Determination of main climate and ground factors controlling vegetation cover regrowth along oil and gas pipelines using multiple, spatial and geographically weighted regression procedures

Abstract

The goal of this research was to determine the primary climate and ground factors controlling Normalize Difference Vegetation Index (NDVI) of grasslands and croplands along the disturbed Right of Way of Baku–Tbilisi–Ceyhan and South Caucasus Pipelines for revegetation planning purposes. The climate factors considered in this research were annual precipitation (PRECIP), annual evapotranspiration (ET), land surface temperature (LST), annual minimum air temperature (Tmin), annual maximum air temperature (Tmax) and solar radiation (SOLRAD). The ground factors considered were elevation, aspect, groundwater, and topsoil depth. A standard multiple regression model detected and eliminated SOLRAD and (Tmax) as the independent variables causing the multi-collinearity in the regression models for grasslands and croplands. The spatial autocorrelation of errors in the regression models was detected using the semi-variogram modeling of regression residuals. This required the use of the global spatial regression model (GSRM) based on the maximum likelihood model of spatial errors. GSRM determined for grasslands that PRECIP, ET, and LST were the most significant factors controlling NDVI. In the case of croplands, PRECIP, ET, and Tmin were observed to be the strongest controlling factors, and LST was not significant. Geographically weighted regression (GWR) determined the ranges of variable regression coefficients predicting NDVI of grasslands and croplands on the local level along the corridor of pipelines. Although no strong correlation was observed between elevation and NDVI, the largest vegetation distribution was identified in the areas with an elevation higher than 150 m. The percentage of high NDVI spatial coverage within the aspect categories was decreasing in the southern directions of slope faces. The threshold of groundwater depth controlling NDVI of grasslands was 1–12 m in the areas lower than 100 m in elevation. Topsoil depth was determined as one of the main ground factors having direct control of NDVI of grassland and croplands. Precipitation was also identified as one of the main factors controlling formation of topsoil depth. The results of GSRM and GWR were validated using low- and high-resolution remotely sensed data.