined e ec (p= 0.991).
These indings ein o ce he impo ance o dayligh du a ion and empe a u e in d i -
ing seasonal cycling ends and can be used o in o m u u e p edic i e modeling e o s.
The analysis highligh s he a ying deg ees o in luence each wea he pa ame e
exe s on cycling ends in Hambu g. Dayligh du a ion and empe a u e eme ged as he
mos signi ican ac o s, consis en ly d i ing seasonal cycling ac i i y, while wind speed
and p ecipi a ion exhibi ed compa a i ely weake impac s. The independen na u e o
empe a u e and dayligh du a ion e ec s sugges s he need o a ge ed in e en ions o
maximize cycling’s po en ial in a o able condi ions.
4. Discussion
Acco ding o he classi ica ion and unde s anding o di e en ela ionships be ween
wea he pa ame e s and hei espec i e s eng hs and weaknesses in in luencing cycling
ac i i y, i becomes e iden ha p edic ing cycling ends o e ime is essen ial. This
s udy add esses his need by ocusing on he yea 2021 and es ablishing Equa ion (1). The
coe icien s o his equa ion, as de ailed in Table 12, demons a e hei high p ecision in
p edic ing he numbe o bicycle ips.
U ban Sci. 2025,9, 41 18 o 30
Table 12. In es iga ing he ela ionship be ween he numbe o ips made by bicycle in he 12 mon hs
o 2017 and he ela ed ma hema ical models.
Pa ame e Es ima ed Sig. Uppe CI Lowe CI
In e cep α+6130 2.840 0.000 +6727 +5874
Tmean β1+9971.9 1.758 0.000 +10,251 +9746
Hmean β2+80 0.102 0.000 +142 +39
Pmean β3−180 −0.647 0.000 −324 −64
DLmean β4+1358.25 1.068 0.000 +1516 +1241
WSmean β5−15,582 −0.266 0.000 −28,321 −12,768
Tmean2γ−317.241 −0.192 0.000 −451 −283
This p oposed equa ion se es as a p ac ical ool o es ima ing he mon hly olume o
bicycle ips by inco po a ing key wea he pa ame e s. I allows o a deepe unde s anding
o how di e en wea he condi ions impac cycling and p o ides a eliable amewo k o
o ecas ing ends in a ious selec ed ime pe iods.
The ela ionship be ween wea he pa ame e s and bicycle ips is modeled using
Equa ion (1), which inco po a es key wea he a iables, such as empe a u e, humidi y,
p ecipi a ion, dayligh du a ion, and wind speed, along wi h a squa ed e m o empe a-
u e o accoun o i s non-linea e ec s. The coe icien s o hese a iables a e de ailed in
Table 12, demons a ing hei es ima ed impac s, signi icance, and con idence in e als.
Equa ion (1): Mon hly P edic ion
BTV = (α+β1×Tmean −β2×Hmean +β3×Pmean +β4×DLmean +β5×WSmean +γ×(Tmean)2) (1)
•BTV =bicycle ip olume;
•α= in e cep ;
•β1— eg ession o nume ical a iables ela ed o mean ai empe a u e;
•β2— eg ession o nume ical a iables ela ed o mean humidi y;
•β3— eg ession o nume ical a iables ela ed o p ecipi a ion;
•β4— eg ession o nume ical a iables ela ed o dayligh ;
•β5— eg ession o nume ical a iables ela ed o mean wind speed;
•γ— eg ession (mean mon hly ai empe a u e);2
•DLmean—means mon hly dayligh du a ion [minu e];
•Tmean—means mon hly ai empe a u e [◦C];
•Hmean—means mon hly humidi y [%];
•WSmean—means mon hly wind speed [kph];
•Pmean—means mon hly p ecipi a ion [mm].
