Academic Edi o : Michael Wink
Recei ed: 27 Feb ua y 2025
Re ised: 21 Ma ch 2025
Accep ed: 23 Ma ch 2025
Published: 26 Ma ch 2025
Ci a ion: Sil a, T.I.d.; Ribei o, J.E.d.S.;
San os, T.S.d.; Co eia, M.R.S.; Ribei o,
M.C.B.d.O.; Mendonça, A.J.T.; Sil a,
A.G.C.d.; Oli ei a, P.H.d.A.; Coêlho,
E.d.S.; Ba os Júnio , A.P.; e al. Can a
Non-Des uc i e Me hod P edic he
Lea A ea o Species in he Caa inga
Biome? Di e si y 2025,17, 234.
h ps://doi.o g/10.3390/d17040234
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
Can a Non-Des uc i e Me hod P edic he Lea A ea o Species
in he Caa inga Biome?
Toshik Ia ley da Sil a 1,*, João E e hon da Sil a Ribei o 2,* , Thainan Sip iano dos San os 1,
Ma cos Robe o San os Co eia 1, Ma ia Ca olina Bo ges de Oli ei a Ribei o 1,
Allysson Jonhnny To es Mendonça 1, An onio Gideilson Co eia da Sil a 2,
Pablo Hen ique de Almeida Oli ei a 2, Es e dos San os Coêlho 2, Au élio Paes Ba os Júnio 2,
Elania F ei e da Sil a
2,3
, Al edo Emilio Rubio-Casal
3
, João L. M. P. de Lima
4
, Thie es Geo ge F ei e da Sil a
5
and Alexand e Maniçoba da Rosa Fe az Ja dim 6,*
1Cen e o Ag a ian, En i onmen al, and Biological Sciences, Uni e sidade Fede al do Recônca o da Bahia,
C uz das Almas 44380-000, BA, B azil; [email p o ec ed] (T.S.d.S.);
ma cos_ [email p o ec ed] (M.R.S.C.); [email p o ec ed] (M.C.B.d.O.R.);
[email p o ec ed] (A.J.T.M.)
2Depa men o Ag icul u al and Fo es y Sciences, Fede al Ru al Uni e si y o he Semi-A id,
Mosso ó 59625-900, RN, B azil; [email p o ec ed] (A.G.C.d.S.);
[email p o ec ed] (P.H.d.A.O.); [email p o ec ed] (E.d.S.C.);
au elio.ba [email p o ec ed] (A.P.B.J.); [email p o ec ed] (E.F.d.S.)
3
Depa men o Plan Biology and Ecology, Uni e si y o Se ille, A . Reina Me cedes, s/n, 41012 Se illa, Spain;
[email p o ec ed]
4MARE—Ma ine and En i onmen al Sciences Cen e, ARNET—Aqua ic Resea ch Ne wo k, Depa men o
Ci il Enginee ing, Facul y o Sciences and Technology, Uni e si y o Coimb a, Rua Luís Reis San os, Pólo II,
3030-788 Coimb a, Po ugal; [email p o ec ed]
5
Depa men o Ag icul u al Enginee ing, Fede al Ru al Uni e si y o Pe nambuco, Dom Manoel de Medei os
A enue, s/n, Dois I mãos, Reci e 52171-900, PE, B azil; [email p o ec ed]
6Depa men o Biodi e si y, Ins i u e o Biosciences, São Paulo S a e Uni e si y—UNESP,
Rio Cla o 13506-900, SP, B azil
*Co espondence: [email p o ec ed] (T.I.d.S.); [email p o ec ed] (J.E.d.S.R.);
[email p o ec ed] (A.M.d.R.F.J.)
Abs ac : Unde s anding he lea a ea is essen ial in plan physiology and ecological s udies,
as i di ec ly in luences pho osyn hesis, anspi a ion and plan p oduc i i y. This s udy
aimed o de elop non-des uc i e allome ic models o es ima e he lea a ea o h ee
species om he Caa inga biome: Cynophalla lexuosa,Libidibia e ea and Tabebuia au ea. A
o al o 1293 lea es we e collec ed om hese species, scanned, and analysed using ImageJ
so wa e o ob ain hei leng h, wid h, and ac ual lea a ea. In addi ion, he p oduc o
leng h and wid h was calcula ed. Linea , powe and exponen ial eg ession models we e
used. The bes equa ions we e chosen based on he coe icien o de e mina ion, Pea son’s
linea co ela ion coe icien , Willmo ’s ag eemen index, mean squa ed e o , oo mean
squa ed e o , mean absolu e e o and mean absolu e pe cen age e o . The bes equa ions
o all species we e cons uc ed using linea and powe models, which we e indica ed
o accu a e p edic ion o lea a ea. These indings con i m he e iciency o allome ic
equa ions as a non-des uc i e me hod o p edic ing lea a ea, p o iding an accessible and
economical al e na i e o ecological s udies in semi-a id en i onmen s.
