Leaf Area Estimation of Cenostigma pyramidale: A Native Brazilian Caatinga Species

Abstract

Global climatic changes are shifting, leading to increased temperatures and unpredictable rainfall changes. Tropical biomes, including the Caatinga in Northeast Brazil, face the risk of aridification. The Caatinga, covering 10.7% of Brazil, is an area that is highly susceptible to desertification, with a projected 5°C temperature rise. The region’s unique biodiversity faces threats from human activities, including deforestation and habitat fragmentation. Cenostigma pyramidale is a Brazilian native species, very common in the Caatinga ecosystem. The plant is vital for its adaptability to dry season and seasonality and presents a potential for use in reforestation programs. However, it also faces challenges like deforestation and habitat fragmentation. This study aims to estimate the leaf area of C. pyramidale using nondestructive and cost-effective methods. The proposed model, based on leaf length and width, provides reliable and accurate leaf area (LA) estimates. Then, this study recommends the equation LA = 0.746 × (LW)^0.979 as the best for estimating leaf area due to its high accuracy and biological consistency. The equation LA = 0.725 × (LW) is recommended as a simpler equation because the linear equation is easier to do mainly in land cases or in more extensive experiments. The equation LA = 0.645 × (L)^1.652 eliminates 50% of the allometric process because only one leaf dimension is used to estimate leaf area. The leaf area estimation model provides a practical tool for researchers studying plant physiology and agronomic decision-making. The findings emphasize the importance of understanding the impact of climate change on species distribution, especially in vulnerable regions like the Caatinga.

This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

1. Introduction

In recent decades, observable shifts in Earth’s climate have become increasingly apparent. Primarily characterized by rising temperatures and fluctuating precipitation patterns, these changes have grown more pronounced and erratic over time. Projections indicate the likelihood of severe climate alterations, particularly within tropical ecosystems. For instance, in South America, the forecast suggests a potential transformation where savannas could encroach upon forested areas, while semiarid zones might extend across Brazil’s northeast region [1].

The Caatinga (dry forests in Northeastern Brazil) occupies about 912,529 km² or 10.7% of Brazilian area [2]. Approximately 94% of the Caatinga is identified as moderately to highly susceptible to desertification [3]. Projections suggest a substantial temperature rise of 5°C under the RCP 8.5 scenario [4]. The amalgamation of dwindling rainfall, escalating temperatures, soil degradation, and the impending risk of desertification collectively position this region among the globe’s most vulnerable to the adverse effects of climate change [5]. The Caatinga has a high rate of endemism and several biological interactions, making this region so valuable. However, the Caatinga is still little known scientifically, and the areas and biological groups studied are limited. In accordance with Araujo et al. [6], Caatinga contains endemic species, which are not found in any other region of the world, presenting itself as a promising source for the discovery of new natural products. The degradation and loss of the native forest cover of the Caatinga, owing at least, in part, to a culmination of human settlement over time, has led to a present landscape characterized by habitat fragmentation, a decline in biodiversity, and a myriad of species facing the risk of extinction and desertification [1, 7]. Because of these conditions, the Caatinga serves as an exemplary tropical ecosystem for investigating the impact of climate change on biodiversity [8].

Cenostigma pyramidale (Tul.) (E. Gagnon and G.P. Lewi) is found in different areas, from humid to more arid locations [9, 10], and easily adapts to multiple different types of soils, from the poorest to the most stony [11], always as pioneer plants in the ecological succession [12]. In the Caatinga ecosystem, this species is considered one of the most common plant species [9]. Its compound leaves present a distinctive form, composed of numerous small leaflets, giving the tree a delicate, feathery appearance. During the dry season, these leaves often shed, highlighting the tree’s adaptability to the fluctuating environmental conditions typical of its native habitat. However, as recently reported [12, 13], the leaf abscission occurs only in severe water privation. In general, 40% of the leaves remain in the mother-plant to sustain photosynthesis [12, 13].

Recent studies show that C. pyramidale is easy to adapt to new environments, viable to use in reforestation program [14, 15] and soil amelioration [16]. Efforts by local communities, conservation organizations, and governmental initiatives aim to protect and sustainably manage the habitats where C. pyramidale thrives. By recognizing its ecological importance and promoting conservation strategies, there is hope for preserving this species so that future generations can admire and benefit from its presence in the intricate tapestry of the Amazon and tropical dry forest. C. pyramidale faces threats due to deforestation, logging, and habitat fragmentation. Conservation efforts are crucial to safeguarding this species and preserving the intricate balance it maintains within its ecosystem.

Leaf area measurements are required for many plant physiology studies such as growth, photosynthesis, transpiration, photon interception, energy balance, fertilizer application, growth rate, and others [17–21]. Due to these features, the leaf area measurement is especially important [17, 18, 22, 23]. However, the direct measurement of leaf area requires expensive equipment [19] such as hand-held equipment or destructive leaf area measurement devices. Low-cost smartphone applications have revolutionized the way leaf area measurements are conducted [24, 25]. Nonetheless, digital leaf area meters are restricted to environments with sparse and robust foliage, limiting their application to controlled laboratory settings and rendering them impractical for field studies [26]. The challenges inherent in directly quantifying leaf area have spurred the exploration of alternative methodologies, driven by laborious procedures, high expenses, and time-intensive protocols, exacerbated by logistical limitations [17–19, 24, 25, 27]. Measuring leaf area through equations based on leaf dimensions provides a cost-effective, swift, and non-invasive method to precisely estimate leaf area, offering a practical alternative to more labor-intensive approaches [19]. Constructing linear models to elucidate the correlations between leaf area (LA) and other leaf dimensions offers a swift, dependable, cost-effective, precise, and non-invasive method, yielding estimations for LA without reliance on destructive techniques [17, 18, 24, 25]. Also, with nondestructive leaf area methods that allow for the replication of measurements during the growth period, it also reduces some of the experimental variability associated to destructive sampling procedures [17, 18, 20, 21, 23, 28]. In crop species, precise assessment of leaf area (LA) assumes heightened significance due to the leaf’s pivotal role in interacting with the environment. Agronomic studies leverage this crucial parameter as a cornerstone for developing essential decision-making tools and methodologies. Also, allometric principles serve as instrumental tools in decoding the intricate interplay between characteristic dimensions like leaf size and resultant leaf area. The aim of this research is to develop a reliable, precise, and nondestructive method for estimating leaf area in Cenostigma pyramidale, an emblematic native plant of the Caatinga biome, through an accurate leaf area equation based on measurable leaf dimensions. Our hypotheses are structured as follows: (i) The formulation of a singular leaf area equation with exceptional reliability and impartiality is feasible; (ii) the utilization of one leaf dimension can yield a leaf area equation that is both dependable and accurate.

