Preliminary Analysis of Snow Water Equivalent Estimation Over La Grande River Watershed Using SSMI data: Comparison of two Neural Network Approach.
Abstract
A Steady flow of information on snow cover extent and water equivalent is crucial for hydrologic forecasting, particularly in region where a large percentage of total precipitation falls as snow. However, because of inaccessibility and the large extent of northern areas, snow surveys are expensive, even more when accurate estimation of the spatial distribution of snow cover variables are required. Combining snow surveys with remote sensing data offers an alternative to estimate snow water equivalent (SWE). Furthermore, using passive microwaves is advantageous for snow mapping, since the microwaves are relatively independent of atmospheric constraints and solar illumination. The main goal of this study is to estimate the snow water equivalent (SWE) over La Grande River watershed in a taiga area. (Northern Quebec, Canada) using SSMI data. More specifically, we have tested the performance of two Artificial Neural Network (ANN) model in SWE estimation: the backpropagation neural network with variable learning rate and the counterpropagation fuzzy neural network. For this purpose the input data include the seven channels of SSM/I sensor in descending mode and the minimum air temperature, while the target examples consisted of SWE measurements conducted by Hydro-Quebec, Alcan and the Churchill Falls and Labrador company (CflCo) from January to March during the period between 1993 and 2002. The preliminary results of training process show that for each ANN model the error is less when the training is made on monthly basis comparatively to an annual basis.
- Publication:
-
AGU Spring Meeting Abstracts
- Pub Date:
- May 2004
- Bibcode:
- 2004AGUSM.H23A..08D
- Keywords:
-
- 1863 Snow and ice (1827)