A Map of the Vegetation of the Territory (Kray) of Khabarovsk

Thomas A. Stone and Peter Schlesinger
The Woods Hole Research Center
Falmouth, MA 02540

Introduction

The goal of this project was to create a 1-km resolution landcover map of Khabarovsk territory (see Figure 1) of the Russian Far East that is geographically accurate and in a digital format. Initially, we suggested that we would provide 10 to 15 landcover classes but that number has been increased to 50 landcover classes to better accommodate the complexity of the landscape and to improve the utility of the map in modeling the biomass and carbon content.

Figure 1. Khabarovsk’s location, close to Japan, N. Korea, and China, make its forests highly vulnerable to exploitation.

Analysis of the region with satellite data allows the creation of maps and information independent of bias, inaccuracy, and uncertainty in the literature. For example, there is immense variance in estimates of the area of Khabarovsk, the current area of forest, and the amount of forest under management. These estimates range from 824,000 km2 (USSR Atlas, 1986) to 790,000 km2 (Newell and Wilson, 1996). Estimates of the area managed by the Russian Forest Fund (Lesnoy Fund) range from 779,000 km2 (Alexeyev and Birdsey 1998) to 700,000 km2 Efremov et al. (1999). However not all areas controlled by the Forest Fund are actually forested; Alexeyev and Birdsey state that 604,000 km2 (78%) are forest covered. At the low end the Sustainable Forestry Pilot Project states that the Forest Fund of Khabarovsk covers 589,505 km2 of which 74% or 435,966 km2 are forested. The variations in these estimates of the Forest Fund area and forest-covered area may be due to uncertainty in boundaries of forestry enterprises (Leskhoz) and overlapping or unclear responsibilities in management of the forests of Khabarovsk.

The forests of the region are managed through the Leskhoz system. In Khabarovsk there are about 45 Leskhoz. In general, the area covered by each Leskhoz is larger northward as the quality and density of timber resources diminishes. The number and boundaries of the forestry enterprises change over time.

The landcover map is based on satellite data from 2000 and 2001 and is independent of any previous map of the region. Fifty initial classes were combined into eight classes by comparison with a Russian forest map. This new map is the first we are aware of that includes fire scars from the extensive fires that burned about 25,000 km2 according to Kasischke et al. (1999) in the region in 1998. The resulting map can be used to estimate the primary productivity of the vegetation of the region. Primary productivity is directly related to the biomass of standing vegetation and its carbon uptake.

Data Used

We chose imagery from the SPOT VEGETATION satellite as the primary data set. This sensor was launched in March of 1998 on board SPOT-IV. Khabarovsk is imaged daily by the VEGETATION satellite, and the images cover a swath 2,200 km wide and with a spatial resolution of 1 km. We used the S10 data set, ten-day synthesis data of the Normalized Differenced Vegetation Index (NDVI). VGT-S10 products are compiled by merging segments (data strips) acquired over ten days and delivered as map-projected data. These products provide data from all spectral bands, the NDVI, and auxiliary data on image acquisition parameters. Three ten-day composites are made during a month. Use of the NDVI is extensive and its strengths and weaknesses are well known (Tucker 1979, Tucker et al., 1991). To reduce the data volume, we chose only data from summer months when the NDVI values are expected to be high. These data covered the period from 6-01-2000 to 8-31-2000 in nine ten-day sections and the period from 6-01-2001 to 8-31-2001, again in nine ten-day periods. Complete documentation of these data including their acquisition, formatting, and production is available at the image processing and archiving centre served by Vito Belgium, with the assistance of the Belgian Science Policy Office.

We obtained complete SPOT VEGTATION S10 data for all of Northern Eurasia and extracted from these data the territory of Khabarovsk plus a 10-km wide buffer around the territory using an ArcView shape file of the administrative regions of Russia. We merged all composite files (nine each from the summers of 2000 and 2001) into one file making an 18-band image of the territory. Due to funding constraints, this project was limited to using either imagery data on hand at WHRC or free satellite data.

We had hoped to use free 500m resolution MODIS data. However, the production of MODIS data by NASA has been much slower than anticipated and MODIS data formats have not stabilized, resulting in the slow development of software tools for data analysis.

Several ancillary data sets included the following.

LANDSAT Data . This extensive data set covers (as of Feb. 1, 2002) 42 Landsat images of the far eastern region of Russia at 30m resolution with dates from 1988 to 2001. All these data were acquired with other funding.