2. Materials and Methods

2.1. Model Construction

For model construction, 1260 leaves were collected from 45 plants grown at Fazenda Buenos Aires near the city of Serra Talhada, Pernambuco, carefully transported at the laboratory of the Federal University of Pernambuco, Brazil, where were processed in accordance with Rodriguez-Paez et al. [29]. The leaves encompassed the broadest range possible as shown in Table 1.

Table 1

Means ± standard deviations (SD), minimum (Min), and maximum (Max) values for the leaf length and width and leaf area of 1260 independent Cenostigma pyramidale leaves.

Features	ẍ	Min	Max
Leaf length	1.97 ± 0.55	0.80	3.96
Leaf width	1.37 ± 0.34	0.51	2.51
L/W	0.71 ± 0.15	0.41	1.37
Leaf area	2.07 ± 1.00	0.32	6.55

2.2. Model Validation

To validate the proposed model, a separate dataset comprising 350 leaves was randomly collected from various strata of the tree canopy, extracted from the branches, and processed in accordance with the methodology outlined in Section 2.1.

2.3. Equation Generation

Nine distinct theoretical models, extensively cited in literature [17, 18, 23, 30–32], underwent rigorous examination. These equations were formulated based on diverse permutations of the constituent elements within leaf area (LA), serving as the dependent variable, in conjunction with corresponding values of both length (L) and width (W), designated as the independent variables. The formulations adhered closely to the principle of parsimony elucidated by Steel and Penny [33]. The main objective of this study was to develop models characterized by simplicity, or alternatively, to encapsulate the most optimal representation of the underlying data. Each equation underwent meticulous calibration across various paradigms, encompassing linear simplicity, modified linearity (attained through the exclusion of ${\beta }_0$), and power models, as delineated in Table 1.

Parameter estimation for each model was conducted using DataFit version 8.0.32, developed by Oakdale Engineering in Oakdale, PA, USA. Model selection was contingent upon various statistical benchmarks, including the following: (i) analysis of variance, employing the F test with a significance threshold of $p<0.001$; (ii) adjusted coefficient of determination ($R^{2}a$); (iii) mean squared error (MSres); (iv) Student’s t-test, where significance was set at $p<0.001$, applied to the absolute mean of errors alongside confidence intervals [34]; and (v) the analysis entailed examining the dispersion pattern of residuals as a percentage (%) of the observed leaf area against the estimated leaf area. This facilitated the identification of the most robust relationship, characterized by the highest $R^{2}a$ value, between observed and estimated leaf area, validated against an independent dataset. Additionally, the examination aimed to assess the equation’s reliability in estimating leaf area for a distinct sample in an unbiased manner, ensuring stability with a high coefficient of determination. This comprehensive scrutiny of residuals spanned the entire dataset. Deviations from normality within the errors were carefully scrutinized, specifically heteroscedasticity as grounds for disqualifying a model, signifying potential inconsistencies in variance across the dataset. These systematic procedures collectively facilitated a thorough evaluation of biases and accuracies inherent in all proposed models, ensuring a rigorous and reliable analysis.

2.4. Statistical Data Analysis

Statistical analyses were conducted using Statistica v. 8.0 (StatSoft, Tulsa, OK, USA), DataFit v. 8.0.32 (Oakdale Engineering, Oakdale, PA, USA), SigmaPlot for Windows v. 11.0 (Systat Software, Inc., San Jose, CA, USA), and R v. 3.3.3 (CoreTeam, 2020). All graphs were generated by GerminaR [35]. To discern significant variations among variables, ANOVA was applied, and mean comparisons were conducted through a Student–Newman–Keuls test, with a threshold of $p<0.01$ considered statistically significant.

3. Results

3.1. Equation Generation

The leaves of C. pyramidale have small leaflets, with complete edge, and a leathery appearance. The size of the sampled leaves ranged widely in both length and width. The length ranged from 0.80 cm to 3.96 cm and the width ranged from 0.51 cm to 2.51 cm (Supplementary Material (available here)), showing that both length and width have a wide plasticity, since the longest leaf is 4.95-fold longer than the shortest (Table 1) while the widest leaf is 4.92-fold wider than the narrowest. So, leaves were wider than longer because the L : W ratio was 0.71 ± 0.15. In any case, the leaf area was determined using an allometric measure with the greatest amplitude, since the largest leaf has a leaf area 20.47-fold larger than the smallest (Table 1).

We proposed nine different biologically based allometric equations to estimate the leaf area of C. pyramidale (Table 2). All equations, with the exception of equation #2, presented an excellent coefficient of determination ($R^{2}a$), ranging from 0.807 to 0.994 (Supplementary Material). Thus, with the exception of equation #2, all other equations presented an excellent ability to estimate the leaf area of C. pyramidale.

