Glacier volume and mass balance in the Ala Archa National Park (Kyrgyz Ala-Too/Kyrgyzstan)

Diplomarbeit von Angela Ender

Introduction:

Glaciers – fascinating elements of nature with the characteristic to be an important source of water and – at the same time – cause of serious natural hazards like outbursts of lakes, which were built potentially during glacier retreats. Changes in glaciers are recognized as important climate indicators since they react strongly on climate changes. They reached their last maximum extent at the end of the Little Ice Age around 1850. From this time onwards, most glaciers all around the world began shrinking. This effect increased during the last 30 years. The ongoing is mainly caused by the increasing global mean temperature and changes in precipitation partitioning (at high elevations more rain than snow falls). As a consequence the situation ‘ [.] may lead to the deglaciation of large parts of mountain ranges by the end of the 21st century‘.

Until the beginning of aerial photography and satellite operations these changes were only sparsely documented. But now the new techniques allow a remotely sensed glacier monitoring, which enables better access especially for distant high mountains like in the Tien Shan Mountain range in Central Asia. Therefore, several regional research studies about climate changes and concomitant glacier recession in the Tien Shan exist. The Ala Archa National Park, which is the subject of this study, is located in the Kyrgyz Ala-Too as a part of the Tien Shan Mountains. The local investigations are in most cases Russian-speaking. According to the author’s information the first English-speaking study was published in 1996 by Aizen et al.

The main goal of this work is to investigate the total volume and mass balance of glaciers in the Ala Archa National Park as well as to gain knowledge about their area change since the second half of the 20th century.

Initial information of the study area and theoretical backgrounds on glaciers are given in section 2. The following sections 3 and 4 are dedicating to the used data and the generation of a time series of digital elevation models (DEM) as well as their evaluation. Methods for glacier mapping from satellite data are explained in section 5 with emphasis on debris cover. By fusion of glacier outlines with DEMs (and DEM-derived products) glacier parameters are obtained, enabling comparisons of glaciers at different points in time. Thereby, the volume change and glacier mass balance is explained in detail. With the help of approximated glacier bed topography, the absolute glacier volume should also be estimated. Furthermore, an ArcGIS-script for deriving glacier branch lines and values for the glacier length is generated and presented. This chapter presents the main task of this thesis. Finally, the cartographic visualization is shown in section 6. The work finishes with the main conclusions as well as an outlook for possible future works.

Table of Content:

	List of tables	iv
	List of figures	vi
	List of abbreviations	ix
1.	Introduction	1
2.	Background	3
2.1	Glaciers – basics	3
2.1.1	Formation and composition of glaciers	3
2.1.2	Movement of glaciers	4
2.1.3	Influence of climatic changes	6
2.1.4	Mass balance of a glacier	7
2.1.5	Estimation of glacier volume	8
2.2	Study area	9
2.2.1	Kyrgyzstan and Tien Shan – geographical overview	9
2.2.2	Ala Archa National Park	10
2.2.3	Climate	12
2.3	GIS technology	13
2.3.1	Data structure	13
2.3.2	Integration of data	13
3.	Data from remote-sensing sensors and other sources	15
3.1	Passive sensors	15
3.1.1	Corona (1960-1972)	16
3.1.2	Hexagon (1971-1986)	18
3.1.3	Landsat (Since 1972)	19
3.1.4	ASTER (Since 1999)	21
3.1.5	RapidEye (Since 2009)	23
3.2	Active Sensors – InSAR (2000)	24
3.3	Additional data sources	26
3.3.1	Topographic maps (1960s/1980s)	26
3.3.2	Aerial images (1971/1988)	27
3.3.3	GPS data (2009)	27
4.	Generation of Digital Elevation Models (DEM)	29
4.1	Basics	29
4.2	DEM from stereo image pairs	30
4.2.1	General procedure	30
4.2.2	Software	31
4.2.3	Master scene	33
4.2.4	Corona-DEM	33
4.2.5	DEM of ASTER	37
4.3	DEM of aerial photos	39
4.4	DTM of topographic maps	41
4.5	Evaluation	41
4.5.1	Reference model	41
4.5.2	Input models	45
5.	Glacier change since 1964	52
5.1	Methods of glacier mapping	52
5.1.1	Spectral properties of snow and ice	52
5.1.2	State of the art	54
5.1.3	Ratio images	55
5.1.4	Debris	55
5.1.5	Glacier masks	57
5.2	Automatic calculation of central branch lines	61
5.2.1	Basics	61
5.2.2	Algorithm	62
5.2.3	Results	66
5.2.4	Calculation the glacier length	67
5.3	Glacier parameters	69
5.3.1	Area	69
5.3.2	Altitude	73
5.3.3	Slope	74
5.3.4	Length	75
5.3.5	Mass balance	76
5.3.6	Absolute volume	81
5.4	Discussion	88
6.	Cartographic visualization	90
7.	Conclusion and outlook	101
	References	xii
	Appendix
A 1:	Tables	xxi
A 2:	Source code ‘flow_lines.py’	xxv

Text Sample:

Chapter 5, Glacier change since 1964:

In this chapter it is shown in detail, how course of time affects the glacier area and volume. This is performed by assigning certain topographic glacier parameters, which are compared to each other.

