Equações alométricas para estimativa da área foliar de espécies lenhosas de floresta ombrófila de terras baixas

Abstract

In forest species the leaf area verification can be a base supply to various types of studies directed to ecology and ecophysiology. The use of allometric equations to estimate the real leaf area presents itself as a fundamental tool to simplify the measurement of this variable. This study aimed to build allometric equations to estimate leaf area of 14 forest species, from linear measurements given in the leaf blade.For each tree species was collected 600, selected at different levels of treetops from, located within experimental plots in a remaining of Lowland Rainforest, in Paulista-PE. All leaves were digitalized and the leaf area (LA) was determined by using the Image-Pro® Plus software, the corresponding value of the maximum length (L) and maximum width (W) were taken. Linear and nonlinear models were generated based on leaves dimensions, where LA was the dependent variable,„L‟, „W‟ and LxW independent variables. Statistical criteria for models selection were based on the examination of the Analysis of variance table (F≤0.01), adjusted coefficient of multiple determination (R²aj), standard error of the estimate (Sxy), mean squared residue (MSR) and graphic criteria of statistical analysis. Models validation was performed by an independent leaves sample (n=200), the real and predicted values of LA were submitted to Student's t-test (p <0.01). To determine the universal equation, LA values were re-estimated using the main equations of each species and the leaves independents, the efficiency of these estimates was evaluated by Student's t-test (p <0.01). In general, 126 adjusted equations present statistical significance (p <0.01) and high Raj 2, showing the existence of a high correlation between LA and the linear dimensions of the leaf blade. However, only 25 equations were validated, including among these, main and alternatives equations, for most species. Simple and modified linear models that consider the use of a single dimension proved to be inappropriate in its entirety. In contrast, all non-linear models, based on the product of the dimensions (LxW), as the potential equation predominantly estimate the leaf area of the 14 forest species. Later the alternative equation that estimated the LA is the linear model, modified by the withdrawal of the interception, can be used without loss of quality of the estimates, is recommended because of its simplicity. Potential models of single dimension, validated to A. discolor (Ŷ=1,9474L1,9287) and H. rubriflora (Ŷ=0,3210C 2,0394) in this study can be used, but with a noticeable loss of accuracy. The proposed equations for each species are unique, represent good estimators of real LA and can be used safely and reliably in several studies. The models based on a single dimension can be used, being recommended for experiments that allow sensitive loss of accuracy. Finally, it was shown that the leaves of E. ovata, L. pisonis, P. heptaphyllum, T. spruceanum and T. guianensis, have a similar leaf blade morphotype, thus, can be estimated by a single universalized equation (Ŷ=0,5822(LW)1,0476) without significant loss of accuracy.