Based on he da a in Table 12, he ollowing indings a e made:
•
In e cep (
α
): Rep esen s he baseline olume o bicycle ips when all he clima ic
ac o s a e held cons an (+6130);
•
Tempe a u e (
β1
) and squa ed e m (
γ
): Tempe a u e posi i ely impac s cycling ac i -
i y (+9971.9), wi h diminishing e ec s a ex eme alues (−317.241);
•
Humidi y (
β2
): A small posi i e impac (+80), e lec ing i s minimal, bu
measu able, in luence;
•
P ecipi a ion (
β3
): A nega i e coe icien (
−
180) indica es a de e en e ec o ain all
on cycling ac i i y;
•
Dayligh (
β4
): A signi ican posi i e impac (+1358.25), con i ming he ole o longe
dayligh hou s in p omo ing cycling;
•
Wind speed (
β5
): The s onges nega i e impac (
−
15,582), showing ha high wind
speeds de e cycling.
U ban Sci. 2025,9, 41 19 o 30
The es ablished Equa ion (1) p o ides a p edic i e amewo k o es ima ing he
mon hly olume o bicycle ips (BTV) based on key wea he pa ame e s, including em-
pe a u e, humidi y, p ecipi a ion, dayligh du a ion, and wind speed. Each pa ame e ’s
eg ession coe icien , as de ailed in Table 12, quan i ies i s indi idual impac on cycling
ac i i y. Addi ionally, he squa ed e m o empe a u e accoun s o i s non-linea e ec
on bicycle ip olumes.
To enhance he p ac icali y o he model, Equa ion (2) is de i ed by mul iplying
he mon hly p edic ions om Equa ion (1) by he numbe o days in each mon h. This
ex ension enables he p edic ion o he a e age daily bicycle ip olume (BTVD). Using
hese equa ions, he numbe o bicycle ips in Hambu g o he yea 2017 is p edic ed, wi h
he esul s summa ized in Table 13. The p edic ed mon hly olumes align closely wi h he
ac ual obse ed da a, demons a ing he obus ness and accu acy o he model.
Table 13. A mon hly a e age o he p edic ed numbe o bicycle ips made in Hambu g in 2017.
Yea Jan Feb Ma Ap May Jun
Bicycle
olume 2017
17,103,710 18,853,286 26,872,945 30,570,713 37,959,189 38,521,977
Jul Aug Sep Oc No Dec
39,588,698 35,168,670 29,305,063 23,432,668 18,436,359 14,337,612
Equa ion (2): Daily P edic ion, o es ima e daily ips, Equa ion (1) is mul iplied by
he numbe o days in each mon h:
BTVD = BTV ×Day (2)
•BTVD = bicycle ip olume pe day;
•Day = numbe o days in he mon h.
To alida e he model, he co ela ion be ween eal and p edic ed bicycle ips in 2017
is assessed. The esul s, p esen ed in Table 14, indica e a Pea son co ela ion coe icien o
0.971 (p< 0.01). This s ong posi i e co ela ion highligh s he p ecision o he model in
cap u ing he ela ionship be ween wea he pa ame e s and bicycle ips, con i ming i s
sui abili y o p edic i e pu poses.
Table 14. Co ela ion be ween he numbe o bicycle ips made in 2017 and he numbe o p edic ed
bicycle ips in 2017, based on Table 13.
Co ela ions Bicycle T ips in 2017 P edic ed Bicycle
T ips 2017
Bicycle ips in 2017
Pea son Co ela ion 1 0.971 **
Sig. (2- ailed) 0.000
N 12 12
P edic ed bicycle
ips 2017
Pea son Co ela ion 0.971 ** 1
Sig. (2- ailed) 0.000
N 12 12
** The co ela ion is signi ican a he 0.01 le el (2- ailed).
The R-squa e alue equals 0.942, acco ding o Table 15, indica ing a high le el o
accu acy o he p oposed model in p edic ing he numbe o bicycle ips (cycling) in he
ci y o Hambu g. Addi ionally, he adjus ed R-squa e alue is posi i e and close o he
maximum amoun (one) and p esen s he signi ican impac o he wea he pa ame e s on
he olume o cyclis s. Figu e 6 u he co obo a es he high accu acy o he simula ion.
U ban Sci. 2025,9, 41 20 o 30
Table 15. Reg ession coe icien s o mon hly bicycle ip olume in 2017 (R2= 0.942).