Keywo ds: biome y; eg ession models; allome ic equa ions; Caa inga; semi-a id egion
Di e si y 2025,17, 234 h ps://doi.o g/10.3390/d17040234
Di e si y 2025,17, 234 2 o 15
1. In oduc ion
Lea es pe o m i al unc ions o plan s, ac ing in ligh in e cep ion, ca bon ixa ion,
gas exchange and wa e egula ion. These unc ions make lea a ea an indispensable
pa ame e in s udies o plan physiology and ecology [
1
]. Th oughou e olu ion, lea es
ha e de eloped a ious shapes and sizes as an adap i e s a egy o cope wi h di e en
en i onmen al condi ions [
2
,
3
]. This di e si y di ec ly e lec s on pho osyn he ic e iciency,
wa e use, and plan p oduc i i y, highligh ing he impo ance o in es iga ing lea a ea o
unde s and he pe o mance o species [4,5].
Lea a ea can be de e mined by di ec me hods, such as measu ing wi h g aph pape
o digi al gauges, and by indi ec me hods, which use ma hema ical models based on linea
measu emen s o lea es [
6
]. Al hough di ec me hods a e accu a e, hey ha e disad an-
ages such as he need o des oy lea es, making epea ed analyses in he same g ow h
cycle un easible [
7
]. Indi ec me hods based on allome ic models eme ge as p ac ical and
economical al e na i es, allowing non-des uc i e and successi e measu emen s h ough-
ou he de elopmen o he plan , in addi ion o p esen ing high accu acy in es ima es [
8
].
Allome ic models elimina e expensi e equipmen [9].
Allome ic models ha use a iables such as he leng h (L) and wid h (W) o lea es
o he p oduc o hese dimensions (LW) ha e been widely used in he es ima ion o lea
a ea (LA) o se e al species. Salaza e al. [
7
] demons a ed he e icacy o using he LW
p oduc o es ima e LA in Theob oma cacao. Simila me hods ha e been success ully ap-
plied in species such as Eus oma g andi lo um [
10
] and Manilka a zapo a [
11
], Dend an hema
g andi lo a [
12
], and Eu e pe ole acea [
13
], ein o cing hei e iciency in di e en mo pholog-
ical con ex s.
Species such as Cynophalla lexuosa,Libidibia e ea, and Tabebuia au ea o e unique
oppo uni ies o s udy physiological and ecological adap a ions o ad e se condi ions,
such as high salini y and wa e sca ci y, cha ac e is ic o he B azilian semi-a id egion.
The a ia ion in lea dimensions ac oss hese species is ad an ageous, as i enhances da a
ep esen a i eness and suppo s he de elopmen o obus allome ic models ha accoun
o di e en phenological s ages and en i onmen al condi ions [14].
Cynophalla lexuosa (L.) J. P esl is a sh ub- ee species na i e o B azil, widely dis ibu ed
in di e en ecosys ems, om he A lan ic coas al ege a ion o he semi-a id egions o he
Caa inga [
15
]. Belonging o he Cappa aceae amily, his species has signi ican ecological
and economic ele ance, being used as o age, o wood p oduc ion, o soil eco e y and
in olk medicine [
15
,
16
]. Wi h a heigh o up o 4 m, C. lexuosa is ole an o di e en
ypes o soil and clima e, being ecognized as a pe ennial species, and is also used as a
sou ce o enewable ene gy [
17
]. Libidibia e ea (Ma . ex Tul.) L.P. Quei oz is a ee species
na i e o B azil ound p edominan ly in he No h and No heas egions [
18
]. Valued by
indigenous, adi ional and u ban communi ies, i s wood is widely used; in addi ion, i is
o en cul i a ed as an o namen al ee in B azil and in opical egions and has an ibac e ial
and an i-in lamma o y ac i i ies, in addi ion o being used in he ea men o pa asi ic
diseases such as leishmaniasis [
19
]. T adi ional communi ies also use i in olk medicine
and eligious i uals [20].
Tabebuia au ea (Sil a Manso) Ben h. & Hook. . ex S. Moo e is a ee species widely
dis ibu ed in he opical and sub opical egions o he Ame icas [
21
]. This species na i e
o B azil occu s in all biomes o he coun y, including he Pan anal, A lan ic Fo es , Ce ado,
Amazon, and Caa inga [
22
]. In he B azilian semi-a id egion, T. au ea is no able o i s
o namen al beau y and he shade p o ided by i s b oad canopy. I s d ough ole ance
makes i aluable o en i onmen al es o a ion p ojec s, while i s du able, high-s eng h
wood is widely used in u ban and u al a o es a ion, ci il cons uc ion, pape p oduc ion,
and ca pen y [
23
]. I s lea es, ba k, and oo s a e widely used in olk medicine, including
Di e si y 2025,17, 234 3 o 15
an i-anemic, an ipy e ic, diu e ic, e mi uge, and pu ga i e ac ions, in addi ion o being
used in he ea men o lu and in lamma o y p ocesses [24,25].
S udies ca ied ou wi h E y h oxylum pau e ense [
26
] and Malus domes ica [
27
] enabled
he cons uc ion o equa ions wi h a high coe icien o de e mina ion. Including a la ge
numbe o samples in s udies is essen ial o educe e o s and inc ease he ep esen a i e-
ness o models [14].
