Estimation of Maximum Daily Fresh Snow Accumulation Using an Artificial Neural Network Model. (8th January 2019)

• Record Type: Journal Article
• Authors: Lee, Gun; Kim, Dongkyun; Kwon, Hyun-Han; Choi, Eunsoo
• Other Names: Porcù Federico Academic Editor.
• Is Part Of: Advances in meteorology. Volume 2019(2019)
• Publication Date: 2019-01-08
• Subjects: Meteorology -- Periodicals; Meteorology; Periodicals; 551.505
• DOI: 10.1155/2019/2709351
• Languages: English
• ISSNs: 1687-9309

Abstract

Abstract : For estimation of maximum daily fresh snow accumulation (MDFSA), a novel model based on an artificial neural network (ANN) was proposed. Daily precipitation, mean temperature, and minimum temperature were used as the input data for the ANN model. The ANN model was regularized and trained using a set of 19, 923 data points, observed daily in South Korea between 1960 and 2016. Leave-one-out cross validation was performed to validate the model. When the input data were known at the gauged locations, the correlation coefficient between the observed MDFSA and the estimated one by the ANN model was 0.90. When the input data were spatially interpolated at ungauged locations using the ordinary kriging (OK) method, the correlation coefficient was 0.40. The difference in correlation coefficients between the two methods implies that, while the ANN model itself has good performance, a significant portion of the uncertainty of the estimated MDFSA at ungauged locations comes from high spatial variability of the input variables that cannot be captured by the network of in situ gauges. However, these correlation coefficients were significantly greater than the correlation coefficient obtained by spatially interpolating the MDFSA values with the OK method ( R = 0.20). These findings suggest that our ANN model significantly reduces the uncertainty of the estimated MDFSA caused by its high spatial variability.