Impacts of corn flowering on estimation of evapotranspiration using remote sensing in Brazilian Savanna

Estimation of evapotranspiration using remote sensing is a promising and low-cost alternative, but there is a lack of studies to calibrate the algorithm for different crops and atmospheric conditions. In this context, the objective of the study was to evaluate the efficiency of the SAFER algorithm in estimating evapotranspiration of corn crops (ETa) in three different sources of surface albedo. The study was carried out in a corn (cultivar AG8700) production (March to Jully) area irrigated by central pivot in Itaberaí-GO, Brazil, in 2021. The region's climate is characterized as Aw, with two well-defined seasons, dry winter and rainy summer, and the soil was classified as Red Oxisol of medium texture. Images from a multispectral and thermal camera model MicaSense Altum and albedo images from the Landsat 8 (each 16 days) and Sentinel 2A (each 10 days) satellites were used to estimate the ETa using the SAFER algorithm. These data were compared with ETc (Crop evapotranspiration) obtained by FAO, Embrapa and climatological water balance methods based on statistical indices. In general, the best correlation with standard methods was the Drone method, mainly the FAO and BHC methods. On average, the EQM (mean square error) was less than 0.22 mm day-1. The agreement index ranged from 0.84 to 0.91. The largest errors were observed in phase III, due to contamination of albedo and NDVI pixels caused by screwing. This error was greater for the DroneLand and DroneSent methods. On average the EQM and EMA (mean absolute error) were close to 1 mm day-1, the confidence index was below 0.74 for all methods. Thus, the use of images from multispectral and thermal cameras proved to be a good tool for estimating evapotranspiration. Corn crop flowering interferes with the estimated values of NDVI and surface albedo, leading to greater underestimation. However, when drone images are used to calculate the three SAFER variables, this effect is mitigated. ETa estimation by remote sensing is not recommended for the reproductive phase of corn crop.