5.1, Methods of glacier mapping:

5.1.1, Spectral properties of snow and ice:

The spectral sensitivity of a sensor provides information about the reflectance properties of surface materials for different wavelengths. In literature, numerous methods for the identification of glaciers in multi spectral remote sensing data exist. There, the specific reflection properties are exploited, since they differ for clean ice, fresh snow or debris. Due to the fact that ice is basically metamorphosed snow, the spectral properties are quite equal for both. Fig. 5.1-1 shows the different reflectance of several glacier surfaces. Each one of them reaches its maximum in the range of visible light (0.4-0.7 µm) and much lower values at the middle-infrared (1.3-3.0 µm). In contrast, for other terrain without glaciers, the reflection in middle-infrared is higher than in visible light.

Fresh snow shows the highest values for reflection. It is therefore one of the brightest natural surfaces. Throughout the aging process, the reflectance is steadily decreasing. On the one hand, this is caused by the increasing grain size during crystallization and on the other hand by increasing pollution by sediments. The dependence of the reflectance to the pollution is shown in fig. 5.1-1 and fig. 5.1-2. Also, the reflectance generally decreased with increasing wavelength in the NIR-range (0.7-1.5 µm).

5.1.2, State of the art:

Basically, the best method for glacier mapping is in-situ measurement. However, since glaciers in high mountain regions are in most cases difficult to access, this method is strongly complicated and time-consuming.

Therefore, to investigate large areas, methods of remote sensing are often used. Various algorithms for deriving glacier boundaries from multi-spectral imagery exist. An accurate method is the manual delineation based on aerial and satellite images. However, this method is strongly time-consuming and depends also on the skill of the surveyor. As an alternative, the often used ratio images are a simple but effective method. Combining these ratio images with Normalized Difference Indices for snow or vegetation, the results can be further improved. Also, by using supervised classifications, good results can be obtained for remote sensing data. compares the different methods and contrasts their efficiency and expense.

However, general difficulties can be found while identifying glacier surfaces with supraglacial debris, since they own similar spectral properties compared to the glacier‘s surrounding. A method based on the slope was published by. More recent researches use the integration of geomorphometric parameters.

In this work, the glaciers were identified by a combination of ratio images and manual post-processing.

5.1.3, Ratio images:

Using threshold values, glacial areas are identified by the division of bands of the near and short wave infrared. In fig. 5.1-4 a glacial map from the ration of TM4/TM5 is presented. In TM4 band, ice is reflected as light gray, while snow is shown white. In TM5, ice becomes black and snow dark gray. In appropriate illuminated regions, good results are obtained for less snow coverage.

Shadowed areas own low reflectance and present therefore some problems. These areas are shown black in both channels. Fortunately, most images show good illuminations. The threshold value of 2 used by Paul is also used here without any further adjustment.

5.1.4, Debris:

Debris is a typical feature at the glacier‘s surface in the ablation area. The largest errors in the glacier masks generation process will arise in these areas. The band ratio detects no glacier under thick debris cover or moraines. However, manual post-processing allows a removal of these mistakes.

By intersecting the pre-defined glacier outlines with the final adjusted ones as it is shown in fig. 5.1-3, the class debris can be obtained (if there are neither clouds nor data gaps).

5.1.5, Glacier masks:

Six glacier masks were created, covering different decades. Four of them are used for further evaluation. Both remaining were rejected. These are the Landsat scene, since it contains too much snow cover, and the topographic map, due to wrong glacier classification. Other available data were finally not used due to bad spatial resolution or too strong snow coverage.

While the delineation of glaciers is no problem for ASTER and RapidEye scenes, certain difficulties exist for Corona. Caused by the equipment, its images have very strong distortions, which make the glacier delineation more difficult. However, with further data (e.g. Hexagon, map or models of volume loss and help expert knowledge, glacier masks could be matched. Using ortho images, DEMs and DEM-derived products, glaciers could be distinguished from other areas. However, some uncertainties still remain as visualized in fig. 5.1-5. On the right-hand-side, the difference model between Corona and SRTM is shown. There, the extents of glaciers seem to be even larger (arrow) compared to the satellite image. Also, proglacial lakes give a hint for finding glacier termini, since they are always located directly next to the glacier. An example for the classification of Top-Karagay glacier is visualized in more detail in chapter 6.