Evaluation of the Performance of the Water Cloud Model and Modified Water Cloud Model in Estimating Soil Moisture. A case study of Kiruuli Village

Abstract

In Uganda, crop yields have been constrained by recurrent droughts and reliance on rain-fed agriculture. As a straightforward measure, irrigation farming has been adopted by the government through its rehabilitation of old schemes and its assistance to farmers in the setting up of micro-irrigation farms. Of consequence is the fact that the maximization of crop yields through irrigation necessitates soil moisture data for irrigation scheduling. Both ground-based measurements and remote sensing techniques can be used to access this information, with the latter holding the advantage of gathering more information over a wider area. Because of its ability to account for vegetation cover, the Water Cloud Model (WCM) ─ a remote sensing-based model ─ has been widely used in earlier studies to estimate soil moisture content over vegetated areas. However, the accuracy of the model is limited by the assumption that vegetation is a homogenous scatterer. Thus, the Modified Water Cloud Model (MWCM) was developed in accordance with the debate that by considering the heterogeneous scattering nature of the vegetation, it would perform better than the WCM. Using Kiruuli Village (in a coffee-growing area), this study compared the performance of the WCM and the Modified Water Cloud Model (MWCM) in estimating soil moisture. The models were implemented using Sentinel 1 and 2 images acquired on 05 September 2021 and 02 August  2021, respectively. Results showed that the MWCM performed slightly better than the WCM with Root Mean Square Errors (RMSEs) of 3.3346 and 3.7482, respectively. The marginality of the results can be attributed to a relatively high vegetation fraction at the time of image acquisition and a reasonably small area of comparison. Generally, more work can be carried out to compare the models across a larger area with a sparser vegetation cover.