Model Summa y
R * R Squa e ** Adjus ed R Squa e ***
0.971 0.942 0.929
*−1<R<+1;**0<R2< 1; *** 0 < Adjus ed R2< 1.
U banSci.2025,9,xFORPEERREVIEW20o 29

Figu e6.Thesigni icanceo  he ela ionshipbe ween henumbe o  ealcycling ipsand henum-
be o p edic edcycling ipsin2017.
Buildingon hesuccessand alida ionin e mso Equa ion(2)inp edic ingbicycle
ipsin2017, hesameequa ionisapplied o o ecas  henumbe o bicycle ipsin2019,
usingwea he da a om ha yea .The esul sin e mso  hisp edic iona ede ailedin
Table16,whichp o ides hemon hlyp edic edbicycle ips o 2019.
Table16.Themon hlya e ageo  hep edic ednumbe o bicycle ipsmadeinHambu gin2019.
Yea JanFebMa Ap MayJun
Bicycle ip ol-
ume2019
15,976,84919,834,91925,413,94532,117,80937,542,24439,943,562
Jul Aug Sep Oc No Dec
39,621,249 36,079,256 29,451,970 24,823,443 19,225,287 15,937,756
Toe alua e heaccu acyo  hissimula ion, heco ela ioncoefficien be ween he
ac ualnumbe o bicycle ipsand hep edic ed aluesiscalcula ed.Asp esen edinTa-
ble17,aPea sonco ela ioncoefficien o 0.961(p<0.01)indica esas ongandsigni ican 
ela ionshipbe ween heac ualandp edic edda a.Thishighle elo ag eemen  u he 
alida es he obus nessand eliabili yo  hep edic i emodelac ossdiffe en yea sand
wea he condi ions .
Table17.Co ela ionbe ween henumbe o bicycle ipsmadein2019and henumbe o p edic ed
bicycle ipsin2019.
Co ela ionsBicycleT ipsin
2019
P edic edBicycleT ips
2019
Bicycle ipsin
2019
Pea sonCo ela ion10.961**
Sig.(2- ailed) 0.000
N1212
P edic edbicycle
ips 2019
Pea sonCo ela ion0.961**1
Sig.(2- ailed)0.000
N1212
**Theco ela ionissigni ican a  he0.01le el(2- ailed).
Figu e 6. The signi icance o he ela ionship be ween he numbe o eal cycling ips and he numbe
o p edic ed cycling ips in 2017.
The p edic i e accu acy o he p oposed model is u he alida ed by i s R-squa e
alue, which equals 0.942 (as shown in Table 15). This high alue demons a es he model’s
excep ional abili y o explain a ia ions in he numbe o bicycle ips based on he wea he
pa ame e s. Fu he mo e, he adjus ed R-squa e alue, which is posi i e and close o one,
con i ms he signi ican con ibu ion o wea he pa ame e s o he p edic i e model and
ensu es i s eliabili y in di e en scena ios.
Addi ionally, Figu e 6illus a es he s ong ag eemen be ween eal and p edic ed
alues o he numbe o bicycle ips in 2017, isually co obo a ing he model’s high le el
o p ecision. This alignmen highligh s he obus ness o he p oposed equa ions and hei
applicabili y o p edic ing cycling ac i i y in Hambu g.
Building on he success and alida ion in e ms o Equa ion (2) in p edic ing bicycle
ips in 2017, he same equa ion is applied o o ecas he numbe o bicycle ips in 2019,
using wea he da a om ha yea . The esul s in e ms o his p edic ion a e de ailed in
Table 16, which p o ides he mon hly p edic ed bicycle ips o 2019.
Table 16. The mon hly a e age o he p edic ed numbe o bicycle ips made in Hambu g in 2019.