In addi ion, he equa ions allow successi e analyses h oughou he plan de el-
opmen cycle, which a e aluable ools in s udies o g ow h, ecology, and sus ainable
managemen [
28
,
29
]. These me hods p o ide ele an in o ma ion o managemen and
conse a ion s a egies o species o ecological and economic impo ance. P e ious s udies
ha e shown ha non-des uc i e me hods a e e ec i e o es ima ing lea a ea in opical
c ops, p omo ing g ea e sus ainabili y and e iciency in ag icul u al managemen [
30
].
Howe e , many species na i e o o adap ed o he semi-a id egion s ill lack scien i ic
app oaches ha explo e hei po en ial.
Gi en hese species’ ele ance o semi-a id ecosys ems and hei adap abili y o se e e
en i onmen al condi ions, s udies ocused on lea a ea es ima ion can help in unde s and-
ing hei ecology and in de eloping sus ainable managemen p ac ices. Non-des uc i e
allome ic me hods allow he e alua ion o he impac o en i onmen al ac o s and ag icul-
u al p ac ices on he g ow h o hese plan s, con ibu ing o hei p ese a ion and a ional
use [5].
These non-des uc i e models seek o imp o e olia analysis echniques, p o iding
aluable ools o ag onomic and ecological s udies. In addi ion, his ype o wo k em-
phasizes he ele ance o e icien and accessible me hods o he sus ainable managemen
o hese species in semi-a id clima e en i onmen s [
27
]. The e o e, his s udy aimed o
de elop allome ic equa ions o es ima e he lea a ea o Cynophalla lexuosa,Libidibia e ea,
and Tabebuia au ea based on he linea dimensions o he lea es.
2. Ma e ials and Me hods
The s udy was ca ied ou in Augus 2024 a he Cen e o Resea ch in Plan Sciences o
he Semi-A id a he Fede al Ru al Uni e si y o he Semi-A id Region, loca ed in he s a e
o Rio G ande do No e, municipali y o Mosso ó, No heas B azil (5
◦
12
′
22
′′
S, 37
◦
19
′
13
′′
W,
al i ude o 21 m). The egion has a ho and d y clima e, cha ac e ized by an a id season
and ain all concen a ed in he summe , classi ied as BSh (hin e land clima e) [
31
]. The
a e age annual ain all is 555 mm, while he a e age annual ai empe a u e is a ound
27.8
◦
C. The p edominan soil in he a ea is classi ied as Eu ophic Red-Yellow Ul isol [
32
].
The samples we e collec ed om mo he ees, and heal hy lea es and lea le s we e
selec ed, ee o damage caused by pes s and diseases o abio ic and bio ic ac o s. Th ee
hund ed six y-eigh lea es o Cynophalla lexuosa, 718 lea le s o Libidibia e ea, and 207
o Tabebuia au a we e used. The lea es we e collec ed om eigh mo he ees o each
species. The numbe o collec ed lea es and lea le s a ied among species due o di e -
ences in lea mo phology and a ailabili y in he sampled ees. The lea samples we e
collec ed unde simila en i onmen al condi ions o minimize ex e nal a iabili y and
ensu e ha he obse ed di e ences in lea dimensions we e p ima ily due o gene ic
and phenological ac o s a he han en i onmen al in luences. This app oach enhances
he eliabili y o he allome ic models by educing con ounding a iables. The samples
we e collec ed andomly o sea ch o lea es o di e en sizes and shapes and o build
models and accu a e equa ions o p edic he lea a ea o hese species. A e collec ion,
he samples we e anspo ed in plas ic con aine s con aining ice o p e en wa e loss
h ough anspi a ion, seeking o mi iga e dehyd a ion. Subsequen ly, he lea es/lea le s
we e de ached and digi ized in a desk op scanne (HP Scanje G2410, Palo Al o, CA, USA)
Di e si y 2025,17, 234 4 o 15
a a maximum esolu ion o 600
×
600 dpi. The images we e p ocessed and con as ed
using ImageJ so wa e e sion 1.53e in he public domain. Then, he leng h (L, cm), co e-
sponding o he dis ance be ween he inse ion o he pe iole and he poin opposi e he
cen al ein, he wid h (W, cm), ob ained as he mos signi ican measu emen pe pendic-
ula o he leng h ein (Figu e 1), and he ac ual lea /lea le a ea (LA, cm
2
) we e de e -
mined indi idually. These da a we e used o calcula e he p oduc s be ween leng h and
wid h (LW, cm2).
Di e si y 2025, 17, x FOR PEER REVIEW 4 o 16
maximum esolu ion o 600 × 600 dpi. The images we e p ocessed and con as ed using
ImageJ so wa e e sion 1.53e in he public domain. Then, he leng h (L, cm),
co esponding o he dis ance be ween he inse ion o he pe iole and he poin opposi e
he cen al ein, he wid h (W, cm), ob ained as he mos signi ican measu emen
pe pendicula o he leng h ein (Figu e 1), and he ac ual lea /lea le a ea (LA, cm
2
) we e
de e mined indi idually. These da a we e used o calcula e he p oduc s be ween leng h
and wid h (LW, cm
2
).