Table 2

Statistical models, regression coefficients (${\beta }_0$ and ${\beta }_1$), standard errors of estimates (SE), mean square error (MS_res), coefficients of determination adjusted for the degrees of freedom (${\mathrm{R}}_{\mathrm{a}}^2$), calculated F (${\mathrm{F}}_{\text{calc}}$), and the derived equations for estimating leaf area in Cenostigma pyramidale leaves based on linear leaf dimensions (length, L, and width, W).

Equation number	Model	Coefficients		SE	MSres	${\mathrm{R}}_{\mathrm{a}}^2$	${\mathrm{F}}_{\text{calc}}$	Estimator
		${\beta }_0$	${\beta }_1$
# 1	$\widehat{Y}={\beta }_1\times \mathrm{W}+{\epsilon }_i$	—	1.573	0.586	0.343	0.934	18,143.22	LA = 1.573 × (W)
# 2	$\widehat{Y}={\beta }_0+{\beta }_1\times W+{\epsilon }_i$	−1.497	2.601	0.461	0.213	0.787	4666.85	LA = −1.497 + 2.601 × (W)
# 3	$\widehat{Y}={\beta }_0\times W^{{\beta }_1}+{\epsilon }_i$	1.126	1.790	0.439	0.193	0.807	5.267	LA = 1.126 × (W)1.790
# 4	$\widehat{Y}={\beta }_1\times \mathrm{L}+{\epsilon }_i$	—	1.093	0.520	0.270	0.948	23,361.20	LA = 1.093 × (L)
# 5	$\widehat{Y}={\beta }_0+{\beta }_1\times L+{\epsilon }_i$	−1.249	1.681	0.400	0.160	0.840	6618.92	LA = −1.249 + 1.681 × (L)
# 6	$\widehat{Y}={\beta }_0\times L^{{\beta }_1}+{\epsilon }_i$	0.645	1.652	0.377	0.142	0.858	7.608	LA = 0.645 × (L)1.652
# 7	$\widehat{Y}={\beta }_1\times LW+{\epsilon }_i$	—	0.725	0.173	0.030	0.994	222,383.86	LA = 0.725 × (LW)
# 8	$\widehat{Y}={\beta }_0+{\beta }_1\times LW+e_i$	0.047	0.712	0.171	0.029	0.971	41,642.48	LA = 0.047 + 0.712 × (LW)
# 9	$\widehat{Y}={\beta }_0\times {LW}^{{\beta }_1}+{\epsilon }_i$	0.746	0.979	0.171	0.029	0.971	41.633	LA = 0.746 × (LW)0.979

LA was more closely dependent on L rather than W, since all equations using $W$ as an independent variable were determined to be unreliable, underestimating the leaf area at values statistically different from zero, then biased; a fact that eliminates the equations #1, #2, and #3 from the process of selecting the best allometric equation to estimate the leaf area of C. pyramidale (Figure 1). Equation #4 also underestimates the LA significantly, a fact that eliminates this equation from the validation process as well (Figure 1).

[figure(s) omitted; refer to PDF]

As a consequence of excluding equations #1, #2, #3, and #4, equations #5, #6, #7, #8, and #9 remain to be validated. Equation #5 presents a relative underestimation of 2.59% and 2.99% of overestimation, totaling a bias of 5.58% (Figure 2), which invalidates it in the search for the best equation to use. Equation #6 has a respective relative underestimation and overestimation rate of 4.92% and 0.79%, totaling a bias of 5.71%, in theory also invalidating this allometric model for estimating the leaf area of C. pyramidale.

[figure(s) omitted; refer to PDF]

With the exclusion of equations #5 and #6, only the models that use the product of length and width (LW) remained to be validated (models #7, #8, and #9). Therefore, these models were used for the allometric estimation of the validation leaf samples. In this case, the proposed allometric equation that generates an estimated leaf area was compared with the measured leaf area (Figure 3). Models #7, #8, and #9, respectively, underestimate the leaf area by 0.87%, 0.63%, and 0.71%, while they overestimate the leaf area by 1.74%, 2.22%, and 1.98% (Figure 3). Thus, the sum of bias of models #7, #8, and #9 are, respectively, 2.61%, 2.85%, and 2.69%. Figure 3 shows that model #9 has an $R^{2}a$ of 0.999, while models #7 and #8 have an $R^{2}a$ of 0.978. Therefore, it is clear that the best model for estimating the leaf area of C. pyramidale is model #9, which returns the equation LA = 0.746 × (LW)^0.979. Furthermore, this model provides a more realistic biological approximation because even with the presence of ${\beta }_0$, if LW is equal to zero, the estimated leaf area will also be zero, a fact that does not happen using model #8, in which, as LW approaches zero would result in a leaf area of 0.047 cm², a biologically inconsistent value.

[figure(s) omitted; refer to PDF]

Between models #7 and #8, we found that both presented an $R^{2}a$ of 0.978 in the validation sample, and even though model #7 results in an $R^{2}a$ of 0.994 in the equation generation phase, model #8 returns an $R^{2}a$ of 0.971. It is common to choose between one of the two models to assist in the allometric determination of the leaf area. In this case, we verified that model #7 is preferred over model #8, since the absence of ${\beta }_0$ in model #7 returns zero value of leaf area if LW is equal to zero, while model #8 returns a positive leaf area, even with LW equal to zero. Thus, as an auxiliary to model #9 utilizing the already selected model #7 could be used. This recommendation comes due to four basic factors: (i) highest $R^{2}a$ among all proposed models; (ii) low mean square error; (iii) being a linear equation, which makes it easier to determine leaf area in large samples in the field; and (iv) absence of ${\beta }_0$ that returns an invalid LA if LW is equal to zero. The linear equations are simpler than the potential equation proposed by model #9. Thus, the linear equation proposed as auxiliary to model #9 returns an equation LA = 0.725 × LW (Table 2).