Yea Jan Feb Ma Ap May Jun
Bicycle ip
olume 2019
15,976,849 19,834,919 25,413,945 32,117,809 37,542,244 39,943,562
Jul Aug Sep Oc No Dec
39,621,249 36,079,256 29,451,970 24,823,443 19,225,287 15,937,756
To e alua e he accu acy o his simula ion, he co ela ion coe icien be ween he
ac ual numbe o bicycle ips and he p edic ed alues is calcula ed. As p esen ed in
Table 17, a Pea son co ela ion coe icien o 0.961 (p< 0.01) indica es a s ong and signi ican
ela ionship be ween he ac ual and p edic ed da a. This high le el o ag eemen u he
U ban Sci. 2025,9, 41 21 o 30
alida es he obus ness and eliabili y o he p edic i e model ac oss di e en yea s and
wea he condi ions.
Table 17. Co ela ion be ween he numbe o bicycle ips made in 2019 and he numbe o p edic ed
bicycle ips in 2019.
Co ela ions Bicycle T ips in 2019 P edic ed Bicycle
T ips 2019
Bicycle ips in 2019
Pea son Co ela ion 1 0.961 **
Sig. (2- ailed) 0.000
N 12 12
P edic ed bicycle
ips 2019
Pea son Co ela ion 0.961 ** 1
Sig. (2- ailed) 0.000
N 12 12
** The co ela ion is signi ican a he 0.01 le el (2- ailed).
Addi ionally, as p esen ed in Table 18, he R-squa e alue o 0.924 u he suppo s
he model’s accu acy in explaining a ia ions in cycling ac i i y. The adjus ed R-squa e
alue o 0.907 ein o ces he model’s eliabili y, conside ing he numbe o p edic o s and
he sample size. These s a is ical measu es indica e ha he p oposed model e ec i ely
cap u es he in luence o wea he pa ame e s on he numbe o bicycle ips. Figu e 7
isually compa es he p edic ed and ac ual cycling numbe s o 2019, showing hei s ong
alignmen and u he co obo a ing he high p ecision o he simula ion.
Table 18. Reg ession coe icien s o mon hly bicycle ip olume in 2019 (R2= 0.942).
Model Summa y
R * R Squa e ** Adjus ed R Squa e *** S d. E o o he
Es ima e
0.961 0.924 0.907 2,840,489.903
*−1<R<+1;**0<R2< 1; *** 0 < Adjus ed R2< 1.
U banSci.2025,9,xFORPEERREVIEW21o 29

Addi ionally,asp esen edinTable18, heR-squa e alueo 0.924 u he suppo s
hemodel’saccu acyinexplaining a ia ionsincyclingac i i y.Theadjus edR-squa e
alueo 0.907 ein o ces hemodel’s eliabili y,conside ing henumbe o p edic o sand
hesamplesize.Theses a is icalmeasu esindica e ha  hep oposedmodeleffec i ely
cap u es hein luenceo wea he pa ame e son henumbe o bicycle ips. Figu e7 is-
uallycompa es hep edic edandac ualcyclingnumbe s o 2019,showing hei s ong
alignmen and u he co obo a ing hehighp ecisiono  hesimula ion.
Table18.Reg essioncoefficien s o mon hlybicycle ip olumein2019(R2=0.942).
ModelSumma y
R*RSqua e**Adjus edRSqua e***S d.E o o  heEs ima e
0.9610.9240.9072,840,489.903
*−1<R<+1;**<R
2
<

1;***<Adjus edR
2
<

1.
Figu e7.Compa isono p edic ed s.ac ualbicycle ipsin2019.
Thenon-linea  eg essionmodelsuccess ullycap u ed hecomplex ela ionshipsbe-
ween hewea he pa ame e sandcyclingbeha io ,pa icula ly h ough heinclusiono 
hesqua ed e m o  empe a u e.Thisapp oachimp o ed hep edic i eaccu acyo  he
model,ase idencedby hehighR
2
 alues,anddemons a es hemodel’s obus nessin
add essingin e dependenciesamong hepa ame e s.
Insumma y, he esul sdemons a e ha  he eisasigni ican  ela ionshipbe ween
wea he pa ame e sand henumbe o cyclis sinHambu g.ThePea sonco ela ionco-
efficien indica esas ongposi i eco ela iono 0.957be weendayligh du a ionand
cyclis numbe s,whilehumidi yexhibi sanega i eco ela iono −0.744,highligh ing
hei  espec i eimpac s.Basedon heANOVA es , heR-squa e alues,and headjus ed
R-squa e alues,dayligh du a ion,wi hanadjus edR-squa e alueo 0.878,has hemos 
subs an ialin luenceoncyclingac i i y.Incon as ,windspeed,wi hanadjus edR-
squa e alueo 0.197,has heleas impac .
