Land Use Changes in the Upper Ca River Basin,
Xiengkhuang Provinces, Lao PDR 1995 – 97:
Effects of Roads and Rivers

ABSTRACT

This paper presents the results of land use and land cover change detection using Remote Sensing and Geographic Information Systems in the Northern Lao PDR. The study area was divided into 16 zones by generating buffers from roads and rivers to compare the land use and cover ratio. In general, the results show that, between 1995 and 1997, forest cover increased by 0.56 %, agriculture-residential land decreased by 0.12 % and wood-shrub land decreased by 0.4 %. It is found that there is low ratio of forest cover close to the road and rivers, where others land uses are more concentrated. The analysis shows that digital technique is very useful for investigating land use and land cover in term of time and investment.

TABLE OF CONTENTS

Introduction

Materials and Methods

Results and discussion

Conclusion

References

INTRODUCTION

Land use and land cover information have become important for land use planning and resource management, yet currently used mapping and monitoring methods cannot address the needs of forest managers and environmentalists. Using or relying on ground surveys and sampling alone requires manpower, expenditure and time. Recent developments in remote sensing technology indicate that, if these methods are carefully combined with reliable ground based data, it is possible to compile detailed inventories of, and to monitor natural resources. Such analyses include the relationships between changes in forest zones and socio-economic development factors. Remote Sensing (RS) and Geographic Information Systems (GIS) are used for more detailed analysis of collected land use information. Remotely sensed data obtained by sensors from a high altitude platform such as the Landsat satellite, would be a good alternative to the ground survey approach. These information systems compile multi-spectral data, which can form a common database for integrated resource inventories. Broad vegetation-type stratification is reported by using Landsat Multispectral Scanner (MSS) and Thematic Mapper (TM) in digital and analogue form; forest types typically have distinctive spectral reflectance patterns. Studies indicate that the technology can provide valuable information with respect to forest resources (Roy et al. 1985, Unni et al. 1985, Jadhav et al. 1987).

This study is part of an environmental assessment of the Ca River Basin, carried out by the National University of Laos (NUOL), Center for Natural Resources and Environmental Studies (CRES) of the National University of Hanoi, the University of Vinh, Vietnam. Only the results of the Lao section are given here and complement the results of a socio-economic survey carried out by NUOL. The main objective of the study is to assess the potential impact of road upgrade on land cover. This paper presents the results of land use and land cover change detection between 1995 and 1997 and to measure the changes close to roads and rivers.

MATERIALS AND METHODS

Materials

Four raw satellite images, taken in 1989, 1992, 1995 and 1997 (Landsat TM band 234, Path 127/128, Row 46/47) were, rectified and provided by World Resources Institute (WRI) for this study. However, the first two dates were not used due to high cloud-covers.

Location of the Study Area

Auxiliary data such as topographic maps scale 1:100- 000 and land use maps from Department of Forestry were used to delineate roads, rivers and to assist image interpretation.

The computer software used to combine and analyze the images were Copenhagen Image Processing System (CHIPS), Version 4.3 and PC ARC/INFO, Version 3.5.1 for the GIS analysis.

Methods

The study area is the Ca River Basin (CRB) located in the northern part of Laos. With an area of 9342 sq. km, it includes parts of Xieng-khuang and Huaphan provinces. Nam Nuen (Ca River), one of the three main rivers, flows from Laos to Vietnam and it is the main river in the watershed.

Landsat images were processed and analyzed by Copenhagen Image Processing System (CHIPS). The supervised maximum likelihood classifier was used to transform multispectral data into thematic maps. Training areas were selected based on visualization of false color composite (FCC), band 432. Training areas for each land cover type ware identified separately using sunshine and shadow perspective to reduce interpretation errors. The two results were then merged.

Each image was classified separately and merged into a change detection image. It was then filtered to remove single pixel artifacts caused by misregistration and system noise. The accuracy of the classification results was evaluated through comparison with the land use maps produced by the Forest Monitoring Project. Randomly selected reference pixels were inspected at the corresponding sites to verify the classification result. A ground truth survey was not carried out to confirm the final image, due to insufficient time and budget.

