Water deficit detection in sugarcane using canopy temperature from satellite images

Abstract

Water deficit (WD) is the main yield gap for sugarcane in Midwest Brazil. Thus, WD detection is essential to quantify yield losses, but field detection requires measurement of soil water content over large areas. In this study, we tested leaf temperature (TL) and land surface temperature (TS) to detect WD in a commercial sugarcane area. The area is located in the central region of Goiás State, Brazil. According to Köppen classification, the climate of the region is Aw (humid tropical, with rainy summer and dry winter). The soil is a Ferralsol (clayey texture). TL was measured by a portable infrared thermometer, and TS was obtained using a spectral image from Landsat 8. Both TL and TS measurements occurred between 28 Jan and 24 Aug 2014 (298-506 DAP). The water balance identified periods of water deficit (WD) and surplus (WS). The difference between TL-Ta was greater than zero (7.11 °C) in WD periods and lower than zero (-2.18 °C) in WS periods. The difference between TS-Ta, in turn, ranged from -0.66 °C to 4.06 °C, but not following the tendency of WD or WS, which is associated with a relative error between TL and TS near 20% for some date. The TS-Ta difference detected soil WD or WS when the relative error was low (362 and 410 DAP) and under higher WD (506 DAP) and WS (394 DAP). This way, TL was able to detect WD and WS along sugarcane growth, while TS showed limited application, requiring improvement based on surface properties to reduce the error in relation to TL. Furthermore, bands 10 and 11 are recommended for surface temperature estimation. Calibration uncertainty increases when the band 11 is used alone, being this band more affected by the absorption of radiation by the atmospheric water vapor, which implies larger errors related to the atmospheric profile in the acquisition of surface temperature.