Theanalysiso co a ianceisu ilized oexamine hesimul aneouseffec so wea he 
pa ame e soncycling olumes.Theob ainedsigni icance alueo 0.991(p>0.05)sugges s
ha wea he pa ame e sdono collec i elyin luencecyclingnumbe s,unde sco ing hei 
independen impac .
Figu e 7. Compa ison o p edic ed s. ac ual bicycle ips in 2019.
The non-linea eg ession model success ully cap u ed he complex ela ionships
be ween he wea he pa ame e s and cycling beha io , pa icula ly h ough he inclusion
o he squa ed e m o empe a u e. This app oach imp o ed he p edic i e accu acy o

U ban Sci. 2025,9, 41 22 o 30
he model, as e idenced by he high R
2
alues, and demons a es he model’s obus ness
in add essing in e dependencies among he pa ame e s.
In summa y, he esul s demons a e ha he e is a signi ican ela ionship be ween
wea he pa ame e s and he numbe o cyclis s in Hambu g. The Pea son co ela ion
coe icien indica es a s ong posi i e co ela ion o 0.957 be ween dayligh du a ion and
cyclis numbe s, while humidi y exhibi s a nega i e co ela ion o
−
0.744, highligh ing
hei espec i e impac s. Based on he ANOVA es , he R-squa e alues, and he adjus ed
R-squa e alues, dayligh du a ion, wi h an adjus ed R-squa e alue o 0.878, has he mos
subs an ial in luence on cycling ac i i y. In con as , wind speed, wi h an adjus ed R-squa e
alue o 0.197, has he leas impac .
The analysis o co a iance is u ilized o examine he simul aneous e ec s o wea he
pa ame e s on cycling olumes. The ob ained signi icance alue o 0.991 (p> 0.05) sugges s
ha wea he pa ame e s do no collec i ely in luence cycling numbe s, unde sco ing hei
independen impac .
To alida e he accu acy o he model, he co ela ion coe icien be ween he ac ual
and p edic ed numbe o cyclis s o 2017 and 2019 is calcula ed. The esul s yielded
coe icien s o 0.971 and 0.961, espec i ely, highligh ing a s ong alignmen be ween he
obse ed and p edic ed alues. The co esponding R-squa e alues, 0.942 o 2017 and
0.924 o 2019, u he ein o ce he high p ecision and eliabili y o he p oposed model in
p edic ing cycling olumes in Hambu g.
This s udy uniquely examines he indi idual and collec i e e ec s o i e key wea he
componen s on cycling le els in Hambu g. The da a om 2017 was ounda ional in
designing he model, while wea he da a om 2019 alida ed he o mula by compa ing
he p edic ed alues wi h ac ual obse a ions. To u he es he model’s accu acy, wea he
da a om 2021 and 2023 a e inco po a ed o es ima e cyclis numbe s and e alua e he
model’s p ecision.
Table 19 p esen s he p edic ed mon hly bicycle ips o 2021 and 2023, compa ed
wi h ac ual obse ed alues. By p edic ing cyclis numbe s o e e y mon h ac oss hese
wo yea s (24 samples in o al), he model’s e ec i eness and dependabili y a e igo ously
assessed. As shown in Table 20, he co ela ion coe icien s o 2021 and 2023 a e 0.945 and
0.959, espec i ely, con i ming he s ong ela ionships be ween he ac ual and p edic ed
alues. The co esponding R-squa e alues, 0.893 o 2021 and 0.919 o 2023, indica e ha
he model explains a high p opo ion o he a ia ion in cycling ac i i y, ein o cing i s
accu acy and eliabili y.
Table 19. The mon hly a e age o he p edic ed numbe o bicycle ips made in Hambu g in 2021
and 2023.
