NDVI, rainfall, modelling, calibration, verification
A model is developed to understand the relationship between satellite-derived NDVI and rainfall data in a large tropical catchment. Two Fourier-based modeling techniques with a seasonal component, viz. a seasonal model (SM) and a linear perturbation model (LPM) are tested, and their performance in reproducing the observed NDVI was evaluated. The methodology makes use of 15 years of 10-day composite time series data of rainfall and NDVI, which is estimated from NOAA-AVHRR data, both of which constitute concurrent data from 1982-96. The models are applied to a large catchment system of the Rufiji basin in Tanzania, with a network of 26 stations rainfall record and Thiessen polygon-interpolated spatially averaged NDVI data. The application of the SM model in forecasting NDVI and the LPM in relating NDVI and Rainfall at the 26 stations in the basin has been tested using the Nash and Sutcliffe (1970) model efficiency criterion. The linear perturbation model performed better than the simple seasonal model. The average model efficiency at the 26 stations considered during calibration and verification, are 0.64 and 0.54 for the LPM, and 0.62 and 0.49 for the SM, respectively. The approach can be used to improve our understanding of vegetation-rainfall relationships as well soil-vegetation-atmospheric processes, thus contributing to enhance hydrologic modeling of tropical watersheds.
BYU ScholarsArchive Citation
"Modeling and Understanding the Relationship between Vegetation and Rainfall of a Tropical Watershed using Remote Sensing Data and GIS,"
Journal of Spatial Hydrology: Vol. 7:
2, Article 4.
Available at: https://scholarsarchive.byu.edu/josh/vol7/iss2/4