Geographically Weighted Structural Equation Models: understanding the spatial variation of latent variables and drivers of environmental restoration effectiveness

This paper describes a methodological extension to Geographically Weighted (GW) models. It develops and applies a GW structural equation model (SEM) to understand the observed and latent drivers associated with effective landscape restoration in Northern China. The paper reviews
recent landscape restoration activities in China, and describes the environmental context of these: soil loss, erosion and land degradation. Restoration effectiveness was described by changes in net primary production and fractional vegetation cover, as recorded in MODIS data for
the period 2000-2012. County level census data provided information on hypothesised latent variables of population pressure, off-farm economy and rural economy. The GW SEM analysis allows the spatial variation in the contributions made by different socio-economic factors to
restoration effectiveness to be evaluated. Although developed with ropey data – the County level population totals were perhaps unrealistically interpolated over a 2km grid and the MODIS data were aggregated over the same – the GW SEM allows the detail of the how what and where to be identified thus supporting local policy and planning. A number of future developments in refining this method are outlined.