P edic ed
bicycle ips
in 2021
Jan Feb Ma Ap May Jun
14,562,752 15,688,310 23,867,039 25,587,013 27,548,571 38,396,805
Jul Aug Sep Oc No Dec
37,699,410 33,396,481 37,925,340 28,449,637 22,638,774 13,560,224
P edic ed
bicycle ips
in 2023
Jan Feb Ma Ap May Jun
17,303,039 18,825,732 23,041,591 28,689,909 35,425,890 37,954,161
Jul Aug Sep Oc No Dec
28,203,362 31,957,853 36,372,011 21,119,856 16,638,136 11,158,314
U ban Sci. 2025,9, 41 23 o 30
Table 20. Mon hly co ela ions be ween obse ed and p edic ed numbe o cycling ips (2021
and 2023).
Co ela ions Bicycle T ips in 2021 P edic ed Bicycle T ips
in 2021
Bicycle ips made
in 2021
Pea son Co ela ion 1 0.945 **
Sig. (2- ailed) 0.000
N 12 12
P edic ed bicycle ips
in 2021
Pea son Co ela ion 0.945 ** 1
Sig. (2- ailed) 0.000
N 12 12
Co ela ions Bicycle ips in 2023 P edic ed bicycle ips
in 2023
Bicycle ips made
in 2023
Pea son Co ela ion 1 0.959 **
Sig. (2- ailed) 0.000
N 12 12
P edic ed bicycle ips
in 2023
Pea son Co ela ion 0.959 ** 1
Sig. (2- ailed) 0.000
N 12 12
** The co ela ion is signi ican a he 0.01 le el (2- ailed).
This inno a i e app oach demons a es he e ec i eness o he p oposed model in
p edic ing cycling olumes ac oss mul iple yea s, achie ing high le els o accu acy and
p ecision. By success ully alida ing he o mula wi h da a om 2017 and 2019 and es ing
i s applicabili y wi h da a om 2021 and 2023, his s udy highligh s he model’s obus ness
and es ablishes a pionee ing amewo k o unde s anding he impac o wea he condi ions
on cycling ends.
The co ela ion analysis e ealed s ong ela ionships be ween ce ain wea he pa am-
e e s, pa icula ly empe a u e and sunligh du a ion. While his may in oduce collinea i y
conce ns, ou model add esses hese in e dependencies h ough non-linea me hods ha
p io i ize he o e all p edic i e pe o mance. The inclusion o co ela ed a iables enables
a comp ehensi e unde s anding o hei join impac on cycling beha io , which would be
o e looked i analyzed independen ly.
The esul s unde sco e he signi ican impac o wea he pa ame e s on u ban cycling
in Hambu g, alida ed h ough he achie emen o high co ela ion coe icien s and R-
squa e alues. Fo ins ance, dayligh du a ion eme ged as he mos in luen ial pa ame e ,
while wind speed exhibi ed he s onges de e en e ec . By alida ing he model wi h
da a om 2017 and 2019 and success ully applying i o o ecas cycling ends o 2021
and 2023, his esea ch demons a es he obus ness and eliabili y o he app oach. These
indings highligh he model’s alue o use in u ban mobili y planning and sus ainable
anspo a ion s a egies.
P e ious esea ch on u ban cycling has o en ocused on isola ed e ec s o wea he
pa ame e s, such as empe a u e [
76
] o p ecipi a ion [
87
]. Howe e , he combined o
non-linea in e ac ions be ween hese pa ame e s ha e been la gely o e looked. This s udy
add esses his gap by in eg a ing i e c i ical pa ame e s in o a uni ied p edic i e model.
This in eg a ion p o ides a mo e holis ic unde s anding o how wea he in luences cycling
beha io , b idging gaps in ea lie s udies.
By in eg a ing mul iple wea he pa ame e s, his model no only alida es i s p edic-
i e powe ac oss mul iple yea s, bu also o e s a solid ounda ion o explo ing addi ional
a iables, such as ai quali y indices, eal- ime wea he o ecas s, o spa ially disagg ega ed
cycling da a. Inco po a ing highe equency da a (e.g., daily o hou ly) could u he
enhance he p ecision o empo al analyses. These ad ancemen s would empowe ci y
U ban Sci. 2025,9, 41 24 o 30
planne s and policymake s o c ea e ac ionable s a egies o sus ainable anspo a ion
and in as uc u e design.