Figu e 1. Rep esen a i e lea es o Cynophalla lexuosa, Libidibia e ea and Tabebuia au ea. Black lea is
an example o a bina y image.
Acco ding o Equa ion (1) [33] and Equa ion (2) [34], he a iance in la ion ac o
(VIF) and he ole ance alue (T) we e calcula ed o e alua e he p esence o
mul icollinea i y in he da a. The VIF quan i ies how much he a iance o a eg ession
coefficien is in la ed due o mul icollinea i y. A he same ime, he ole ance alue (T) is
ecip ocal, indica ing how much o a a iable’s a iance is no explained by o he
p edic o s in he model. High VIF alues (>10) o low T alues (<0.1) sugges p oblema ic
mul icollinea i y. When VIF is less han 10 and T is g ea e han 0.1, leng h and wid h da a
can be used o es ima e he a ea o lea le s using eg ession models [34].
VIF= 1
1−𝑟²
(1)
T =
(2)
whe e ep esen s he co ela ion coefficien be ween L and W.
Linea and nonlinea eg ession models we e es ed o es ima e LA, conside ed a
dependen a iable, as a unc ion o lea dimensions (L, W, and LW) as independen
a iables. To cons uc he allome ic equa ions, linea (ŷ = β
0
+
β
1
× x), powe (ŷ = β
0
× x
β1
)
and exponen ial (ŷ = β
0
× β
1x
) models we e used, in which he alues o ŷ ep esen he
es ima ed lea a ea, he alues o he linea dimensions o he lea es x, and he eg ession
coefficien s β
0
and β
1
.
The c i e ia adop ed o selec ing he bes equa ion o p edic ing he lea a ea o he
h ee species we e based on he highes coefficien o de e mina ion (R
2
), Pea son’s linea
co ela ion coefficien ( ), and Willmo ag eemen index (d), and lowes mean squa ed
e o (MSE), oo mean squa ed e o (RMSE), mean absolu e e o (MAE) and mean
absolu e pe cen age e o (MAPE). When VIF is less han 10 and T is g ea e han 0.1,
leng h and wid h da a can be used o es ima e he a ea o lea le s using eg ession models
[34].
Figu e 1. Rep esen a i e lea es o Cynophalla lexuosa,Libidibia e ea and Tabebuia au ea. Black lea is
an example o a bina y image.
Acco ding o Equa ion (1) [
33
] and Equa ion (2) [
34
], he a iance in la ion ac o (VIF)
and he ole ance alue (T) we e calcula ed o e alua e he p esence o mul icollinea i y in
he da a. The VIF quan i ies how much he a iance o a eg ession coe icien is in la ed
due o mul icollinea i y. A he same ime, he ole ance alue (T) is ecip ocal, indica ing
how much o a a iable’s a iance is no explained by o he p edic o s in he model. High
VIF alues (>10) o low T alues (<0.1) sugges p oblema ic mul icollinea i y. When VIF is
less han 10 and T is g ea e han 0.1, leng h and wid h da a can be used o es ima e he
a ea o lea le s using eg ession models [34].
VIF =1
1− 2(1)
T=1
VIF (2)
whe e ep esen s he co ela ion coe icien be ween L and W.
Linea and nonlinea eg ession models we e es ed o es ima e LA, conside ed a
dependen a iable, as a unc ion o lea dimensions (L, W, and LW) as independen a i-
ables. To cons uc he allome ic equa ions, linea (
ˆ
y =
β0
+
β1×
x), powe (
ˆ
y = β0×xβ1
)
and exponen ial (
ˆ
y =
β0×β1x
) models we e used, in which he alues o
ˆ
y ep esen he
es ima ed lea a ea, he alues o he linea dimensions o he lea es x, and he eg ession
coe icien s β0and β1.
The c i e ia adop ed o selec ing he bes equa ion o p edic ing he lea a ea o he
h ee species we e based on he highes coe icien o de e mina ion (R
2
), Pea son’s linea
co ela ion coe icien ( ), and Willmo ag eemen index (d), and lowes mean squa ed e o
(MSE), oo mean squa ed e o (RMSE), mean absolu e e o (MAE) and mean absolu e
Di e si y 2025,17, 234 5 o 15
pe cen age e o (MAPE). When VIF is less han 10 and T is g ea e han 0.1, leng h and
wid h da a can be used o es ima e he a ea o lea le s using eg ession models [34].
R2=1−∑n
i=1(yi−ˆyi)2
∑n
i=1(y′i)2(3)
=∑n
i=1(yi−y)(xi−x)
q∑n
i=1(yi−y)2∑n
i=1(xi−x)2(4)
d=1−∑n
i=1(ˆyi−yi)2
∑n
i=1(|ˆy′i|+|y′i|)2(5)
MSE =∑n
i=1(ˆyi −yi)2
n(6)
RMSE =s∑n
i=1(ˆyi −yi)2
n(7)
MAE =∑n
i=1|yi−ˆ
yi|
n(8)
MAPE =100
n∑n
i=1
yi−ˆyi
yi
(9)
whe e
ˆ
yi
is he es ima ed alues o lea a ea;
yi
is he obse ed alues o lea a ea;
yi
is he
a e age o obse ed alues;
ˆ
y′i=ˆ
yi −y
;
y′i=yi −y
;
n
is he o al obse a ion numbe s;
xi
and yi a e he i- h obse a ions o he independen and dependen a iables, espec i ely;
yand xa e a e ages o he a iables yand x.