ASTER Data. ASTER is the highest resolution sensor on board the new NASA TERRA platform (launched Dec 18, 1999). We have five scenes of data along the northern Khabarovsk coast. ASTER's special strength is in thermal imaging. Data are acquired on demand.

Leskhoz Data. We have complete Leskhoz forest boundary data of the Russian Far East from Russian colleagues. We do not know the dates of the vectors, but we assume they are for 1993. Dimity Dmitry Efremov has provided us with a graphic of the current (2002) boundaries that we shall convert for digital use. We have not been provided the attributes of the Leskhoz polygons beyond their names and locations.

We have a Habitat Map of Southern Khabarovsk (and all Primorye) constructed in part by tiger researcher Dr. Dale Miquelle (of Hornocker Wildlife Institute - World Conservation Society) and Russian colleagues.

More than 40 NOAA AVHRR LAC (1 km resolution) images have been provided by Dr. Nikolay Minko of the Russian Institute of Solar Terrestrial Physics, Center of Remote Sensing in Irkutsk. NOAA AVHRR images are weather satellite data commonly used for daily mapping of fires. Dr. Minko has also provided us with plot locations of forest fires in the entire Russian Far East for the years 1998, 1999, and 2000, based on his research.

Topographic Data. Our topographic data were obtained from the EROS Data center of the USGS (see http://edcdaac.usgs.gov/gtopo30/hydro/as_dem.html ). These data have been resampled to 1 km resolution. After extracting the territory of Khabarovsk from the Asia data, we converted it to an ERDAS Imagine image using an ArcView Shapefile of the administrative regions of Russia. Elevation in the image ranges from 0 to 2810 meters. Khabarovsk has a significant amount of relief and little flat land. Despite bordering the Pacific Ocean over its entire N-S extent of about 1,800 km, over 50% of the territory is above 500m in elevation, 30% is between 500m and 1000m in elevation, and 20% is above 1000m in elevation. Therefore, conditions for significant forest growth and production are limited.

From the topographic data, we created an aspect (the compass direction of the slope) image. Aspect is very important particularly in northern latitudes, as it is highly correlated with the presence or absence of permafrost, which can exert a fundamental control on the type and vigor of vegetation. This file is available for future work.

Methods

Classification of the SPOT VEGETATION data was done as a simple unsupervised classification (ISODATA) using ERDAS Imagine software and the 18-band SPOT S10 NDVI image, (consisting of 18 ten-day composites). We chose to create 50 landcover classes based on the SPOT Imagery. As the SPOT data are NDVI data, over the course of a summer they will follow the phenological cycle of the vegetation with the only exceptions being areas of bare soil, snow, or water. Consequently, for vegetated areas each of the classes developed will have a similar phenology.

Figure 2. Aspects of the 50 landcover classes of Khabarovsk as defined in this work. In general, the higher the summed NDVI, the higher the vegetation productivity over the two seasons. The greater the normalized standard deviation, the more variable the NDVI values were over the two summers, perhaps due to weather or variables related to deciduous versus evergreen forest types.

The 50 landcover classes we developed are in a continuum from very low summed NDVI (meaning no or very low vegetation vigor), to moderate summed NDVI with a wide range of standard deviations, to very high summed NDVI with low standard deviations (Figure 2). In general, we would expect evergreen forests to exhibit very small NDVI changes and lower standard deviations over the summer. We expect deciduous forests to exhibit high variability and hence higher standard deviations over the summer.

Finally, we have taken the summed NDVI values (column 4, Table 1) of the 50 classes derived and have separated them into seven levels of relative vegetation productivity (null, very low, low, medium low, medium high, high, and very high). This division shows the geographic distribution of the most and least productive regions (Figure 4) and offers insight into the geographic distribution of biomass throughout the region.

Results

We have created two versions of a one-kilometer resolution digital map of Khabarovsk territory’s current landcover based on phenology, elevation, and information about recent forest fires. The fifty-class landcover map demonstrates the significant complexity of the region’s landcover combined with its high variability in terrain, a component not recognized in any other landcover map of the region.

The eight class version of the map shows clearly the relative distribution of simplified forest species complexes (e.g. larch-conifer, spruce-Korean pine) interspersed with areas of non-forested land heavily impacted by human activity in the south, especially along bays, rivers, and other transport corridors. This map minimizes the terrain component to a regional level, showing the general effect of the region’s physical geography on its forest resources. The map highlights the growing lack of contiguity in the larch-conifer component, due to fire, logging, and, to a much lesser extent, agriculture.