Fu u e esea ch could compa e di e en ime pe iods o iden i y long- e m ends,
analyze b oade beha io al pa e ns among cyclis s, and inco po a e addi ional anspo
modes. Tes ing he model in ci ies wi h di e se clima es and in as uc u e o expanding
da ase s would enhance i s gene alizabili y and accu acy. These e o s would deepen
insigh s in o cycling dynamics and sus ainable u ban mobili y s a egies.
5. Conclusions
This esea ch on wea he -d i en cycling o e s a comp ehensi e explo a ion o how
wea he pa ame e s in luence u ban bicycle usage, wi h a ocus on he ci y o Hambu g.
The s udy’s inno a ion lies in i s holis ic app oach, in eg a ing his o ical cycling da a wi h
i e key clima ic ac o s: empe a u e, humidi y, p ecipi a ion, wind speed, and dayligh
du a ion. By inco po a ing a comp ehensi e se o wea he pa ame e s and add essing
hei in e dependencies, his s udy con ibu es o he li e a u e by p o iding new insigh s
in o he collec i e in luence o wea he condi ions on cycling beha io . Fu u e esea ch
could ex end his app oach by explo ing addi ional ac o s, such as ai quali y and eal- ime
wea he o ecas s. Unlike p e ious s udies ha ypically analyze only one o wo ac o s,
his esea ch de elops a obus p edic i e model capable o u u e p ojec ions. This model
equips u ban planne s wi h aluable insigh s o de ising be e s a egies o imp o e
cycling condi ions ac oss di e en ime pe iods.
The key indings om his s udy include:
Wea he condi ions: They play a pi o al ole in de e mining cycling ac i i y. Longe
dayligh hou s and mode a e empe a u es encou age cycling, while high wind speeds,
hea y p ecipi a ion, and ele a ed humidi y ac as signi ican de e en s. This s udy es ab-
lishes clea connec ions be ween hese a iables and cycling ends.
P edic i e model: The o mula ion o a mul i-pa ame e equa ion signi ican ly en-
hances ou unde s anding o how wea he condi ions collec i ely impac cycling ends.
This model su passes p e ious app oaches by p o iding a mo e de ailed and in eg a ed
analysis. I enables accu a e p edic ions o cycling ac i i y in di e se wea he condi ions
and ime ames.
The lexibili y o he p oposed model ex ends i s applicabili y o u u e esea ch. While
his s udy ocuses on wea he pa ame e s, he model can be expanded o include addi ional
indica o s, such as geog aphic ea u es, land use pa e ns, popula ion densi y, and socioeco-
nomic ac o s. This adap abili y allows he app oach o be u ilized o bo h sho - e m and
long- e m u ban planning goals, om ac ion-o ien ed in e en ions o isiona y p ojec s.
The me hodology can be applied ac oss a ious scales, namely buildings, neighbo hoods,
ci ies, o en i e egions, p o iding a p ac ical amewo k o p edic ing u u e ends in
di e en con ex s. Fu u e esea ch could build upon his s udy by in eg a ing addi ional
a iables, such as ai quali y indices and eal- ime wea he o ecas s, o enhance he p edic-
i e powe and applicabili y o he model. Fu he mo e, explo ing al e na i e app oaches
wi h adjus ed weigh s and p io i ies, based on new a iables, could lead o mo e p ecise
and con ex -speci ic insigh s.
To op imize cycling condi ions u he , especially in ad e se wea he scena ios, in-
co po a ing addi ional c i e ia in o he model will yield s onge scien i ic indings and
inno a i e solu ions. This esea ch se s a ounda ion o mo e comp ehensi e analyses
ha combine en i onmen al, social, and economic ac o s o c ea e adap i e and inclusi e
u ban s a egies.