Ini ially, a desc ip i e da a analysis was pe o med, including he minimum, maxi-
mum and mean alues, o al ampli ude, asymme y coe icien , ku osis and coe icien o
a ia ion o he sampled pa ame e s. The c i e ia adop ed o selec ing he bes equa ion
o p edic ing he lea a ea o he h ee species we e based on s a is ical pe o mance me -
ics. The equa ion wi h he highes coe icien o de e mina ion (R
2
) and Pea son’s linea
co ela ion coe icien ( ) we e p e e ed, indica ing a s ong ela ionship be ween p edic ed
and obse ed alues. The Willmo ag eemen index (d) was also conside ed o assess he
model’s o e all p edic i e accu acy. To ensu e low p edic ion e o , he equa ion wi h he
lowes mean squa ed e o (MSE), oo mean squa ed e o (RMSE), mean absolu e e o
(MAE), and mean absolu e pe cen age e o (MAPE) was selec ed. The no mali y o he
da a was assessed using he Shapi o–Wilk es . The obse ed lea a ea (ac ual) and he
es ima ed a ea we e compa ed using he S uden ’s - es o pai ed samples (p
≤
0.01). Da a
analysis was pe o med using R so wa e ( e sion 4.1.2) [35].
3. Resul s
The esul s o he desc ip i e analysis o he h ee species e alua ed, Cynophalla lexu-
osa,Libidibia e ea and Tabebuia au ea, highligh ma ked di e ences in lea cha ac e is ics.
Fo C. lexuosa, he leng h o he lea es a ied be ween 3.482 and 9.690 cm, wi h an a e age
o 7.303 cm and an ampli ude o 6.208 cm, while he wid h a ied om 2.199 o 6.168 cm,
wi h an a e age o 4.457 cm and an ampli ude o 3.969 cm. The p oduc o leng h and wid h
p esen ed alues be ween 8.545 and 59.768 cm
2
, wi h a mean o 33.165 cm
2
and ampli ude
o 51.223 cm
2
. The lea a ea o his species anged om 6.634 o 47.695 cm
2
, wi h a mean o
25.754 cm2and a ange o 41.061 cm2(Table 1).
Di e si y 2025,17, 234 6 o 15
Table 1. Desc ip i e analysis o leng h (W), wid h (L), p oduc o leng h and wid h (LW), and lea
a ea (LA) o Cynophalla lexuosa,Libidibia e ea, and Tabebuia au ea.
Desc ip i e S a is ic L W LW LA
Cynophalla lexuosa
Minimum 3.482 2.199 8.545 6.634
Maximum 9.690 6.168 59.768 47.695
Ampli ude 6.208 3.969 51.223 41.061
Mean 7.303 4.457 33.165 25.754
S anda d de ia ion 1.098 0.678 9.227 7.315
Coe icien o a ia ion 15.0 15.2 27.8 28.4
Asymme y a−0.381 −0.511 −0.012 0.006
Ku osis + 3 b2.939 2.965 2.596 2.629
Shapi o–Wilk 0.003 ** <0.0001 ** 0.321 ns 0.384 ns
Libidibia e ea
Minimum 1.534 0.756 1.289 1.030
Maximum 3.784 2.328 8.574 6.803
Ampli ude 2.250 1.572 7.285 5.773
Mean 2.438 1.336 3.383 2.651
S anda d de ia ion 0.479 0.298 1.421 1.102
Coe icien o a ia ion 19.6 22.3 42.0 41.6
Asymme y a0.558 0.837 1.086 1.049
Ku osis + 3 b2.688 3.104 3.652 3.526
Shapi o–Wilk <0.0001 ** <0.0001 ** <0.0001 ** <0.0001 **
Tabebuia au ea
Minimum 1.898 1.002 2.014 1.546
Maximum 28.967 6.518 188.807 126.440
Ampli ude 27.069 5.516 186.793 124.894
Mean 12.468 3.180 45.602 33.396
S anda d de ia ion 6.023 1.091 34.161 23.583
Coe icien o a ia ion 48.3 34.3 74.9 70.6
Asymme y a0.360 0.060 1.246 1.027
Ku osis + 3 b2.596 2.841 5.160 4.392
Shapi o–Wilk 0.003 ** 0.030 * <0.0001 ** <0.0001 **
a
Asymme y di e s om ze o by he - es a 5% p obabili y;
b
Ku osis di e s om h ee by he - es a 5%
p obabili y; ** Signi ican a 1% p obabili y; * Signi ican a 5% p obabili y; ns No signi ican .
The species T. au ea p esen ed he highes alues and he mos signi ican a ia ion.