Figure 3a. (left) Classified SPOT S10 data of Khabarovsk with color assignments for 50 landcover classes (See Table 1). We used 9 ten-day periods of NDVI data from the summers of 2000 and 2001 and an unsupervised classification algorithm to create the map. This image is in a UTM map projection. Figure 3b (right) shows an eight class version of Figure 3a. Here, non-forest is gray, rocks and outcrops are dark red, open larch is pale yellow, larch is light brown, larch/conifer mix is green, spruce/larch mix is purple, spruce/Korean pine mix is bright red, and larch/birch mix is light blue.

Table 1: Class descriptions, dominant Russian Forest Map Type, area (km2 ) and 2 year summed vegetation index (VI) for each of the 50 landcover classes defined.

Figure 4. Vegetative productivity levels in Khabarovsk based on two growing seasons of SPOT Vegetation data. The darker green the color the higher the vegetation productivity. Not surprisingly, the regions of high productivity are in the southern part of Khabarovsk.

Discussion

Using the two summer time series data from SPOT allows an analysis that emphasizes phenological change of the vegetation cover over time. As the NDVI composites use the maximum NDVI value over a ten-day period, this does not preclude the possibility of cloud or smoke contamination or degradation of the NDVI signal when fires persist.

The frequency of forest fires in this region is highly variable and their effects are very widespread. It is therefore critical to use the most recent satellite and field data. Field data or satellite data from years 3 or 4 years previous, for example, leaves the researcher open to badly overestimating the extent of current standing forest. Forest regrowth is extremely slow in this region, and there is little likelihood of its overestimation.

Estimates of forest loss from fire in this region are uncertain. Similarly, the amount of logging is poorly known. Illegal logging may be as much as three times the legal amount, resulting in a significant loss of timber exports to China and Japan (Newell and Lebedev, 2000).

Conclusions

SPOT Vegetation data can be used to create a current 50-class landcover classification at 1 km resolution of the entire territory of Khabarovsk in the Russian Far East that allows for an independent method of estimating vegetative productivity, carbon uptake, and fire extent.

This result is enhanced with comparisons of lower resolution forest cover maps and by interactive comparison of derived classes with topographic data. We have also produced a simplified version of the map with eight landcover classes and produced a vegetative productivity map that has divided Khabarovsk into seven regions for the 2000-2001 growing season. The results, summarized in Table 1, can be used to estimate the standing biomass of the region with field data on forest biomass and productivity throughout the territory of Khabarovsk.

These maps provide a fresh review of the vegetation of the territory of Khabarovsk, uncomplicated by the biases of other generations; they capture the complexity of the land surface, the variability of its terrain, and forest fire impacts. The work can be repeatedly improved, successively with new data readily available each year; information content might be further increased with additional imagery data of higher resolution. Still, these maps represent the best current understanding of the landcover of Khabarovsk.

References

Alexeyev, V., and R. Birdsey (eds.), 1998. Carbon Storage in Forest and Peatlands of Russia. USFS Northeastern Research Station, General Tech. Report NE-244, 137 pp.

Efremov D. F, L. Carlsson, M. Olsson, and A. Sheingauz, 1999. Institutional Change and Transition in the Forest Sector of Khabarovsk Krai, IR-99-068, IIASA, Laxenburg, 72 pp.

Kasischke, E.S., K. Bergen, R. Fennimore, F. Sotelo, G. Stephens, A. Janetos, and H.H. Shugart. 1999. Satellite imagery gives a clear picture of Russia's boreal forest fires. EOS Transactions of the American Geophysical Union 80, 141, 147. 1999.

Newell, J., and E. Wilson, 1996. The Russian Far East. Friends of the Earth – Japan, Tokyo, 200 pp.

Newell, J., and A. Lebedev, 2000. Plundering Russia’s Far Eastern Taiga. Friends of the Earth Japan & Bureau of Regional Oriental Campaigns & Pacific Environment and Resources Center, 47 pp.

Tucker, C.J. 1979. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sensing of the Environment, 8:127-150.

Tucker, C. J., W. W. Newcomb, S. O. Los, and S. D. Prince, 1991. Mean and inter-year variation of growing-season normalized difference vegetation index for the Sahel 1981-1989. International Journal of Remote Sensing, 12:1113-1115.

Soviet Ministry of Cartography and Geodesy, 1986. Atlas USSR (in Russian), Moscow. 259 pp.

©Woods Hole Research Center, 2007