It is also quite common to propose an allometric model that uses only one of the linear measurements of the leaf. In this case, we should return to models #5 and #6. Model #6, in comparison to model #5, is preferable do to the observation even when ${\beta }_0$ present in both models returns a zero leaf area if the leaf length is zero, while model #5 returns a leaf area of −1.249 cm², even with L equal to zero, i.e., a nonbiological effect. Therefore, it would be recommended to use model #6 which returns an equation LA = 0.645 × L^1.652.

4. Discussion

The use of regression equations to estimate LA is a nondestructive, simple, quick, accurate, reliable, and inexpensive method. In this study, we describe a reliable method to estimate C. pyramidale leaf area with high accuracy and lack of uncertainty using simple mathematic equations. Here, we describe three equations, all with high accuracy, unbiased, and dependable. From these three equations, it is up to the researcher to choose the method that best adapts to their needs at that particular moment, since the equations proposed here are nondestructive, allowing the researcher to estimate the leaf area of a given crop throughout its normal growth cycle. Measurements of leaf length and/or width have been used to estimate leaf area for several fruit and nut trees and vines such as grape [36–38], avocado [39], pistachio [40], cherry [41], myrtle [42], peach [27], strawberry [43], chestnut [44], coffee [18], pitanga [45], pecan [30, 46], sugar beet [47], cacao [32], faba bean [48], kiwi [49], watermelon [50], banana [51], and many others species [19, 26]. As highlighted by Robbins and Pharr [52], the process of selecting a model necessitates finding harmony between the predictive accuracy of the model and the frugality of including only the essential variables required for forecasting LA.

In accordance with Silva et al. [53], C. pyramidale reaches a maximum 60 mm of leaf length. However, this finding does not agree with our dataset of 1260 leaves where the maximum leaf length registered was 39.6 mm. However, our dataset does agree with Maia [54] which describes C. pyramidale as reaching a maximum of 30 mm.

In our study, we observed that the leaves tended to be wider than they were narrow. However, it is important to note that this observation remains speculative, as the ratio of length to width varies significantly among species, primarily due to the intricate diversity in leaf shapes [19]. However, the ratio of length to width of 0.71 ± 0.15 can be described by a shape between an ellipse (0.78) and a triangle (0.5). Our shape ratio agreed closely with 0.69 for pepper [55], 0.63 for sunflower [20], 0.59 for vineyard [36], 0.62 for soybean [56], 0.73 for corn [57], and 0.56 for ficus [58].

In the vast majority of studies with leaf area estimation, both leaf length and leaf weight were considered the best model to explain leaf area without bias and in a reliable method [17, 18, 23, 31, 44]. Montero et al. [36] described that vineyard equations relating LA with LW had one of the highest correlations coefficients with LA and was preferable to other models, which are slightly more accurate. In this study, we verified that all equations that used the product of leaf length and width showed the highest $R^2$, lowest MS_res, and best coefficients of determination in the validation sample. Frequently, the inclusion of a constant (commonly known as the intercept) in linear models presents difficulties in precisely estimating LA [59], especially when dealing with smaller leaves. This disparity tends to diminish in significance as leaf size increases. To address this, Antunes et al. [18] proposed a practical bias correction method, suggesting the elimination of the intercept when estimating leaf area across different genotypes of coffee plants.

Our research highlights the inherent uncertainty associated with the utilization of simplistic linear models, particularly when incorporating a ${\beta }_0$ term into the equation. As a result, we propose a adoption of straightforward linear models that can offer more reliable and interpretable results, like $Y_i{=\beta }_1{X}_i+{\mathrm{\varepsilon }}_i$. The superiority of the nonlinear model #9 can be conceptually explained by using the term ${\text{LW}}^{{\beta }_1}$. This term conveniently facilitates an interactive adjustment that compensates for the penalty incurred when the LW product deviates from the shape of a rectangle. In this study, the term ${\text{LW}}^{{\beta }_1}$ was always close to 1. Equation type $\widehat{Y}={\beta }_0\times {\text{LW}}^{{\beta }_1}+{\epsilon }_i$ was also considered the best for estimating LA of black pepper [60], ginger [61], grapevine [37], dwarf coconut tree [62], coffee [18], purging nut [17], squash [63], triticale [64], and others [23].

Da Vitória [65] agrees with this study because we describe that equations using the product of leaf length and leaf width have a closer approximation than those obtained by one of the other leaf dimensions. Potdar and Pawar [51] argue that the models incorporating both L and W were the best because this gives the highest correlation coefficients ($R^{2}a$) and the lowest coefficients of variation (CV), a fact that disagrees with our study, because the highest CV was shown when we use LW as an independent variable (48.7%), different from the CV obtained using W (CV = 24.9%) or L (27.6%) based equations. Potdar and Pawar [51] describe that for bananas, the best models were those combining L and W, but between those bases in only one leaf dimensions, those based on W data alone were simple and convenient, but the accuracy of LA estimation was diminished to some extent.

Polynomial models that integrate the ${\beta }_0$ parameter (representing the constant, as seen in equations #5 and #8) are inadequate for leaf area estimation, as the model then lacks biological validity when ${\beta }_0$ deviates from zero. This underscores the limitations of relying solely on linear models throughout the entire span of leaf development, emphasizing the crucial need to transition towards the adoption of nonlinear models in future applications [66]. Regardless of its apparent precision and accuracy, the linear model with intercepts, as developed, has also proven to be inadequate, particularly for small leaves [17, 18, 67]. Uncertainties in leaf area measurements using small leaves have previously been reported for several other species [17, 23, 68, 69].

