Olga V. Toutoubalina
Remote sensing of environmental
degradation in the north:
Case Study of the Non-Ferrous Metal Industry in Noril'sk,
northern Siberia
Summary of PhD dissertation
Large scale industrial and military developments in some regions
of the Russian, Canadian and Scandinavian Arctic and sub-Arctic
have caused considerable environmental problems. Non-ferrous
metal processing is one of the greatest environmental threats
generated in the Arctic by civil industry because of the high
toxicity and durability of heavy metal pollution. In addition,
the high-latitude vegetation is made particularly vulnerable by
the extreme climatic conditions, including permafrost.
This case study concentrates on the mining and smelting complex of Noril'sk city, the largest single industrial source of sulphur dioxide emissions in the world (over 2,000,000 tonnes annually) and also a large heavy metal polluter. As a result of the atmospheric emissions, more than 6,000 km2 of forest have been destroyed south-east of the city, following the direction of the prevailing winds. Mapping and monitoring of such large areas in conditions of extreme climate, difficult terrain, and absence of roads is greatly facilitated by satellite remote sensing, which allows one to obtain inexpensive and regular environmental information on a regional scale.
I have reviewed the environmental research accomplished to date in the Noril'sk region. My analysis has identified the lack of spatial detail in characterisation of the damage to natural ecosystems. I have employed American and Russian satellite imagery to bridge this gap and compile a detailed map of the health of vegetation in the Noril'sk region. This type of environmental analysis has not previously been attempted for an area so remote and climatically extreme, and therefore presents several methodological challenges.
I have experimented with a combination of recently developed digital image analysis methods and Geographical Information Systems technology and demonstrated that the Normalised Difference Vegetation Index is not sufficient to characterise industrially damaged high-latitude vegetation; its remaining potential is in change detection.
After field visits to the Noril'sk area, I formulated a algorithm to produce a comprehensive and faithful vegetation map: the hybrid multispectral, multisource, and multiseasonal classification. Such a classification requires several inputs: results of unsupervised spectral clustering of the main satellite image; training data based on field investigations; existing vegetation maps; and phenological correction data from an additional satellite image taken at a different stage of the vegetation season. The result is a 20-class map of the contemporary vegetation state in the Noril'sk region, compiled from 1995-1998 satellite data. Of the total 8529 km2 study area, 1963 km2 were found to be occupied by dead and heavily damaged vegetation. This includes 838 km2 predominantly covered by dead larch forest with varied vegetation underneath. I organised a validation expedition to the area in 1998 and found that the classification accuracy is about 70%. It could be even higher in reality, because the validation was hindered by the limited location accuracy of Global Positioning System receivers.
Then I analysed phenological changes in the Noril'sk area using data from early July (the beginning of the vegetation season) and early August (just after the peak of the vegetation season). Prior to this I improved image comparability by linear regression based on calibration targets not affected by phenology. The analysis reveals important increases of the vegetation vigour on mountain slopes after the snowmelt, and its decreases in the lowland part of the area, as a result of both natural and pollution-induced senescence. The derived data allowed me to develop a novel technique of phenological correction for multi-year change detection analysis, using 10-day composites of Normalised Difference Vegetation Index images from the AVHRR sensor. By brief change-detection analysis of 1972, 1985, and 1995 satellite data I have shown that without such a correction phenological differences may completely disguise the effects of industrial degradation. The analysis also allowed me to identify an area of considerable degradation south-west of Noril'sk that appeared after 1972. Degradation immediately south-east of Noril'sk was relatively small, since it was already a technogenic barren area in 1972. Examination of 1961 satellite photography of the immediate vicinities of Noril'sk reveals, however, a 21% decrease in green vegetation by 1995.
Finally, I have proposed a methodological framework for assessing the remediation potential of the ecosystems and for planning remediation in the Noril'sk region within a Geographical Information System (GIS). Such assessment should be based on joint analysis of geological, geocryological, soil and botanical factors, with the use of satellite and aerial images, maps, field materials, first-hand knowledge of the area and in co-operation with local scientists. The positive remediation experience of the nickel and copper smelter complex in Sudbury, Canada is also very instructive. In conclusion, I outline further areas of research, which include further improvements to the classification and phenological correction algorithms, design of the satellite analysis component for the proposed GIS; and the analysis of SAR and ASAR data to improve the identification of dead forests and to reduce the dependence of the monitoring techniques on weather conditions.
Written by Dr Olga Tutubalina, August 2001