Uso do satélite Sentinel na correlação de parâmetros físico hídricos do solo em área cafeeira

Abstract

Agriculture uses several techniques to increase crop productivity. The use of irrigation is a technique that allows the increase and guarantee of crop productivity. The use of a set of remote sensing techniques can improve water management at different spatial and temporal scales, allowing the monitoring of large irrigated areas. In this context, the objective of this work is to apply spectral indices derived from Sentinel 2 on three images of different dates (06/24/2017, 06/29/2018 and 07/09/2019), taking advantage of the Minimum Sequential Optimization (SMOreg) algorithms and Simple Linear Regression (RLS), to constitute, within the agricultural sciences, prediction models capable of representing the physical-water parameters available water capacity of plants (CAD), field capacity humidity (CC) and permanent wilting point (PMP ) quickly and accurately. For this purpose, 21 points were georeferenced in a 4.1 ha coffee plantation area. At each point, an undisturbed soil sample was collected in the 0 to 0.2 m layer. The models were generated using the Waikato Environment for Knowledge Analysis (WEKA) software. In all models generated, on the three dates, the non-parametric models stood out with the Square Root Mean Error (RMSE) smaller than the parametric models. There was a predominance of use of the Normalized Moisture Difference Index (NDMI) to estimate WC, use of the Single Ratio Moisture Stress Index (MSI) to estimate PMP and predominant use of NDMI and MSI in CAD models. Analyzing the confidence index (c), the models generated on the three dates to estimate CAD were the only satisfactory ones, and for both CAD, CC and PMP, Near Infrared (NIR) and shortwave infrared bands were used (SWIR).