Sensitivity of Vegetation in India to Climatic Factors at Multiple Timescales

Abstract

In order to predict and model the responses of vegetation to future climate change, it is essential to understand vegetation responses to climate. In this study I used kNDVI, temperature, vapor pressure deficit (VPD), and rainfall data to calculate the sensitivity of vegetation to climate in the Indian mainland. Here, vegetation sensitivity to a climate variable refers to the response of vegetation to a unit change in that variable. Sensitivity is estimated as the regression coefficients from a fitted multiple linear regression. I used STL decomposition to separate kNDVI time series into 3 timescales—trend, seasonal and remainder. The seasonal component is dominant and explains 80% of variation in the original time series. The remainder component explains about 15%. To examine sensitivity to climate at different timescales, each of the STL components (as well as the complete kNDVI time series) was regressed against standardized climate variables. The sensitivity data—for each climate variable and kNDVI component—was examined spatially, and also correlated against the other sensitivities to reveal any interesting relationships. The sensitivities of the seasonal-component and complete kNDVI models are strongly correlated. For all models/components, temperature-sensitivity is negatively correlated with both rain- and VPD-sensitivity, while rain-sensitivity is positively correlated with VPD-sensitivity. The sensitivities of vegetation to climate variables were then split by vegetation type—there were no significant differences in mean sensitivity. All vegetation types in India seem to respond similarly to climate. To check for relations between sensitivity and long-term or background climate, the vegetation sensitivity data were correlated with mean climate data and squared climate data. No clear relationships were identified. More study is needed to fully characterize vegetation dynamics in India.