Modeling of satellite data to identify the seasonal patterns and trends of vegetation index in Kathmandu Valley, Nepal from 2000 to 2015

Abstract

Normalised difference vegetation index (NDVI) data were analysed to identify the seasonal patterns and the time series trends of vegetation in Kathmandu. The data were managed in three steps: reordering, removal of unreliable values and validating. A cubic spline function was used to examine annual seasonal patterns that revealed regular seasonal peaks (day 225 to 280) and troughs (day 50 to 81) of vegetation and start of greening from April and of browning from November. Linear regression models were fitted to seasonally adjusted NDVI, which statistically showed 40.70% of the grid cells had a significant increase and 24.71% of it had decreasing trends. To adjust for autocorrelation, generalized estimating equations (GEE) were fitted to the data for whole area that showed, the overall vegetation has been significantly declining at a rate of -0.005 ̊C and -0.006 ̊C per decade for 2000-2004 and 2010-2015 respectively. The recent period of decline is alarming for a growing city like Kathmandu.