Many scholars [17, 18, 31, 67] affirm that the use of only L or W to estimate LA is not satisfactory because the mean square error (MSE) increases significantly. However, in our study, the MSE ranged from 0.029 to 0.343, with the highest value being 11.8-fold greater than the smallest, an increase that does not, per se, explain the bias associated with equations using only L or W as independent variables, mainly that the equation with highest MSE shows $R^{2}a$ in reliable values. Employing a singular parameter, the models using leaf length were considered the best estimate for the leaf area of some plant species; but this method potentially results in a diminished precision as previously reported in other tree species [17, 18, 22, 31]. In many other species, leaf width provides a better result than leaf length [50, 58, 68]. Moreover, in accordance with Toebe et al. [70], one of the reasons why models originating from a single leaf dimension are less reliable to predict the leaf area is due to the fact that there is no normalizing the dispersion of these data. Moreover, by using only one dimension, it was possible to estimate LA with considerable time savings.

In certain scenarios, the utilization of equation #7, such as $\widehat{Y}={\beta }_1\times LW+{\epsilon }_i$, proves more advantageous compared to equation #9, $\widehat{Y}={\beta }_0\times {LW}^{{\beta }_1}+{\epsilon }_i$ because it involves simpler computations [39]. Equation #7 exhibited notable dispersion of residuals, demonstrating consistent homoscedasticity. This characteristic enhances the viability of equation #7 as the preferred option for modeling the allometry of C. pyramidale. This finding corresponds with the trends observed with the power equation, underscoring the strength and effectiveness of our proposed model [16, 17, 63]. To summarize, equations #6, $\widehat{Y}={\beta }_0\times L^{{\beta }_1}+{\epsilon }_i$, and #9, proposed in our study, adhere to statistical principles, with the exception that equation #9 may demonstrate greater stability during validation than equation #6.

5. Conclusions

Our findings suggest that employing LW as an independent variable (equation #9; LA = 0.746 × (LW)^0.979) provides a precise estimation of Cenostigma pyramidale leaf area. This equation offers a practical solution, allowing agronomists and researchers to accurately assess the leaf area of C. pyramidale plants, eliminating the need for costly tools such as leaf area planimeters or digital cameras. Here, we described a reliable method to estimate leaf area through a nondestructive method. So, the researcher could estimate with high accuracy the leaf area using a simple ruler. In larger experiments, the researcher could need a smartphone or a tablet to input the data at the same time, avoiding draft and postinput data in a PC. The results indicated that basic linear measurements could be employed to predict C. pyramidale leaflet and overall leaf area. To turn cause the process to be quicker, the leaf area estimation could be estimated using equation #7 (LA = 0.725 × (LW) because the linear equation is easier than power models, proposed by equation #9, especially in field conditions. Furthermore, we could recommend the equation #6 (LA = 0.645 × L^1.652), which requires the measurement of only one leaf dimension (leaf length) as independent variable. However, the best method is dependent on how much plant material can be measured, the accuracy required, and the time and personnel available. This decision is researcher-dependent.

Authors’ Contributions

MFP conceptualized, validated, and investigated the study, developed methodology and software, performed formal analysis, collected resources, wrote, reviewed, and edited the article, administered the project, and involved in funding acquisition. A.J.-O. conceptualized and validated the study, collected resources, wrote, reviewed, and edited the article, and involved in funding acquisition. J.D.J.-N conceptualized the study, wrote, reviewed, and edited the article, administered the project, and involved in funding acquisition. Y.Y.P.-R.: software and formal analysis. L.A.R.-P. conceptualized the study, wrote, reviewed, and edited the article, and involved in funding acquisition. All authors have read and agreed to the published version of the manuscript.

References

[1] R. Pixoto, "Como a recente descoberta do primeiro clima árido no Brasil pode impactar o restante do país," 2024.

[2] J. M. C. Silva, I. R. Leal, M. C. Tabarelli, The Largest Tropical Dry forest Region in South America, 2017.

[3] R. M. Vieira, J. Tomasella, A. A. Barbosa, M. A. Martins, D. A. Rodriguez, F. S. Rezende, F. Carrielo, M. O. Santana, "Desertification risk assessment in Northeast Brazil: current trends and future scenarios," Land Degradation and Development, vol. 32, pp. 224-240, DOI: 10.1002/ldr.3681, 2020.

[4] J. A. Marengo, R. R. Torres, L. M. Alves, "Drought in Northeast Brazil-past, present, and future," Theoretical and Applied Climatology, vol. 129, pp. 1189-1200, DOI: 10.1007/s00704-016-1840-8, 2017.

[5] IPCC, "Deforestation accounts for about 20% of CO2 emissions globally," 2021. https://www.climatecentral.org/what-we-do#wwd

[6] H. F. P. Araujo, N. F. Canassa, C. C. C. Machado, M. Tabarelli, "Human disturbance is the major driver of vegetation changes in the Caatinga dry forest region," Scientific Reports, vol. 13,DOI: 10.1038/s41598-023-45571-9, 2023.

[7] J. B. de Jesus, D. G. de Oliveira, W. S. Araújo, L. S. da Cruz, T. M. Kuplich, "Influence of anthropization on the floristic composition and phytosociology of the Caatinga susceptible to desertification in the state of Sergipe, Brazil," Tropical Ecology, vol. 63, pp. 398-408, DOI: 10.1007/s42965-021-00201-1, 2022.

[8] G. S. R. Gonçalves, P. V. Cerqueira, D. P. Silva, L. B. Gomes, C. F. Leão, A. F. A. de Andrade, M. P. D. Santos, "Multi-temporal ecological niche modeling for bird conservation in the face of climate change scenarios in Caatinga, Brazil," PeerJ, vol. 11,DOI: 10.7717/peerj.14882, 2023.

