Monitoring Spatio-temporal Variations of Vegetation Responses to Drought with Remotely Sensed Data

Studies on monitoring vegetative responses to drought by integrating remote sensing and meteorological data sets are of great practical meaning in mitigating losses, making decisions and managing agriculture.;A field experiment was conducted to test the hyperspectral sensitivity of corn responses to drought in controlled gradients. Statistical analysis shows that leaf reflectance at 510 nm and 690 nm under artificial illumination mode can indicate early drought significantly. The sensitivity of different hyperspectral indices and position-based variables on vegetative chlorophyll change under drought was also comprehensively compared.;China was divided into 7 regions based on the monthly anomaly of NDVI, precipitation, temperature and sunshine duration from 1982 to 2006 with hierarchical clustering. The impact of precipitation, temperature and sunshine duration on crop NDVI in spring and summer (average R2 >=0.35) is higher than that in autumn and winter (average R2 < =0.14). The full-year crop NDVI trend for region 1∼3 increases (p<0.1), which is related with temperature increase as temperature is the limiting climatic factor for crop growth in these regions.;Time-frequency domain analysis was conducted to denoise multi-temporal remotely sensed data. These methods can be used for retrieving cloud-free phenological curves from multi-temporal satellite data. The two drought indices TVDI and VHI were spatio-temporally compared with the meteorological drought index, SPI. TVDI was subjected to extracted ‘wet edge’ and ‘dry edge’ in the study area while VHI was subjected to noises on NDVI and LST time series. SPI3∼6 and remotely sensed drought indices can reach relative stable relationship, which may be due to the fact that most areas in the present study were covered by forest.;Carbon dynamics for forest ecosystem in drought can be quantified with carbon parameters from models and multi-temporal MODIS NDVI data sets. Sub-pixel percent forest cover was extracted based on principle component analysis (PCA) to multi-temporal MODIS NDVI. The estimated carbon change is highly correlated with the remotely sensed PsnNet change (r=0.32 p <0.001). Satellite remote sensing in conjunction with ecological models can retrieve many important variables caused by global change.