Water equivalent of snow retrieved from data of passive microwave scanning with the use of artificial neural networks over the Russian Federation territory
| Autor: | D. O. Petrov, D. A. Kostyuk, A. A. Volchek |
|---|---|
| Rok vydání: | 2016 |
| Předmět: |
Global and Planetary Change
microwave remote sensing Artificial neural network Meteorology snow storage Science Vegetation Snow Water equivalent Set (abstract data type) Geography Geochemistry and Petrology snow water equivalent Russian federation artificial neural network Microwave Realization (probability) Earth-Surface Processes Water Science and Technology Remote sensing |
| Zdroj: | Lëd i Sneg, Vol 56, Iss 1, Pp 43-51 (2016) |
| ISSN: | 2412-3765 2076-6734 |
| Popis: | Using of the Chang model for calculation of the snow water equivalent on the basis of measurements of the Earth thermo-microwave radiation by means of scanning polarimeters (SMMR, SSM/I, AMSR-E) from board of orbital satellites does not allow obtaining the accuracy needed hydrological purposes. Low accuracy of the calculations is caused by both simplified character of the mathematical model, and due to significant influence of the surface characteristics (relief, vegetation and complex structure of snow thickness) upon the microwave radiation propagation. This work was aimed at finding a way to increase accuracy of calculations of the snow water equivalent on the Russian Federation territory with its different climate conditions by means of application the neural network approach for processing of results of the passive microwave scanning of the Earth surface. Feed-forward multi-layer artificial neural network was trained by back-propagation algorithm using SSM/I data and results of snow water equivalent in situ measurements obtained at 117 meteorological stations during the period from January 1st, 1988 till December 31st, 1988. Validation was performed using data from the same sources collected during 7 years (1992–1998). Results of performed numerical experiments and obtained values of rootmean-square error (σ = 24.9 мм; r = 0.39±0,01) allow coming to conclusion that the best estimation of water equivalent of a snow cover is provided by artificial neural network using as the input data a set of the SSM/I channels 19.35, 37.0, 85.5 GHz of horizontal and vertical polarizations with meteorological data differentiated by types of the snow survey route.It is shown that low correlation coefficients ( |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|