Assessment of random forest method in pixel-based snow cover classification in Alpine region, Tatra mountains and Kaçkar mountains

Download

10512158-kutuphane.pdf

Date

2022-11-21

Author

Aksu, Cansu

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

281
views

167
downloads

For most countries in the Northern Hemisphere, the amount of usable water throughout the year is roughly determined by the amount of snow. Climate change and increasing demand on drinking and industrial water due to population growth make the monitoring of snow cover even more crucial than it was in the past. Today, to observe the amount of snow cover, different algorithms are being used on remote sensing data for classification of snow, aside from in-situ data collection techniques. This study presents the evaluation of the performance of Random Forest (RF) algorithm for snow cover classification on Sentinel-2 imagery over three selected mountainous regions: Alpine Region, Tatra Mountains, and Kaçkar Mountains, with different input combinations as independent variables (i.e., predictors). The combinations have been evaluated for three different times of the year for a much better assessment – to observe the differences in when snow cover starts to form (November to December), the time with roughly the maximum amount of snow is observed (January to March), and the time when it starts to melt (April to June). The confusion matrices, overall accuracy (OA) and Kappa coefficient were used for accuracy assessment. Overall, principle component bands combination (Pca) yielded the most accurate results. Pca combination also provided the shortest computation time out of all combinations, excluding the process of obtaining principal components, as the combination has the least amounts of input to the RF model, as compared to the other combinations. The overall results revealed that RF algorithm works well with appropriate numbers of principal component bands with NDSI, NDVI and NDWI indices for complex terrains over mountainous areas.

Subject Keywords

Sentinel-2, Remote sensing of snow, Classification, Machine learning, Snow hydrology, Random forest

URI

https://hdl.handle.net/11511/101226

Collections

Graduate School of Natural and Applied Sciences, Thesis

Suggestions

OpenMETU
Core

Special issue on remote sensing of snow and its applications Arslan, Ali Nadir; Akyürek, Sevda Zuhal (2019-06-01) Snow cover is an essential climate variable directly affecting the Earth's energy balance. Snow cover has a number of important physical properties that exert an influence on global and regional energy, water, and carbon cycles. Remote sensing provides a good understanding of snow cover and enable snow cover information to be assimilated into hydrological, land surface, meteorological, and climate models for predicting snowmelt runoff, snow water resources, and to warn about snow-related natural hazards. Th...
Monitoring the snow covered Areas from NOAA AVHRR images in the Eastern Part of Turkey Akyürek, Sevda Zuhal (2002-04-01) Monitoring snow-covered areas and estimating the snow water equivalent play an important role in predicting discharges during spring months, especially in regions where snow is an important resource.^This study has been conducted in the Upper Euphrates River basin, of 10 200 knr area, and elevation range of 1125— 3500 m. In estimating snow-covered areas, besides semi-supervised multispectral classification of NOAA-AVHRR data, a theta algorithm, developed by the US National Weather Service, has been used. Th...
Analysis of future climate change impacts on snow distribution over mountainous watersheds in Northern California by means of a physically-based snow distribution model Ishida, K.; Ercan, Ali; Trinh, T.; Kavvas, M. L.; Ohara, N.; Carr, K.; Anderson, M. L. (2018-12-01) The impacts of climate change on snow distribution through the 21st century were investigated over three mountainous watersheds in Northern California by means of a physically-based snow distribution model. The future climate conditions during a 90-year future period from water year 2010 to 2100 were obtained from 13 future climate projection realizations from two GCMs (ECHAM5 and CCSM3) based on four SRES scenarios (A1B, A1FI, A2, and B1). The 13 future climate projection realizations were dynamically down...
A machine learning-based accuracy enhancement on EUMETSAT H-SAF H35 effective snow-covered area product Kuter, Semih; Bolat, Kenan; Akyürek, Sevda Zuhal (2022-04-01) Snow is a major element of the cryosphere with significant impact on the Earth's water cycle and global energy budget. Acquiring consistent and long time series data on the spatial extent of snow cover doubtlessly plays a key role in our understanding and modeling of the current and future environmental dynamics. Remote sensing offers a powerful tool for continuous retrieval of snow cover information by utilizing snow's contrasting reflectance characteristics at optical wavelengths. The pre-operational H35 ...
Estimation of seasonal runoff using remote sensing satellite data Sorman, AU; Saydam, C (1997-05-03) Snow hydrology has been a relatively less investigated part of the hydrologic cycle. In the last two decades there has been a growing interest and scientific research to study the role of melting snow fbr the estimation of runoff mainly contributing to the storage in large reservoirs and to make the procedure operational on a real time basis. The paper describes a project related to satellite snow cover mapping and snowmelt runoff computations, which is currently carried out in eastern Turkey.

Citation Formats

C. Aksu, “Assessment of random forest method in pixel-based snow cover classification in Alpine region, Tatra mountains and Kaçkar mountains,” M.S. - Master of Science, Middle East Technical University, 2022.