Prediction of vegetation anomalies over an inland river basin in north-western China

Vegetation cover plays an important role in linking the atmosphere, water, and land and is deemed as a key indicator in the terrestrial ecological system. Therefore, it is of great importance to monitor vegetation dynamics and understand the mechanisms of vegetation change, including that driven by climate change. This study examines (a) the evolution of vegetation dynamics over the Heihe River Basin in the typical arid zone in north-western China using nonparametric Mann-Kendall test and Thiel Sen's slope; (b) the relationships between remotely sensed vegetation indices (normalized difference vegetation index [NDVI] and enhanced vegetation index [EVI]) and hydroclimatic variables based on correlation analysis; and (c) the prediction of vegetation anomalies using a multiple linear regression model. For the analysis, the Moderate Resolution Imaging Spectroradiometer NDVI/EVI product and the gridded daily meteorological data at a spatial resolution of 0.125 degrees over the period 2001-2010 are considered. The results indicate that vegetation cover improved over a large proportion during 2001-2010, with a significant trend towards warm and wet, characterized by an increase in average annual temperature and precipitation by 0.042 degrees C/year and 5.8mm/year, respectively. We test the feasibility of NDVI and EVI in quantifying the responses of vegetation anomaly to climate change and develop a statistical model to predict vegetation dynamics in the basin. The NDVI-based model is found to be more reliable than the EVI-based model, partly due to the vegetation characteristics and geomorphologic properties of the study region. The proposed model performs well when there is no lag time between meteorological factors and vegetation indices for grassland and cropland, whereas 1-month lead time prediction is found to be best for forest. The soil water content is introduced as an extra explanatory variable, which effectively improves the prediction accuracy for different land use types. In general, the predictive ability of the proposed model is stable and satisfactory, and the model can provide useful early warning information for regional water resources management under changing climate.