[9] I. M. C. Ribeiro, E. C. P. Arruda, A. F. M. Oliveira, J. S. Almeida, "Physical and chemical defenses of Cenostigma pyramidale (Fabaceae): a pioneer species in successional caatinga areas," Rev Caatinga, vol. 34, pp. 398-409, DOI: 10.1590/1983-21252021v34n216rc, 2021.

[10] J. A. B. Oliveira, J. S. Almeida, "Do the successional stage and plant architecture influence herbivory by insects in Cenostigma pyramidale in a dry tropical forest?," International Journal of Environmental Analytical Chemistry, vol. 6, pp. 352-362, DOI: 10.24221/jeap.6.4.2021.3616.352-362, 2021.

[11] C. L. Wright, J. B. West, A. L. A. de Lima, E. S. Souza, M. Medeiros, B. P. Wilcox, "Contrasting water use strategies revealed by species-specific transpiration dynamics in the Caatinga dry forest," Tree Physiology, vol. 2023,DOI: 10.1093/treephys/tpad137, 2023.

[12] A. C. B. Pinho-Pessoa, K. R. Mendes, A. Jarma-Orozco, M. P. S. Pereira, M. A. Santos, R. S. C. Menezes, J. P. Ometto, E. C. Pereira, M. F. Pompelli, "Interannual variation in temperature and rainfall can modulate the physiological and photoprotective mechanisms of a native semiarid plant species," Indian Journal of Science and Technology, vol. 11,DOI: 10.17485/ijst/2018/v11i42/130972, 2018.

[13] M. F. Pompelli, A. Jarma-Orozco, J. d. D. Jaraba-Navas, Y. Y. Pineda-Rodriguez, L. A. Rodríguez-Páez, "Interannual variation in gas exchange and leaf anatomy in Cenostigma pyramidale is exacerbated through El Niño and La Niña climate events," Forests, vol. 15,DOI: 10.3390/f15020346, 2024.

[14] R. T. Maciel Filho, A. V. Lacerda, A. C. Gomes, J. S. O. Araújo, F. M. Barbosa, "Efeito sazonal na vegetação lenhosa do estrato regenerante em mata ciliar no cariri Paraibano," Potencialidades do Bioma Caatinga: Marcas de convivência e resistência, pp. 16-19, 2016.

[15] A. V. Lacerda, F. Barbosa, "Woody component structure of a riparian forest in an intermittent stream in the semi-arid region of Paraíba, Brazil," Floresta, vol. 50, pp. 1565-1574, DOI: 10.5380/rf.v50i3.64520, 2020.

[16] J. Kaal, A. Martínez Cortizas, K. G. J. Nierop, "Characterisation of aged charcoal using a coil probe pyrolysis-GC/MS method optimised for black carbon," Journal of Analytical and, vol. 85, pp. 408-416, DOI: 10.1016/j.jaap.2008.11.007, 2009.

[17] M. F. Pompelli, W. C. Antunes, D. T. R. G. Ferreira, P. P. G. S. Cavalcante, H. C. L. Wanderley-Filho, L. Endres, "Allometric models for non-destructive leaf area estimation of the Jatropha curcas," Biomass and Bioenergy, vol. 36, pp. 77-85, DOI: 10.1016/j.biombioe.2011.10.010, 2012.

[18] W. C. Antunes, M. F. Pompelli, D. M. Carretero, F. M. DaMatta, "Allometric models for non-destructive leaf area estimation in coffee ( Coffea arabica and Coffea canephora )," Annals of Applied Biology, vol. 153, pp. 33-40, DOI: 10.1111/j.1744-7348.2008.00235.x, 2008.

[19] S. K. Pandey, H. Singh, "A simple, cost-effective method for leaf area estimation," Journal de Botanique, Le, vol. 2011,DOI: 10.1155/2011/658240, 2011.

[20] Y. Rouphael, G. Colla, S. Fanasca, F. Karam, "Leaf area estimation of sunflower leaves from simple linear measurements," Photosynthetica, vol. 45, pp. 306-308, 2007.

[21] H. Demirsoy, "Leaf area estimation in some species of fruit tree by using models as a non-destructive method," Fruits, vol. 64, pp. 45-51, 2009.

[22] F. F. Blanco, M. V. Folegatti, "Estimation of leaf area for greenhouse cucumber by linear measurements under salinity and grafting," Scientia Agricola, vol. 62, pp. 305-309, DOI: 10.1590/S0103-90162005000400001, 2005.

[23] J. N. B. Santos, A. Jarma-Orozco, W. C. Antunes, K. R. Mendes, J. M. Figueiroa, L. M. Pessoa, M. F. Pompelli, "New approaches to predict leaf area in woody tree species from the Atlantic Rainforest, Brazil," Austral Ecology, vol. 46, pp. 613-626, DOI: 10.1111/aec.13017, 2021.

[24] J. Schrader, G. Pillar, H. Kreft, "Leaf‐IT: an android application for measuring leaf area," Ecology and Evolution, vol. 7, pp. 9731-9738, DOI: 10.1002/ece3.3485, 2017.

[25] R. Adhikari, C. Li, K. Kalbaugh, K. Nemali, "A low-cost smartphone controlled sensor based on image analysis for estimating whole-plant tissue nitrogen (N) content in floriculture crops," Computers and Electronics in Agriculture, vol. 169,DOI: 10.1016/j.compag.2019.105173, 2020.

[26] P. Kaushik, P. K. Pati, M. L. Khan, P. K. Khare, "A quick and simple method for estimating leaf area by leaf weight," International Journal of Botany Studies, vol. 6, pp. 1286-1288, 2021.

[27] H. Demirsoy, L. Demirsoy, S. Uzun, B. Ersoy, "Non-destructive leaf area estimation in Peach," European Journal of Horticultural Science, vol. 69, pp. 144-146, 2004.

