Assessing Field - Level Sorghum Yield Variability in South Sudan Using Remote Sensing

Abstract

Linear mixed-effects models (LME) include both fixed-effects and random-effects variables. The use of LME is powerful in analyzing repeated measurements, longitudinal data or unbalanced data and the variations between and within-subject observations can be captured by random effects. In this study, remote sensing data were used to understand sorghum yield variability in a context of low input low output extensive farming system such in South Sudan. LME modelling approach helped understanding the sorghum yield variation between the two states of interest in this study and between two different agricultural seasons. The unbalanced nature, the repeated measures on same statistical units for remotely derived parameters and the longitudinal nature of the data dictated the choice and the appropriateness of Linear Mixed-Effects models (LME) for statistical analysis in this study. The random-effects structures were used to describe the spatio (between states) and temporal (between seasons) specific variations of the sorghum yield during the two agricultural seasons (2018-2019); while the size of cultivated land and the households’ size as a proxy of labor were used as fixed-effects variables in addition to remotely derived variables.