Variability of satellite derived phenological parameters across maize producing areas of South Africa

22 Oct 2018

Changes in phenology can be used as a proxy to elucidate the short and long term trends in climate change and variability. Such phenological changes are driven by weather and climate as well as environmental and ecological factors. Climate change affects plant phenology largely during the vegetative and reproductive stages. The focus of this study was to investigate the changes in phenological parameters of maize as well as to assess their causal factors across the selected maize-producing Provinces (viz: North West, Free State, Mpumalanga and KwaZulu-Natal) of South Africa. For this purpose, five phenological parameters i.e., the length of season (LOS), start of season (SOS), end of season (EOS), position of peak value (POP), and position of trough value (POT) derived from the MODIS NDVI data (MOD13Q1) were analysed. In addition, climatic variables (Potential Evapotranspiration (PET), Precipitation (PRE), Maximum (TMX) and Minimum (TMN) Temperatures spanning from 2000 to 2015 were also analysed. Based on the results, the maize-producing Provinces considered exhibit a decreasing trend in NDVI values. The results further show that Mpumalanga and Free State Provinces have SOS and EOS in December and April respectively. In terms of the LOS, KwaZulu-Natal Province had the highest days (194), followed by Mpumalanga with 177 days, while NorthWest and Free State Provinces had 149 and 148 days, respectively. Our results further demonstrate that the influences of climate variables on phenological parameters exhibit a strong space-time and common covariate dependence. For instance, TMN dominated in North West and Free State, PET and TMX are the main dominant factors in KwaZulu-Natal Province whereas PRE highly dominated in Mpumalanga. Furthermore, the result of the Partial Least Square Path Modeling (PLS-PM) analysis indicates that climatic variables predict about 46% of the variability of phenology indicators and about 63% of the variability of yield indicators for the entire study area. The goodness of fit index indicates that the model has a prediction power of 75% over the entire study area. This study contributes towards enhancing the knowledge of the dynamics in the phenological parameters and the results can assist farmers to make the necessary adjustment in order to have an optimal production and thereby enhance food security for both human and livestock.