The leng h o he lea es a ied widely be ween 1.898 and 28.967 cm, wi h an a e age o
12.468 cm and a wid h o 27.069 cm. The wid h a ied be ween 1.002 and 6.518 cm, wi h a
mean o 3.180 cm and a wid h o 5.516 cm. The p oduc o leng h and wid h anged om
2.014 o 188.807 cm
2
, wi h a mean o 45.602 cm
2
and ampli ude o 186.793 cm
2
. The lea a ea
anged om 1.546 o 126.440 cm
2
, wi h a mean o 33.396 cm
2
and a ange o 124.894 cm
2
.
Thus, i is no ewo hy ha T. au ea has he la ges lea es wi h mo e signi ican a iabili y
in hei dimensions, L. e ea has he mos mino and mos uni o m lea es, while C. lexuosa,
in u n, has in e media e alues wi h less dispe sion o he da a (Figu e 2).
The linea and nonlinea associa ion pa e ns be ween he independen and dependen
a iables we e used o cons uc eg ession models ha es ima e he lea a ea (LA) o C.
lexuosa,L. e ea and T. au ea (Figu e 3). These esul s highligh he impo ance o using
di e en app oaches, linea and nonlinea , depending on he a iables unde analysis and
he species s udied, ensu ing mo e obus eg ession models o lea a ea es ima ion.
Di e si y 2025,17, 234 7 o 15
Di e si y 2025, 17, x FOR PEER REVIEW 7 o 16
Figu e 2. Rela ionship be ween he numbe o lea es and pe cen age o lea a ea di ided in o
diffe en size classes. (A) Cynophala lexuosa; (B) Libidibia e ea; (C) Tabebuia au ea.
Figu e 2. Rela ionship be ween he numbe o lea es and pe cen age o lea a ea di ided in o di e en
size classes. (A)Cynophala lexuosa; (B)Libidibia e ea; (C)Tabebuia au ea.
Di e si y 2025, 17, x FOR PEER REVIEW 7 o 16
Figu e 2. Rela ionship be ween he numbe o lea es and pe cen age o lea a ea di ided in o
diffe en size classes. (A) Cynophala lexuosa; (B) Libidibia e ea; (C) Tabebuia au ea.
Figu e 3. Ma ix plo and equency his og ams be ween lea pa ame e s. (A)Cynophala lexuosa;
(B)Libidibia e ea; (C)Tabebuia au ea.
Di e si y 2025,17, 234 8 o 15
The esul s o he eg ession models o es ima e he lea a ea o he species C. lexuosa,
L. e ea and T. au ea highligh he e iciency o di e en equa ions as a unc ion o lea
dimensions (Table 2). The analyses we e pe o med based on c i e ia such as coe icien
o de e mina ion (R
2
), Pea son’s co ela ion coe icien ( ), Willmo ag eemen index (d),
mean squa e e o (RMSE), mean absolu e e o (MAE) and mean absolu e pe cen age
e o (MAPE).
Table 2. Reg ession model, equa ions, coe icien o de e mina ion (R
2
), Pea son co ela ion coe icien
( ), Willmo ’s conco dance index (d), mean squa ed e o (MSE), oo mean squa e e o (RMSE),
mean absolu e e o (MAE) and mean absolu e pe cen age e o (MAPE) ob ained as a unc ion o
measu emen s o lea dimensions o Cynophalla lexuosa,Libidibia e ea, and Tabebuia au ea.
Equa ion
Code Model R2 dMSE RMSE MAE MAPE Es ima o o LA (ˆy)
Cynophalla lexuosa
1 Linea 0.8997 0.9485 0.9729 5.367 2.316 1.862 0.0914 ˆ
y=−20.393 +6.319 ×L
2 Linea 0.9489 0.9005 0.9732 5.322 2.306 1.827 0.0844 ˆy =−19.860 +10.230 ×W
3 Linea 0.9950 0.9975 0.9987 0.266 0.516 0.406 0.0164 ˆy =−0.472 +0.790 ×LW
4 Powe 0.9108 0.9544 0.9763 4.769 2.184 1.751 0.0758 ˆy =0.591 ×L1.888
5 Powe 0.9153 0.9567 0.9776 4.532 2.129 1.676 0.0712 ˆy =1.477 ×W1.898
6 Powe 0.9950 0.9975 0.9987 0.265 0.515 0.405 0.0163 ˆy =0.723 ×LW1.019
7 Exponen ial 0.9077 0.9527 0.9750 4.946 2.224 1.762 0.0764 ˆy =3.903 ×1.288L
8 Exponen ial 0.9147 0.9564 0.9770 4.567 2.137 1.662 0.0688 ˆy =3.808 ×1.522W
9 Exponen ial 0.9147 0.9564 0.9770 4.567 2.137 1.662 0.0688 ˆy =9.480 ×1.095LW
Libidibia e ea
