Evaluation of the performance of the water cloud model and the modified water cloud model in estimating soil moisture.

Abstract

Periodic droughts and reliance on rain-fed agriculture in Uganda have limited the crop yield obtained in the country. Irrigation farming as a straightforward measure has been adopted by the government through rehabilitation of old schemes and helping farmers set up micro-irrigation farms. The maximization of crop yield through irrigation however necessitates soil moisture information for irrigation scheduling. The water cloud model currently investigated by (Abdulrahamani, 2019) to aid in attaining soil moisture data over wide vegetated areas has a limitation with its accuracy owing to the assumption it makes that vegetation canopy is a homogeneous scatterer. The new MWCM that takes into account the vegetation fraction had not yet been assessed in contrast to the WCM anywhere in Uganda. This study therefore aimed to evaluate the performance of the MWCM and the WCM in estimation of soil moisture over a coffee-growing village. In situ data was collected in the topmost layer (5cm depth) and computed using the thermo-gravimetric method. Soil moisture was computed using parameters determined from Sentinel-1 and Sentinel-2 images. Root mean square errors (RMSE) of 3.7482 and 3.3346 and coefficients of determination (R2) of 0.6093 and 0.6175 were obtained for the WCM and MWCM respectively. This showed the MWCM had a better performance than the WCM in estimating soil moisture over the study area. The difference on the other hand was found not to be large because of the relatively high vegetation fraction at that time. Results can be improved by the use of L-band SAR data which has a higher penetrating ability than the C-band SAR data used in this research. Additionally, future works can look at the incorporation of surface roughness parameters and usage of the SAFY model to estimate vegetation characteristics to improve the accuracy of the results. The setting up of in situ soil moisture monitoring stations to provide larger datasets over longer periods for the Spatio-temporal evaluation of models is also suggested.