skip to primary navigation skip to content

Development of a phenological correction algorithm for remote sensing of industrial impact on high-latitude vegetation. Case study: Noril'sk, Northern Siberia

Development of a phenological correction algorithm for remote sensing of industrial impact on high-latitude vegetation. Case study: Noril'sk, Northern Siberia

This is a previous project which was run at SPRI within the former Remote Sensing Group. The content below was written in November 2001.



Multispectral imaging from satellites is increasingly used to monitor northern vegetation. The cost of satellite images has been decreasing and access to them becoming easier since the mid-1990s: low and medium spatial resolution imagery such as AVHRR (and more recently MODIS, MISR and the high-resolution ASTER) is distributed freely over the Internet; Landsat 7 ETM+ can be freely redistributed, and the improved Landsat acquisition plan led to a multiple cloudless land coverage in 1999-2000 (except some of the Antarctic).

Prior to the mid-1990s there were few cloudless images for remote northern regions. This study focuses on the town of Noril'sk and its surroundings in Northern Siberia, an area of extreme damage to vegetation, caused mainly by sulphur dioxide emissions (approximately 2 million tonnes per year). For this region, the black-and white CORONA photos of July 1961, the Landsat MSS images of September 1971, the Landsat TM images of July 1985 and 1995, and a number of Russian multispectral MK-4 photos are available.

The early July images and photos show the beginning of the vegetation season when most of the north-facing slopes are still snow covered and shrub vegetation on the slopes has not started to vegetate. Images from late July-early August picture the peak of the vegetation period with snow pertaining only above 800 m in mountains, the maximum overall development of vegetation, and its early senescence in the areas of intensive industrial impact.

It may be difficult to discriminate multi-year industrial damage to vegetation from its phenological dynamics if the multi-year imagery comes from different stages of the vegetation season. The phenological stage of vegetation is defined not only by the height and aspect of the terrain, but by prevailing vegetation species, microclimate, and soil,; and information about these is often insufficient. As a result, maps of vegetation dynamics may have significant "phenological noise".

Aims of the project

To correct maps of vegetation dynamics using a model of phenological changes based on images of low spatial and high temporal resolution.

Current results

A correction algorithm has been developed which currently consists of four stages:

  • Radiometric calibration of the images (including system correction and linear regression of the radiances of one image to the other using phenologically stable calibration targets).
  • Determination of the phenological position of the images (approximated by the number of frost-free days since the beginning of the season, obtained from meteorological data).
  • Finding the low spatial resolution phenological 'analogies' from a one-year series. The 10-day, 1 km AVHRR composites for 1995 from the USGS dataset were used. The vegetation indices (NDVI) were used instead of spectral band radiances, to compensate in part for the spectral range differences between Landsat MSS/TM and the AVHRR.
  • Subtraction of vegetation index differences (multi-year difference minus the low resolution phenological difference) to yield a dynamics image almost free of "phenological noise".
Analysis of the multi-year vegetation degradation around Noril'sk
[Click for larger version]
Diagram as described adjacent

This algorithm has allowed the analysis of the multi-year vegetation degradation around Noril'sk using images of 9 July and 13 September, where a direct comparison was not previously possible. (See diagram.)

Further work

It is intended to further improve the algorithm using daily AVHRR and MODIS imagery to account for non-linear phenological development, daily meteorological data and a detailed digital terraim model.


  • O.V.Tutubalina and W.G.Rees. 2001. Vegetation degradation in a permafrost region as seen from space: Noril'sk, 1961-1999. Accepted by Cold Regions Science and Technology.
  • O.V.Tutubalina. 2001.Phenological correction of satellite imagery to facilitate change detection of northern vegetation in the areas of industrial impact. In: Abstracts of the V International Conference on the Development of the North and Problems of Nature Restoration. Syktyvkar: Komi Science Centre of RAS, 5-8 June 2001, pp. 256-259.
  • The algorithm is explained in more detail in this thesis: Olga V. Toutoubalina. 2000. Remote sensing of environmental degradation in the north: case study of the non-ferrous metal industry in Noril'sk, Northern Siberia. Dissertation submitted in partial fulfilment of the requirements for the PhD Degree at the University of Cambridge (held in the University Library).