ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., IV-4/W4, 179-187, 2017
https://doi.org/10.5194/isprs-annals-IV-4-W4-179-2017
© Author(s) 2017. This work is distributed under
the Creative Commons Attribution 4.0 License.
 
13 Nov 2017
FRACTIONAL SNOW COVER MAPPING BY ARTIFICIAL NEURAL NETWORKS AND SUPPORT VECTOR MACHINES
B. B. Çiftçi1, S. Kuter2,4, Z. Akyürek3,5, and G.-W. Weber4,6 1Cankiri Karatekin University, Graduate School of Natural and Applied Sciences, Department of Forest Engineering, 18200, Cankiri, Turkey
2Çankırı Karatekin University, Faculty of Forestry, Department of Forest Engineering, 18200, Çankırı, Turkey
3Middle East Technical University, Faculty of Engineering, Department of Civil Engineering, 06800, Ankara, Turkey
4Middle East Technical University, Institute of Applied Mathematics, 06800, Ankara, Turkey
5Middle East Technical University, Graduate School of Natural and Applied Sciences, Department of Geodetic and Geographic Information Technologies, 06800, Ankara, Turkey
6Poznan University of Technology, Faculty of Engineering Management, Department of Marketing and Economic Engineering, 60-965, Poznan, Poland
Keywords: Fractional Snow Cover Mapping, MODIS, Landsat ETM+, Artificial Neural Networks, Multilayer Perceptron, Support Vector Machines, Support Vector Regression Abstract. Snow is an important land cover whose distribution over space and time plays a significant role in various environmental processes. Hence, snow cover mapping with high accuracy is necessary to have a real understanding for present and future climate, water cycle, and ecological changes. This study aims to investigate and compare the design and use of artificial neural networks (ANNs) and support vector machines (SVMs) algorithms for fractional snow cover (FSC) mapping from satellite data. ANN and SVM models with different model building settings are trained by using Moderate Resolution Imaging Spectroradiometer surface reflectance values of bands 1–7, normalized difference snow index and normalized difference vegetation index as predictor variables. Reference FSC maps are generated from higher spatial resolution Landsat ETM+ binary snow cover maps. Results on the independent test data set indicate that the developed ANN model with hyperbolic tangent transfer function in the output layer and the SVM model with radial basis function kernel produce high FSC mapping accuracies with the corresponding values of R = 0.93 and R = 0.92, respectively.
Conference paper (PDF, 1625 KB)


Citation: Çiftçi, B. B., Kuter, S., Akyürek, Z., and Weber, G.-W.: FRACTIONAL SNOW COVER MAPPING BY ARTIFICIAL NEURAL NETWORKS AND SUPPORT VECTOR MACHINES, ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci., IV-4/W4, 179-187, https://doi.org/10.5194/isprs-annals-IV-4-W4-179-2017, 2017.

BibTeX EndNote Reference Manager XML