P ac ical implica ions: The indings om his esea ch ha e subs an ial implica ions
o u ban planne s and policymake s. By unde s anding he ela ionship be ween wea he
U ban Sci. 2025,9, 41 25 o 30
condi ions and cycling beha io , wea he -adap i e cycling in as uc u e can be designed
and implemen ed in ci ies. This includes shel e ed bike lanes, imp o ed d ainage sys ems,
and be e ligh ing o sho e dayligh pe iods, ensu ing he sa e y and con enience o
cyclis s all-yea ound.
Fu he mo e, his model can be adap ed o o he anspo a ion modes and u ban
planning challenges, making i a e sa ile ool o add essing di e se needs. By le e aging
his amewo k, he quali y o u ban li e can be enhanced in ci ies, sus ainable mobili y
can be p omo ed, and esilience agains wea he - ela ed challenges can be ensu ed. The
in eg a ion o wea he conside a ions in o u ban design is no only p ac ical, bu essen ial
o os e ing ac i e anspo a ion and c ea ing li able, u u e- eady ci ies.
This s udy demons a es how a da a-d i en, mul i- ace ed app oach can lead o
meaning ul insigh s and ac ionable s a egies, se ing as a empla e o simila esea ch
ac oss di e en egions and con ex s. By expanding he scope and inco po a ing addi ional
ac o s, u u e s udies can build on his ounda ion o u he ad ance he science o u ban
mobili y and planning.
This s udy has ce ain limi a ions ha should be acknowledged. Fi s , he a ailabili y
o ecen li e a u e and comp ehensi e da ase s was limi ed, which may ha e impac ed he
scope o he analysis. Second, he in eg a ion o mul iple me hodologies and non-linea
p edic i e modeling equi ed signi ican ime and compu a ional esou ces, making he
analysis p ocess in ensi e. Thi d, access o high- equency wea he da a, such as hou ly
obse a ions, was cons ained, which could ha e u he imp o ed he model’s empo al
accu acy. Finally, he s udy ocused on a single case s udy (Hambu g), which may es ic
he gene alizabili y o he esul s o o he ci ies wi h di e en clima ic and in as uc u al
condi ions. Add essing hese limi a ions in u u e esea ch could enhance he obus ness
and applicabili y o simila s udies. While his s udy ocuses on Hambu g due o i s unique
and challenging wea he condi ions, he p oposed model o e s signi ican po en ial o
b oade applica ions. Fu u e esea ch could apply he model o ci ies wi h a ying clima ic
and in as uc u al con ex s o e alua e i s adap abili y and gene alizabili y. Addi ionally,
compa a i e analyses be ween Hambu g and o he u ban a eas could p o ide aluable
insigh s in o op imizing u ban cycling s a egies and e ining he model in ega d o
di e se en i onmen s.
Au ho Con ibu ions: Concep ualiza ion, N.F. (Nahid Falah), N.F. (Nadia Falah), and J.S.-G.; me hod-
ology, N.F. (Nahid Falah), N.F. (Nadia Falah); so wa e, N.F. (Nahid Falah).; alida ion, N.F. (Nadia
Falah) and J.S.-G.; o mal analysis, N.F. (Nahid Falah), N.F. (Nadia Falah); in es iga ion, N.F. (Nahid
Falah), N.F. (Nadia Falah); esou ces, N.F. (Nahid Falah), N.F. (Nadia Falah), and J.S.-G.; da a cu a ion,
N.F. (Nadia Falah) and J.S.-G.; w i ing—o iginal d a p epa a ion N.F. (Nahid Falah), N.F. (Nadia
Falah), and J.S.-G.; w i ing— e iew and edi ing, N.F. (Nadia Falah) and J.S.-G.; isualiza ion, N.F.
(Nahid Falah), N.F. (Nadia Falah); supe ision, J.S.-G. All au ho s ha e ead and ag eed o he
published e sion o he manusc ip .
Funding: This esea ch ecei ed no ex e nal unding.
Da a A ailabili y S a emen : The o iginal con ibu ions p esen ed in his s udy a e included in he
a icle: u he inqui ies can be di ec ed o he co esponding au ho .
Con lic s o In e es : The au ho s decla e ha he e a e no con lic s o in e es .