Home Articles Forest resources study in Mongolia using Remote Sensing and GIS

Forest resources study in Mongolia using Remote Sensing and GIS

ACRS 1994

Forestry

Forest Resources Study in Mongolia
using Remote Sensing and GIS

B. Enkhtuvshin, M. Ganzorig, D. Amarsaikhan, H. Tulgaa

Informatics Centres, Mangolian Academy of Sciences
Av. Enkhtaivan – 54B, Ulaanbaatar-51, Mangolia


Abstract

Forest inventories and studies require and extensive use of RS data. To make a rapid decision making the forest management agencies need the real time up-to-date information. The aim of our study is to detect the forest changes and carry out the related analysis. For this purpose, multitemporal RS and GIS data were compiles within the ERDAS environment. As a result of the analysis forest changes for more than 10 years and the anthropogenic influences have been determined.


Introduction

Forest management requires reliable inventory data and the maps indicating current state of the forest area. Forests and forest land are an important natural resource in many part of the world and provide the raw material for a wide range of wood based industries (5). Deforestation has become a global problem in many developing countries and it has a direct correlation with population density, and the resource on forest resources is caused mainly by ever increasing populations and some development activities (94). Mongolia has abundant forest recourse in comparison with its over 2 million inhabits.

However, the forests are decreasing because of different natural and anthropogenic influences. One of the reasons could be a rapid growth of population and urbanization process for example, during the last decades, the population and area of Ulaanbataar have been significantly increased. RS provides a real time information on the status and condition of the forests. Integrating RS data with other traditional and ground truth information one can performa a truth information one can perform a through analysis and advise the forest managers for better planning.

Test Area and Data Sources

As a rest site ‘Bogd-Uul’ mountain situate in the central part of Mongolia, near the Ulaanbaatar city was selected. The mountain is the nature protected area covered by the forest ecosystem having about 60% of uniformly distributed deciduous forest. (Fig. 1).


Figure 1 Location of the area

For the forest change study the following data have been used :

  • Forest taxonomical map of 1963, scale 1:50, 000
  • Forest taxonomical map of 1988, scale 1:50, 000
  • Topographic map, scale 1:50,000
  • AFA analogue panchromatic data of 1962, 1963, 1975
  • SPOT-XS of 1986
  • SPOT PAN of 1989
  • Land sat TM (7 bands) of 1988
  • Ground truth data

ACRS 1994

Forestry

Forest Resources Study in Mongolia
using Remote Sensing and GIS


Forest Change Analyses

At the beginning the taxonomical maps and AFA photographs of the study area were converted to digital form by the use f CCD came and scanjet Plus scanner, respectively. Then, screen digitizing of the maps was done using the DIGSCRN’ – module. To detect the major forest changes, an air photograph of 1963 and a taxonomical map of 1963 were compared with the air photograph of 1975 and taxonomical map of 1988.


Figure 2 Comparison of multitemporal data

The analysis indicated, that the fire, ripe age deciduous forest was restored and instead, the young age forest has been formed. To investigate the further restoration process we have used Landsat – TM, SPOT-XS and SPOT-PAN data taken in different seasons.

By analysis of multitemporal air photos and satellite information anthropogenic influence as forest fires, felling, cattle herding etc the modern condition of the forest ecosystems have been detected.

SPOT-XS and Landsat-TM data were classified into coniferous, deciduous & mixed forest area and in further the coniferous was classified into subclassed by area. For the quality improvement of Landsat-TM and SPOT-XS a high spatial resolution of SPOT-PAN has been used. To do this, at first both multispectral data were normalized and then multiplied by panchromatic data.

The further analyses require a creation of a DEM because it was necessary to see the forest changes and burned area from various sides by different angles for this purpose, contours were digitized from the topographic map scale 1:50,000 and then interpolated. The created DEM integrated with SPOT-XS image which indicates the burned areas, is shown in figure 3.


Figure 3 DEM and detection of the burned area

For the identification of the forest areas the following techniques have been applied to the RS images:

1. Normalized Difference Vegetation Index (NDV) maps were estimated for the SPOT-XS as

NDVI = (b3 – b2)/(b3 + b20
NDVI = (b2*b3/b1**b2

and for the Landsat TM as

NDVI = (b4 – b3) / (b3 + b4)
NDVI = b2 **2/b1*b3
NDVI = b3 * b4/b2**2, etc

More details about application of this technique are described in (3). Data cloud of SPOT-XS data was almost in the origin and was rotated 24 degree. Fro the intensity of TM data 6 bands, excluding the thermal band have been used and all the transformations applied. For the final affine transformation of the images the following matrices were used:

3. Principle Component Analysis (PCA) for TM data and the Following PCs were

PCI=89.21%; PC2=6.8%; PC3=2.21%; PC4=1.11%; PC5 =0.72%; PC6= 0.04%

4. Rational and supervised classification using maximum likelihood decision rule. Almost same procedures used in (3) were used here.

Colour differences in the images indicate such forms of the forest spreads, like missed forest, deciduous forest, coniferous forest, reservoir and pasture. After applying saturation enhancement in TM data different hues of the coniferous forest have been formed.

Analysis of the hues indicated that there is the correlation between the colour tones age classes. Thus, in the image 4 groups of age classes : your, ripe old and pasture/deciduous have been distinguished.


Conclusions

As seen from the analyses, multitemporal air and satellite data together with other cartographic and ground information can be effectively used for the study of the condition of the forest ecosystem.

In result of the digital image processing, GIS analyses and visual interpretation of the available RS data it is concluded that primary satellite data (Scale 1:100, 000 and less) can be used to detect the structure of the forest formation and macro-geocomplexes, where as RS data with enlarged scale (Scale 1:50,000 n larger) can be used to interpret the structure of classes and groups of forest associations as well as meso and microgeocomplexes (2).


References

  • Brown, A., Hiong, J.T and Weir, M.J 1989, An Experimental Forest Map Design based on the FAO/UNEP Forest Classification System, ITC Journal 1989-3/4, 212-2217.
  • Gannzzorig, M., Enkhtuvshin, B., Amarsaikhan, D., and Tuglaa, H., 1994, Application of RS and GIS for Regional Feature Change Monitoring, Paper to be presented at the International Symposium on Space Technology and Applications for Susbtainable Development, Beijing, China.
  • Ganzorig, M. Amarsaikhan, D., Enkhtuvshin, B, and Tuglaa, H., 1994 Design of a multilevel Database using RS and GIS Techniques, Paper to be presented at the 15th ACRS, Bangalore, India.
  • Sudhakar, S., Krishan., Das P.K and Raha A.K, 1992 Forest Resources Management Survey with IRS-IA, Is Journal 1992-3, 277-285.
  • Susilawati, S., and Weir, M.J 1990 GIs Applications in Forest Management in Indonesia ITC Journal 1190-3, 236-245.