[28] Y. Rouphael, A. H. Mouneimne, E. Mendoza-de-Gyves, C. M. Rivera, G. Colla, "Modeling individual leaf area of rose ( Rosa hybrida L.) based on leaf length and width measurement," Photosynthetica, vol. 48, 2010.

[29] L. A. Rodriguez-Paez, A. Jarma-Orozco, J. d. D. Jaraba-Navas, Y. Y. Pineda-Rodriguez, M. F. Pompelli, "Leaf area estimation of yellow oleander Thevetia peruviana (pers.) K. Schum using a non-destructive allometric model," Forests, vol. 15,DOI: 10.3390/f15010057, 2024.

[30] J. L. Whitworth, A. Mauromoustakos, M. W. Smith, "A nondestructive method for estimation of leaf area in pecan," Horticultural Science (Stuttgart), vol. 27, 1992.

[31] A. Shabani, A. R. Sepaskhah, "Leaf area estimation by a simple and non-destructive method," Iran Agricultural Research, vol. 36, pp. 101-105, 2017.

[32] J. C. S. Salazar, L. M. Melgarejo, E. H. D. Bautista, J. A. Rienzo, F. Casanoves, "Non-destructive estimation of the leaf weight and leaf area in cacao ( Theobroma cacao L.)," Scientia Horticulturae, vol. 229, pp. 19-24, 2018.

[33] M. Steel, D. Penny, "Parsimony, likelihood, and the role of models in molecular phylogenetics," Molecular Biology and Evolution, vol. 17, pp. 839-850, 2000.

[34] G. Cumming, F. Fidler, D. L. Vaux, "Error bars in experimental biology," The Journal of Cell Biology, vol. 177, 2007.

[35] F. Lozano-Isla, O. B. Alfaro, M. F. Pompelli, "Germinar: An r package for germination analysis with the interactive web application 'germinaquant for r," Ecological Research, vol. 34, pp. 339-346, DOI: 10.1111/1440-1703.1275, 2019.

[36] F. J. Montero, J. A. de Juan, A. Cuesta, A. Brasa, "Nondestructive methods to estimate leaf area in Vitis vinifera L," Horticultural Science (Stuttgart), vol. 35, pp. 696-698, DOI: 10.21273/HORTSCI.35.4.696, 2000.

[37] L. Williams, T. E. Martinson, "Nondestructive leaf area estimation of ‘Niagara’ and “DeChaunac” grapevines," Scientia Horticulturae-Amsterdam, vol. 98, pp. 493-498, DOI: 10.1016/S0304-4238(03)00020-7, 2003.

[38] G. R. Sepúlveda, W. M. Kliewer, "Estimation of leaf area of two grapevine cultivars ( Vitis vinifera L.) using laminae linear measurements and fresh weight," American Journal of Enology and Viticulture, vol. 34, pp. 221-226, 1983.

[39] S. Uzun, H. Çelik, "Leaf area prediction models (Uzçelik-I) for different horticultural plants," Turkish Journal of Agriculture and Forestry, vol. 23, pp. 645-650, 1999.

[40] A. Ranjbar, P. Damme, "Estimation of leaf area by nondestructive methods in three ranian pistachio species ( Pistacia mutica subsp. Cabulica, Pistacia khinjuk subsp. blonda and Pistacia khinjuk subsp. Populifolia)," Mededelingen-Faculteit Landbouwkundige en Toegepaste BiologischeWtenschappen Universiteit Gent, vol. 64, pp. 49-56, 1999.

[41] H. Demirsoy, L. Demirsoy, "A validated leaf area prediction model for some cherry cultivars in Turkey," Pakistan Journal of Botany, vol. 35, pp. 361-367, 2003.

[42] G. Gugliuzza, G. Fascella, M. M. Mammano, M. Militello, "Non-destructive leaf area estimation in Myrtus communis plants," Acta Horticulturae, vol. 1104,DOI: 10.17660/ActaHortic.2015.1104.14, 2015.

[43] H. Demirsoy, L. Demirsoy, A. Öztürk, "Improved model for the non-destructive estimation of strawberry leaf area," Fruits, vol. 60, pp. 69-73, 2005.

[44] Ü. Serdar, H. Demirsoy, "Non-destructive leaf area estimation in chestnut," Sci Hortic-Amsterdam, vol. 108, pp. 227-230, 2006.

[45] M. F. Pompelli, J. M. Figueirôa, F. Lozano-Isla, "Allometric models for non-destructive leaf area estimation in Eugenia uniflora (L.)," Journal of Agronomy, vol. 2,DOI: 10.21704/pja.v2i2.1133, 2018.

[46] S. I. Torri, C. Descalzi, E. Frusso, "Estimation of leaf area in pecan cultivars ( Carya illinoinensis )," Ciencia e Investigacian Agraria, vol. 36, pp. 53-58, 2009.

[47] J. T. Tsialtas, N. Maslaris, "Leaf area estimation in a sugar beet cultivar by linear models," Photosynthetica, vol. 43, pp. 477-479, 2005.

[48] E. Peksen, "Non-destructive leaf area estimation model for faba bean ( Vicia faba L.)," Sci Hortic-Amsterdam, vol. 113, pp. 322-328, 2007.

[49] E. Mendoza-de-Gyves, Y. Rouphael, V. Cristofori, R. Mira, "A nondestructive, simple and accurate model for estimating the individual leaf area of kiwi ( Actinidia deliciosa )," Fruits, vol. 62, pp. 171-176, 2007.

[50] Y. Rouphael, A. H. Mouneimne, C. M. Rivera, M. Cardarelli, A. Marucci, G. Colla, "Allometric models for nondestructive leaf area estimation in grafted and ungrafted watermelon ( Citrullus lanatus Thunb)," Journal of Food Agriculture and Environment, vol. 8, pp. 161-165, DOI: 10.1234/4.2010.1475, 2010.

