Evaluation of MOD16 Algorithm over Irrigated Rice Paddy Using Flux Tower Measurements in Southern Brazil

Vanessa De Arruda Souza,Débora Regina Roberti,Tamíres Zimmer,Marcelo Bortoluzzi Diaz,Daniela Santini Adamatti,Rita De Cássia Marques Alves,Osvaldo L L De Moraes,Luis Gustavo G De Gonçalves,Anderson Luis Ruhoff

doi:10.3390/w11091911

Abstract

Evapotranspiration (ET) is an important component of the hydrological cycle. Understanding the ET process has become of fundamental importance given the scenario of global change and increasing water use, especially in the agricultural sector. Determining ET over large agricultural areas is a limiting factor due to observational data availability. In this regard, remote sensing data has been used to estimate ET. In this study, we evaluated the Moderate-Resolution Imaging Spectroradiometer (MODIS) land surface ET product estimates (hereafter MOD16 ET – MODIS Global Terrestrial Evapotranspiration Product) over two rice paddy areas in Southern Brazil, through the ET measured using the eddy covariance technique (hereafter EC). The energy balance components were evaluated during fallow and flooded seasons showing latent heat flux dominates in both seasons. The results showed that MOD16 ET underestimated EC measurements. Overall, the RMSE (root mean square error) ranged between 13.40 and 16.35 mm 8-day−1 and percent bias (PBIAS) ranged between −33.7% and −38.7%. We also assessed the ET (measured and estimated) main drivers, with EC yielding higher correlation against observed net radiation (Rn) and global radiation (Rg), followed by air temperature (Temp) and vapor pressure deficit (VPD), whilst MOD16 ET estimates yielded higher correlation against leaf area index (LAI) and fraction of photosynthetically active radiation (fPAR). The MOD16 algorithm was forced with meteorological measurements but the results did not improve as expected, suggesting a low sensitivity to meteorological inputs. Our results indicated when a water layer was present over the soil surface without vegetation (LAI around zero), the largest differences between EC measurements and MOD16 ET were found. In this period, the expected domain of soil evaporation was not observed in MOD16 ET physical processes partition, indicating the algorithm was not able to detect areas with high soil moisture. In general, the MOD16 ET product presented low accuracy when compared against experimental measurements over flooded rice paddy, suggesting more studies are necessary, in order to reduce uncertainties associated to the land cover conditions.

Highlights

The increasing competition for water associated to its scarcity is becoming a threat to sustainable development [1]
Our results indicated when a water layer was present over the soil surface without vegetation (LAI around zero), the largest differences between eddy covariance (EC) measurements and MOD16 ET were found
According to Timm et al [37] when ∆G is not included in the energy balance analysis, lower r coefficients are typically found, suggesting that the lower energy balance closure is due to uncertainties in the estimate of heat storage capacity, especially during the irrigation season and changes in water temperature measured in a single depth [66]

Summary

Introduction

The increasing competition for water associated to its scarcity is becoming a threat to sustainable development [1]. In a climate changing scenario associated with an increasing global water demand for cropland production, that currently accounts for approximately 70% of the world water consumption [2], water footprint quantification is fundamental to improve water management and mitigate future freshwater scarcity [3]. One of the most challenging issues in modern global agricultural management is the proper control of the amount of water made available for plantation. A correct estimation of ET over irrigation-based areas is critical to improve productivity while reducing water use as well as to mitigate the impact of extreme events in agriculture, such as droughts [4]. Determining ET over large agricultural areas is a challenging task due to limited observational data availability. Remote sensing techniques have been largely used to estimate ET over large areas with high spatial and temporal resolution and high accuracy [6]

Objectives

Methods

Results

Conclusion