Artificial neural network for the estimation of soil moisture using earth observation datasets

Abstract

Surface Soil Moisture (SSM) is an important variable in agricultural water management, required for irrigation water demand, scheduling, etc. In this chapter, the estimation of SSM is carried out using Artificial Neural Network (ANN) model trained by MODIS land surface temperature (LST) and normalised difference vegetation index (NDVI) feature spaces and validated using the in-situ data. The ANN model is trained, validated and tested using the three different combinations of input-output datasets. The first combination of datasets is considered as MODIS LST (input) and in-situ SSM (output) datasets for ANN-I model. The second combination of datasets is considered as MODIS NDVI (input) and in-situ SSM (output) datasets for ANN-II model. The third combination of datasets is considered as MODIS LST and NDVI (input) and in-situ SSM (output) datasets for ANN-III model. The performance of ANN-I, ANN-II and ANN-III models are evaluated in terms of correlation coefficient (r), bias and root mean squared error. In overall, the performance of ANN-II model was found good for SSM estimation. © 2021 Elsevier Inc. All rights reserved.