HR: 0800h
AN: H21D-1378    [Abstracts]
TI: Analysis and improvement of estimated snow water equivalent (SWE) using Artificial Neural Networks
AU: * E Azar, A
EM: aeazar@ce.ccny.cuny.edu
AF: NOAA-CREST, City University of New York, New York, NY 10031
AU: Ghedira, H
EM: ghedira@ce.ccny.cuny.edu
AF: NOAA-CREST, City University of New York, New York, NY 10031
AU: Khanbilvardi, R
EM: rk@ce.ccny.cuny.edu
AF: NOAA-CREST, City University of New York, New York, NY 10031
AB: The goal of this study is to improve the retrieval of SWE/Snow depth in Great lakes area, United States using passive microwave images along with Normalized Difference Vegetation Index NDVI and Artificial Neural Networks (ANNs). Passive microwave images have been successfully used to estimate snow characteristics such as Snow Water Equivalent (SWE) and snow depth. Despite considerable progress, challenges still exist with respect to accuracy and reliability. In this study, Special Sensor Microwave Imager (SSM/I) channels which are available in Equal-Area Scalable Earth Grid (EASE-GRID) format are used. The study area is covered by a 28 by 35 grid of EASE-Grid pixels, 25km by 25km each. To have a comprehensive data set of brightness temperatures (Tb) of SSM/I channels, an assortment of pixels were selected based on latitude and land cover. A time series analysis was conducted for three winter seasons to assess the SSM/I capability to estimates snow depth and SWE for various land covers. Ground truth data' were obtained from the National Climate Data Center (NCDC) and the National Operational Hydrological Remote Sensing Center (NOHRSC). The NCDC provided daily snow depth measurements reported from various stations located in the study area. Measurements were recorded and projected to match EASE-GRID formatting. The NOHRSC produces SNODAS dataset using airborne Gamma radiation and gauge measurements combined with a physical model. The data set consisted of different snow characteristics such as SWE and snow depth. Landcover characteristics are introduced by using Normalized Difference Vegetation Index (NDVI). An Artificial Neural Network (ANN) algorithm has been employed to evaluate the effect of landcover in estimating snow depth and Snow Water Equivalent (SWE). The model is trained using SSM/I channels (19v, 19h, 37v, 37h, 22v, 85v, 85h) and the mean and standard deviation of NDVI for the each pixel. The preliminary time series results showed various degrees of correlations between snow depths and SWE with combinations of different channels. However a consistent relationship is observed for many channels but the highest sensitivity is observed for GTV (37v-19v) for a three year period, showing correlation of 70 percent for some pixels. Also, introducing the landcover as an input to the ANN, increases the accuracy of SWE and snow depth estimations. This is in agreement with the literature though it indicates necessity of further research to establish an algorithm to estimate SWE/Snow depth for our study area.
DE: 1836 Hydrological cycles and budgets (1218, 1655)
DE: 1854 Precipitation (3354)
DE: 1860 Streamflow
DE: 1894 Instruments and techniques: modeling
DE: 9820 Techniques applicable in three or more fields
SC: Hydrology [H]
MN: Fall Meeting 2005