1 Linea 0.9146 0.9563 0.9772 0.103 0.322 0.250 0.1060 ˆ
y=−2.715 +2.201 ×L
2 Linea 0.9315 0.9651 0.9819 0.083 0.288 0.220 0.0892 ˆy =−2.123 +3.573 ×W
3 Linea 0.9951 0.9975 0.9987 0.005 0.077 0.058 0.0221 ˆy =0.033 +0.773 ×LW
4 Powe 0.9314 0.9650 0.9820 0.083 0.288 0.216 0.0841 ˆy =0.424 ×L2.011
5 Powe 0.9331 0.9660 0.9822 0.081 0.285 0.216 0.0817 ˆy =1.560 ×W1.727
6 Powe 0.9951 0.9975 0.9987 0.005 0.076 0.058 0.0220 ˆy =0.797 ×LW0.986
7 Exponen ial 0.9254 0.9620 0.9799 0.090 0.301 0.219 0.0837 ˆy =0.412 ×2.090L
8 Exponen ial 0.9113 0.9546 0.9753 0.108 0.329 0.256 0.0973 ˆy =0.567 ×3.039W
9 Exponen ial 0.9113 0.9546 0.9753 0.108 0.329 0.256 0.0973 ˆy =1.176 ×1.253LW
Tabebuia au ea
1 Linea 0.9223 0.9603 0.9794 43.190 6.572 4.958 0.6849 ˆ
y=−13.491 +3.761 ×L
2 Linea 0.8926 0.9447 0.9709 59.706 7.727 5.752 0.6037 ˆy =−31.560 +20.430 ×W
3 Linea 0.9902 0.9950 0.9975 5.459 2.336 1.856 0.0877 ˆy =2.070 +0.687 ×LW
4 Powe 0.9424 0.9708 0.9850 32.011 5.657 4.375 0.1460 ˆy =0.675 ×L1.511
5 Powe 0.9394 0.9692 0.9838 33.837 5.817 4.241 0.1497 ˆy =2.562 ×W2.114
6 Powe 0.9923 0.9961 0.9980 4.295 2.072 1.591 0.0594 ˆy =1.002 ×LW0.923
7 Exponen ial 0.9167 0.9574 0.9758 48.088 6.934 5.655 0.2393 ˆy =9.280 ×1.096L
8 Exponen ial 0.8973 0.9472 0.9677 61.331 7.831 6.408 0.2431 ˆy =6.159 ×1.641W
9 Exponen ial 0.8973 0.9472 0.9677 61.331 7.831 6.408 0.2431 ˆy =18.671 ×1.012LW
Fo C. lexuosa, he linea (
ˆ
y =
−
0.472 + 0.790
×
LW) and powe (
ˆ
y = 0.723
×
LW
1.019
)
models we e he mos accu a e, p esen ing R
2
o 0.9950, o 0.9975, and do 0.9987, indica -
ing ha app oxima ely 99.5% o he a ia ion in lea a ea could be explained by he adjus ed
equa ions. The RMSE and MAE o hese models we e low (0.516 and 0.406 o he linea
model and 0.515 and 0.405 o he powe model), ein o cing he accu acy o he es ima es.
In he species L. e ea, linea (
ˆ
y = 0.033 + 0.773 ×LW
) and powe (
ˆ
y = 0.797 ×LW0.986
)
models also s ood ou , wi h R
2
o 0.9951, o 0.9975 and do 0.9987. The RMSE and MAE
alues we e he lowes eco ded (0.077 and 0.058 o he linea model and 0.076 and 0.058
o he powe model), sugges ing high eliabili y in he p edic ions o hese equa ions.
Di e si y 2025,17, 234 9 o 15
Fo T. au ea, he linea (
ˆ
y = 2.070 + 0.687
×
LW) and powe (
ˆ
y = 1.002
×
LW
0.923
)
models again showed be e pe o mance, wi h R
2
o 0.9902 and 0.9923, o 0.995 and
0.9961, and d o 0.9975 and 0.998, espec i ely. The associa ed e o s (RMSE and MAE)
we e signi ican ly low (2.336 and 1.856 in he linea model and 2.072 and 1.591 in he powe
model), indica ing high accu acy in he es ima ion o lea a ea. Models based on linea
equa ions wi hou in e cep (model 3) and powe (model 6) we e he mos app op ia e o
es ima e hese species’ lea a ea, ega dless o hei mo phological di e ences.
Figu e 4shows he ela ionship be ween lea a ea (LA) and he p oduc o leng h
and wid h (LW) o he species C. lexuosa,L. e ea, and T. au ea. Visual analysis o he
esiduals’ dispe sion and he applica ion o he bes models con i med he accu acy o he
adjus ed equa ions.
Di e si y 2025, 17, x FOR PEER REVIEW 9 o 16
he linea model and 0.515 and 0.405 o he powe model), ein o cing he accu acy o he
es ima es. In he species L. e ea, linea (ŷ = 0.033 + 0.773 × LW) and powe (ŷ = 0.797 ×
LW
0.986
) models also s ood ou , wi h R
2
o 0.9951, o 0.9975 and d o 0.9987. The RMSE
and MAE alues we e he lowes eco ded (0.077 and 0.058 o he linea model and 0.076
and 0.058 o he powe model), sugges ing high eliabili y in he p edic ions o hese
equa ions.
