Streamflow data play a key role in water resources management; however these data are
not often available. One of the alternatives then is to use the rainfall-runoff models, but in most cases
the required inputs such as rainfall and evapotranspiration are not available to use these models.
Freely available remote sensing data, which represent features of the above input variables, can be
used to generate streamflow data as an alternative. This project uses daily Moderate Resolution
Imaging Spectrometer (MODIS) data to generate daily streamflow for the Thomson catchment in
Victoria in Australia through an Artificial Neural Network (ANN) model. Daily MODIS reflectance and
radiance data were first converted to Normalized Difference Vegetation Index (NDVI) and cloud top
temperature (CTT) respectively. Several ANN models with one hidden layer were then developed
using combinations of present day NDVI and CTT variables, and several daily lags of these variables.
Results showed that a seasonally stratified model with five inputs had given predictions comparable to
observed streamflow. Five inputs were present day NDVI and CTT, and three past days of CTT