Exploring the Landscape Scale Influences of Tree Cover on Crop Yield in an Agroforestry Parkland Using Satellite Remote Sensing and Spatial Statistics

Abstract

Agroforestry in so called parklands, characterized by coexsitance of trees and crops, is the dominating agriculture practice in the Sudano-Sahelian zone of West Africa, and this landuse system is therefore the backbone of food security in the region. Trees in parklands influence crops both through competitive and facilitative mechanism, but the effects on crop production  are challenging to disentangle due to the complexity of the system resulting from high variability in tree cover structure (tree sizes and densities), tree species diversity and different crop combinations. Previous field-based research on this topic has provided key knowledge about the mechanisms driving the influence, but has not been able to clarify how the influence of individual trees scale to the landscape. Focusing on an agroforestry parkland landscape in central Burkina (Saponé) dominated by Vitellaria paradoxa and Parkia biglobosa and crop combinations of millet, sorghum and cowpea, this paper examines how tree cover structure influences crop yield at landscape scale using satellite remote sensing, field data and spatial statistical analysis. The following questions guided this study: i) to what extent does tree cover structure inside agricultural fields and in the surrounding landscape explain crop yield variability?, ii) what is the direction and magnitude of the identified relationships?, and iii) how important are soil properties for explaining crop yield variability? The overall aim is to clarify what tree cover structure maximizes crop yield production in this type of agroforestry parkland. Our analysis is based on data from two years (2017 and 2018) with clear differences in rainfall to assess the stability in identified relationships. Our findings showed that tree canopy cover and tree density inside the field tended to decrease crop yield because of the competition, but also that these variables when considering the surrounding landscape exerted an opposite effect because of their buffering effects. The explanatory variables representing soil properties did have limited effects on crop yield in this study. These patterns were consistent during the two years of monitoring. Overall, our results suggest that farmers in this area might manage the tree cover in a way that optimizes sustainable yields as canopy cover and tree density in most parklands is below the limits identified here where competition outweight the facilitative effects.

Keywords: Agroforestry, remote sensing, tree-crop interactions, Landscape, smallholder agriculture, Vitellaria paradoxa