The potential of remote sensing technologies for monitoring the productivity of dry season irrigated rice

Abstract

Rice is an important staple crop and significantly contributes to the dietary needs of the global population. From the African context, Nigeria and most sub-Saharan countries rely on rice importation to meet the rising consumption demands, mainly due to low yield returns. As such, monitoring rice yield and yield indicators (e.g. Leaf Area Index, biomass) in order to understand patterns and trends of rice growth is fundamental to improve yield outcomes. In this PhD rice yield and yield indicators were monitored using a combination of proximal, airborne and satellite sensors, across a range of spatial scales with the overall aim of furthering our understanding of the dynamics of irrigated rice yields. The first main objective of the thesis investigated the relative merits of structural and multispectral information for estimating centimetre scale rice above ground biomass from very high spatial resolution drone imagery. The focus was the reproductive and ripening stages of rice growth due to the strong relationship between biomass and yield at these stages. Results indicated that crop structural information, derived from a consumer-grade RGB camera, was of greater importance for rice biomass estimation than multispectral information. The second object of the thesis explored the potential of a hybrid gaussian process regression (GPR) - radiative transfer model (RTM) to estimate the phenological dynamics of rice Leaf Area Index (LAI). Sentinel-2 spectral bands were simulated from field spectroscopy data and combined with extensive in situ field measurement to develop and test a hybrid LAI prediction model. Results were also compared with the satellite-derived Sentinel-2 LAI standard. The findings demonstrated the potential of the proposed hybrid model for predicting within-season dynamics of rice LAI. The third objective of the thesis determined the relative importance of the spatial and spectral resolution of Sentinel-2 for estimating rice yield across a range of spatial extents. This section investigated the suitability of Sentinel-2 for predicting within and between field yield variability across varying spatial extents. The results demonstrated that the spatial resolution of Sentinel-2 data was more important than the spectral resolution for predicting within field yields. Results also demonstrated the potential of Sentinel-2 data for estimating rice yields across smallholder rice farms.

Date of Award	1 Aug 2021
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Angela Harris (Supervisor) & Gareth Clay (Supervisor)