There has been a recent proliferation of remote sensing-based trend analysis for monitoring regional desertification. These show contradictory results. All of them claim to have been “validated” through expert interpretation, in the absence of sufficient field data. The authors suggest that such an approach is not sufficiently rigorous. Therefore, they demonstrate an approach which simulates land degradation so that the intensity, rate and timing of the reduction in NDVI can be controlled, in order to quantitatively evaluate the ability of methods to detect these known changes. The results show that linear trend analysis is rather insensitive to previously observed levels of NDVI reduction due to degradation in the well-studied communal lands in the Lowveld of South Africa. The period of investigation, has a large but rather unpredictable influence on the linear trends. This casts doubts over the ability of linear trend analysis, to detect relatively subtle, slowly-developing degradation in semiarid rangelands.
Reference:
Wessels, KJ, Van den Bergh, F, Scholes, RJ. 2011. Can linear trend analyses of NDVI time series data truly detect land degradation? Simulations may provide the answer. 34th International Symposium for Remote Sensing of Environment. Sydney Convention & Exhibition Centre Sydney, Australia, 10-15 April 2011
Wessels, K. J., Van den Bergh, F., Scholes, R., & Miteffa, S. (2011). Can linear trend analyses of NDVI time series data truly detect land degradation? Simulations may provide the answer. http://hdl.handle.net/10204/5170
Wessels, Konrad J, F Van den Bergh, RJ Scholes, and S Miteffa. "Can linear trend analyses of NDVI time series data truly detect land degradation? Simulations may provide the answer." (2011): http://hdl.handle.net/10204/5170
Wessels KJ, Van den Bergh F, Scholes R, Miteffa S, Can linear trend analyses of NDVI time series data truly detect land degradation? Simulations may provide the answer; 2011. http://hdl.handle.net/10204/5170 .