Grass Reference Lysimetric Evapotranspiration Measurements Compared to Satellite Data Observations

Abstract

We present one of the first research work for the study site (Botucatu-SP, Brazil) that compares in-situ grass reference evapotranspiration measurements obtained by weighing lysimetry technique with estimated data from this same parameter by the use of satellites. Data from two datasets available at the Google Earth Engine platform were used to compare monthly average values of reference evapotranspiration (ET0): 1. FLDAS (Famine Early Warning Systems Network (FEWS NET) Land Data Assimilation System); 2. MOD16A2.006 (Terra Net Evapotranspiration 8-Day Global 500m). Statistical analyses were carried out to analyze statistical variability between the compared data. At a 5% significance level, the test for homogeneity of variances (Bartlett’s test) indicated that the variances cannot be considered homogeneous (p-value of 4.3%); the Normality test (Shapiro-Wilk’s test) indicated that the observed data follow a normal distribution (p-value of 62.4%); the analysis of variance (ANOVA; Welch Test) with the F-Snedecor’s test indicated at a 5% significance level that the observed mean values do not differ from each other (p-value of 34.5%; Fcalc < F). Annual ET0 average values and their respective standard deviations are presented here: 2.3 ± 1.1 mm.day-1 (Lysimeter); 2.7 ± 0.8 mm.day-1 (MOD16A2.006); 3.2 ± 1.7 mm.day-1 (FLDAS); the coefficient of variation obtained was of 45.7%. Estimated data from MOD16A2.006 dataset overestimates lysimetric measurements by 8% (R² = 0.94), which we consider to be an exceptional approximation between estimated and measured ET0, whereas estimated data obtained from the FLDAS dataset overestimates lysimetric measurements by 33% (R² = 0.90).