Fo T. au ea, he linea (ŷ = 2.070 + 0.687 × LW) and powe (ŷ = 1.002 × LW
0.923
) models
again showed be e pe o mance, wi h R
2
o 0.9902 and 0.9923, o 0.995 and 0.9961, and
d o 0.9975 and 0.998, espec i ely. The associa ed e o s (RMSE and MAE) we e
signi ican ly low (2.336 and 1.856 in he linea model and 2.072 and 1.591 in he powe
model), indica ing high accu acy in he es ima ion o lea a ea. Models based on linea
equa ions wi hou in e cep (model 3) and powe (model 6) we e he mos app op ia e o
es ima e hese species’ lea a ea, ega dless o hei mo phological diffe ences.
Figu e 4 shows he ela ionship be ween lea a ea (LA) and he p oduc o leng h and
wid h (LW) o he species C. lexuosa, L. e ea, and T. au ea. Visual analysis o he
esiduals’ dispe sion and he applica ion o he bes models con i med he accu acy o he
adjus ed equa ions.
Figu e 4. Rela ionship be ween LA and LW and he bes models. (A) Cynophala lexuosa; (B) Libidibia
e ea; (C) Tabebuia au ea.
Fo C. lexuosa, a well-de ined posi i e ela ionship be ween LA and LW was
obse ed, wi h low da a dispe sion and homogenei y o he esiduals. The bes models
iden i ied we e he linea non-in e cep and powe models wi h R
2
alues o 0.995,
indica ing ha LW could explain 99.50% o he a ia ion in LA. These models
demons a ed high efficiency in p edic ing lea a ea o his species. In he case o L. e ea,
he same models (linea wi hou in e cep and powe ) also showed a good i , wi h a clea
posi i e ela ionship be ween LA and LW. The esiduals we e homogeneous, and he
dispe sion was minimal, ein o cing he applicabili y o he p oposed equa ions. The R
2
alue was also high (0.9951), showing high explana o y capaci y.
Fo T. au ea, he esul s showed simila beha io , wi h he linea non-in e cep and
powe models p o iding he bes i s. Despi e he g ea e ampli ude in he LW and LA
alues, he posi i e ela ionship be ween hese a iables was consis en . The R
2
alues
we e sligh ly lowe (0.9923 and 0.9902), bu i s ill indica es ha he a ia ion in LA can
be explained by he LW p oduc . The linea non-in e cep and powe models
demons a ed high applicabili y and p ecision in es ima ing LW lea a ea (LA) om he
h ee species s udied. The da a’s low dispe sion and he esiduals’ homogenei y alida e
he use o hese models in p ac ical applica ions and compa a i e s udies.
The models chosen (linea and powe ) o es ima e he lea a ea (LA) o C. lexuosa, L.
e ea and T. au ea showed a high co ela ion wi h he obse ed LA alues (R
2
close o 0.99,
Figu e 4. Rela ionship be ween LA and LW and he bes models. (A)Cynophala lexuosa; (B)Libidibia
e ea; (C)Tabebuia au ea.
Fo C. lexuosa, a well-de ined posi i e ela ionship be ween LA and LW was obse ed,
wi h low da a dispe sion and homogenei y o he esiduals. The bes models iden i ied
we e he linea non-in e cep and powe models wi h R
2
alues o 0.995, indica ing ha LW
could explain 99.50% o he a ia ion in LA. These models demons a ed high e iciency in
p edic ing lea a ea o his species. In he case o L. e ea, he same models (linea wi hou
in e cep and powe ) also showed a good i , wi h a clea posi i e ela ionship be ween LA
and LW. The esiduals we e homogeneous, and he dispe sion was minimal, ein o cing
he applicabili y o he p oposed equa ions. The R
2
alue was also high (0.9951), showing
high explana o y capaci y.
Fo T. au ea, he esul s showed simila beha io , wi h he linea non-in e cep and
powe models p o iding he bes i s. Despi e he g ea e ampli ude in he LW and LA
alues, he posi i e ela ionship be ween hese a iables was consis en . The R
2
alues
we e sligh ly lowe (0.9923 and 0.9902), bu i s ill indica es ha he a ia ion in LA can be
explained by he LW p oduc . The linea non-in e cep and powe models demons a ed
high applicabili y and p ecision in es ima ing LW lea a ea (LA) om he h ee species
s udied. The da a’s low dispe sion and he esiduals’ homogenei y alida e he use o hese
models in p ac ical applica ions and compa a i e s udies.
The models chosen (linea and powe ) o es ima e he lea a ea (LA) o C. lexuosa,
L. e ea and T. au ea showed a high co ela ion wi h he obse ed LA alues (R
2
close o
0.99, Figu e 5). Fo C. lexuosa, he linea model (Figu e 5A) showed a s ong ela ionship
be ween he obse ed lea a ea (OLA) and he es ima ed lea a ea (ELA), wi h a high
coe icien o de e mina ion (R
2
= 0.9950), indica ing ha he model was highly e ec i e
o ep esen ing he ela ionship be ween he a iables. The powe model (Figu e 5D) also
