Detection of Snowmelt Areas using Synthetic Aperture Radar (SAR) Data in High Mountainous Central Asia

Kurzfassung

In the High Mountainous Central Asia, ecosystems are characterized by unique and diverse flora and fauna. These provide the population with clean water, pasture and forest products, and protect them against floods. Global warming, as well as no strategic agricultural use, is causing damage to fragile ecosystems and may reduce their ability to regenerate. However, rural populations are directly dependent on natural resources and must adapt to the negative impacts of climate change while, if possible, taking steps to mitigate those impacts. In mountain regions, one of the main indicators of global warming is an earlier snowmelt initiation and a shortened snow cover. However, snow and glacier melt play a central role for the water inflow of rivers and thus for the water supply of the population. To obtain a large-scale monitoring of snowmelt, remote sensing is a good way to detect a change in snow properties. Due to the ability to penetrate clouds and hence to deliver information during cloudy seasons, radar observation from the European Space Agency (ESA) satellite Sentinel-1 data with a VH polarization in a C-band frequency is used. In this thesis, snowmelt in the Naryn river Basin in Kyrgyzstan and its surrounding mountains is researched. Therefore, the wet snow detection approach by Nagler - invented in 2000 - is applied, which classifies wet snow using Synthetic Aperture Radar (SAR) data. The approach is based on amplitude comparison between a backscatter coefficient of a geocoded reference image in dry mid-summer and an image during snowmelt. The study site is located between 40.8 to 42.4◦N and 74.4 to 78.5◦E, and is divided into four equally sized areas for processing. Due to geographical and climatological differences between the test area of this thesis and the Alps, where Nagler (2000) conducted his experiments for the algorithm in the procedure, the period of the reference image as well as the threshold value for classification have to be adjusted. For this purpose, snowmelt data from April to August in the years 2017 to 2020 are utilized. The reference image is set to the first week of February 2018; the consistently coldest period without any snowmelt. The adjusted algorithm is then implemented in a MATLAB program. This will allow the processing of current data from 2021 onwards and its evaluation. The output of the MATLAB program is a geocoded binary map classifying wet snow and non-wet snow; the Thomas Nagler algorithm and the use of SAR imagery cannot determine whether the remaining area is dry snow or ground surface. Finally, in the evaluation, snowmelt initiation times of each year can be derived, and the melting duration at different altitudes is analyzed. The results of the MATLAB evaluation and analysis are verified by snow maps by the Global SnowPack, DLR as well as hydrological data.