[51] M. V. Potdar, K. R. Pawar, "Non-destructive leaf area estimation in banana," Sci Hort Amsterdam, vol. 45, pp. 251-254, DOI: 10.1016/0304-4238(91)90070-F, 1991.

[52] N. S. Robbins, D. M. Pharr, "Leaf area prediction models for cucumber from linear measurements," HortScience, vol. 22, pp. 1264-1266, DOI: 10.21273/HORTSCI.22.6.1264, 1987.

[53] J. A. Silva, J. E. F. Bezerra, L. H. Silva, V. E. Marques, A. V. Lacerda, "Analysis of phytosociological parameters and spatial distribution of Cenostigma pyramidale (tul.) e. Gagnon and G.P. Lewis in a regenerating stratum of a riparian forest in the semi-arid region of Paraíba," Revista Geografar, vol. 40, pp. 255-267, DOI: 10.51359/2238-6211.2023.258023, 2023.

[54] G. N. Maia, Caatinga: Árvores e Arbustos e Suas Utilidades, 2004.

[55] E. A. M. De Swart, R. Groenwold, H. J. Kanne, P. Stam, L. F. M. Marcelis, R. E. Voorrips, "Non-destructive estimation of leaf area for different plant ages and accessions of Capsicum annuum L," The Journal of Horticultural Science and Biotechnology, vol. 79, pp. 764-770, DOI: 10.1080/14620316.2004.11511840, 2004.

[56] J. V. Wiersma, T. B. Bailey, "Estimation of leaflet, trifoliate, and total leaf areas of soybean," Agronomy Journal, vol. 67, pp. 26-30, DOI: 10.2134/agronj1975.00021962006700010007x, 1975.

[57] G. W. McKee, "A coefficient for computing leaf area in hybrid corn," Agronomy Journal, vol. 56, pp. 240-241, 1964.

[58] M. Giaccone, A. Pannico, P. Scognamiglio, C. M. Rivera, C. Cirillo, Y. Rouphael, S. De Pascale, B. Basile, "Regression model for leaf area estimation in Ficus carica L," Acta Horticulturae, vol. 2017,DOI: 10.17660/ActaHortic.2017.1173.27, 2017.

[59] S. Chatterjee, A. S. Hadi, Regression Analysis by Example, 2006.

[60] K. Kandiannan, C. Kailassam, K. K. Chandaragiri, N. Sankaran, "Allometric model for leaf area estimation in black pepper ( Piper nigrum L.)," Journal of Agronomy and Crop Science, vol. 188, pp. 138-140, DOI: 10.1046/j.1439-037X.2002.00540.x, 2002.

[61] K. Kandiannan, U. Parthasarathy, K. S. Krishnamurthy, C. K. Thankamani, V. Srinivasan, "Modeling individual leaf area of ginger ( Zingiber officinale Roscoe) using leaf length and width," Sci Hortic-Amsterdam, vol. 120, pp. 532-537, DOI: 10.1016/j.scienta.2008.11.037, 2009.

[62] E. F. Sousa, M. C. Araújo, R. P. Posse, E. Detmann, S. Bernardo, P. A. Berbert, P. A. Santos, "Estimating the total leaf area of the green dwarf coconut tree ( Cocos nucifera L.)," Scientia Agricola, vol. 62, pp. 597-600, 2005.

[63] M. Toebe, R. R. Souza, A. C. Mello, P. J. Melo, A. Segatto, A. C. Castanha, "Leaf area estimation of squash ‘Brasileirinha’ by leaf dimensions," Ciência Rural, vol. 49,DOI: 10.1590/0103-8478cr20180932, 2019.

[64] M. Toebe, P. J. de Melo, R. R. de Souza, A. C. Mello, F. L. Tartaglia, "Leaf area estimation in triticale by leaf dimensions," Revista Brasileira de Ciência Avícola, vol. 14,DOI: 10.5039/agraria.v14i2a5656, 2019.

[65] E. L. da Vitória, I. L. J. Freitas, T. LocatellI, E. G. Lacerda, J. M. Valle, R. C. Pereira, P. F. P. de Almeida, R. Z. da Vitória, C. P. Simon, A. A. Fernandes, "Mathematical models for leaf area estimates of guava," Journal of Agricultural Science, vol. 10,DOI: 10.5539/JAS.V10N12P272, 2018.

[66] A. F. Zuur, N. I. Elena, C. S. Elphick, "A protocol for data exploration to avoid common statistical problems," Methods in Ecology and Evolution, vol. 1, 2009.

[67] M. C. Souza, C. L. Amaral, "Non-destructive linear model for leaf area estimation in Vernonia ferruginea Less," Brazilian Journal of Biology, vol. 75, pp. 152-156, 2015.

[68] V. H. Ramirez-Builes, T. G. Porch, E. W. Harmsen, "Development of linear models for estimation of leaflet area in common bean ( Phaseolus vulgaris L.)," Journal of Agriculture of the University of Puerto Rico, vol. 92, pp. 171-182, DOI: 10.46429/jaupr.v92i3-4.2634, 2008.

[69] W. C. Antunes, K. R. Mendes, A. R. M. Chaves, J. P. Ometto, A. Jarma-Orozco, M. F. Pompelli, "Spondias tuberosa trees grown in tropical, wet environments are more susceptible to drought than those grown in arid environments," Rev Col Ciên Hortic, vol. 10,DOI: 10.17584/rcch.2016v10i1.4456, 2016.

[70] M. Toebe, A. Cargnelutti Filho, C. Burin, A. L. Fick, I. M. M. Neu, G. Casarotto, B. M. Alves, "Leaf area prediction models for jack bean by leaf dimensions," Bragantia, vol. 71, pp. 37-41, DOI: 10.1590/S0006-87052012005000010, 2012.