Authors: Narcisa G Pricope*, University of North Carolina Wilmington, Andrea Gaughan, University of Louisville, Kyle Woodward, University of North Carolina Wilmington, Forrest Stevens, University of Louisville, Michael Drake, University of Colorado Boulder
Topics: Human-Environment Geography, Remote Sensing, Africa
Keywords: remote sensing, participatory mapping, Random Forest, NTFP, grazing
Session Type: Virtual Paper
Start / End Time: 1:30 PM / 2:45 PM
Room: Virtual 44
Presentation File: No File Uploaded
Recently there has been an increased focus on the integration of socio-ecological research with remote sensing and land systems modeling to understand human-environment interactions. The roles of non-timber forest products (NTFP) and livestock ownership in natural resource-dependent livelihoods is a growing topic in socio-ecological research. However, few of these studies incorporate remote sensing and spatial modeling into analyses of household-level natural resource use. Relying on field mapped resource areas and household-level resource use data from previous participatory mapping research, we compiled Landsat-derived vegetation proxies and geospatial datasets from WorldPop to spatially model grazing and NTFP collection activities within a transboundary southern African landscape encompassing three community-based organizations (CBO). We constructed Random Forest classifiers to assign pixel-level predictions of each resource use activity of interest, then assessed each model’s performance and relative covariate importance. Then for each resource type, we used the model outputs to compare total predicted land area under use to the total used land area according to the household-level resource use data. Finally, we investigate the similarities and differences in land use patterns between each CBO and identify the lessons learned in modeling land use in this and similar landscapes. Our mixed methods approach could extend analyses of subsistence-level land use activities beyond the areas covered on the ground by researchers. Remotely sensed data can provide useful context in determining where people are most likely to use particular natural resources, which could aid in future land management decision-making.