Buffers from Roads

Other information were extracted from topographic maps. Information on watershed boundary, village location and the drainage network was digitized into Geographic Information System (GIS).

The study area was divided into 16 zones, moving outward from roads and rivers. Each zone was two kilometers wide except the last one.

These zones were created to overlay with the classified images to generate statistics and compare the changes.

Flowchart of the analysis procedure'

RESULTS AND DISCUSSION

Results of the RS/GIS analysis, which develop an overview of current forest and land-use patterns, were used to examine the relationship between road and land cover in the watershed. Table 1 presents the relationship between various land-cover types in the Lao section, based on interpretation of 1995 and 1997 satellite images.

Table 1: Proportions of land cover types in 1995 and 1997

Note: Cloud cover from the 1995 image was added to the 1997 image to facilitate comparison.

Some confusion was noted between barren land, which includes active shifting cultivation, and residential areas. This is due to the fact that the high spectral signature reflectance in both land uses is similarly. Therefore, they were merged into one category of land use type "Agriculture and Residential". "Wood and Shrub-land" primarily constitutes regrowth, which is fallow or abandoned land within a shifting cultivation cycle, for technical reasons, a minor proportion of paddy-land may be classified under Wood and Shrubland. When combined, these two classes made up 50.33 % of land cover in the catchment area in 1997, indicating the total area of land recently involved in some stage of an agricultural production cycle. This does not include natural grassland, which is used as pasture for livestock.

The data indicate that forest cover around 45% in both 1995 and 1997. This figure is close to earlier estimates of forest cover for the Nam Mat/Neun and Nam Mo/Khien Watershed Management Units (Table 2). However, when compared different data analysis methods on forest cover produced different conclusion. As table 2 illustrates, it is therefore not possible to say whether the Lao section of the Ca River Basin lies above or below the national average.

Table 2: Various forest cover estimates for Lao PDR and Ca River Basin, Lao section

The images showed, however that forest cover is quite fragmented in the catchment. Large areas are composed of non-forest land, while others show extensive tracts of forest - particularly in the South of Nam Mo watershed, and the Northern part of Nam Neun watershed. The area including the Nam Mat Watershed generally shows a lower forest cover, particularly along Road 7.

In comparing data between the 1995 and 1997, there appears to be a slight overall increase in forest cover (0.56%), while Agriculture/Residential and Wood and Shrubland cover has decreased slightly (0.12% and 0.44% respectively).

A preliminary forest cover analysis of 1989–1997 change carried out under the Ca River Basin Environmental Assessment showed a similar slight increase (0.31%) in the Lao section (WRI 1998). In the larger picture, the MRC/GTZ analysis of the whole of the Lao PDR also showed an overall decline (-0.9%) from 1993 to 1997.

Relationships between roads, rivers and land cover types

In order to examine the relationship between agricultural activity, forest cover and the presence of major roads in the Ca River catchment, land cover percentages were calculated for 16 individual "buffer-zones".

The data suggest a close relationship between agricultural activity and distance from major roads in the Lao section of the Ca River Basin. The zone closest to Roads 7 shows 16% "Agriculture/Residential land" in 1997, falling to some 4 % in the zone furthest from these roads. "Wood and Shrubland" which is mainly fallow land, corresponds with the first three zones, "Wood and Shrubland" follows the relative variations of Agricultural/Residential land. By contrast, "Forest" increases as the distance from road increases - growing from almost 23 % within the first two kilometers of roads in 1997, to 54% in the zone 28-30 kms distance from roads.

Variations between the two extremes are also evident in and around the "middle" zones. Trends reverse briefly in both years, with Forest declining and Agriculture/Residential and Wood/Shrubland rising. It is likely that these variations are caused by settlements near rivers in areas removed from roads.

In order to examine this more closely, further buffer-zone analysis was carried out on the major rivers in the catchment. The data indicate that proximity to the rivers in important for agricultural activity in the Lao section of the Ca River Basin. Agriculture/Residential and Wood/Shrubland land is highest in the two-kilometer zone closest to rivers, while Forest dominates the zones furthest away.

The overlap between roads and rivers in the Ca River Basin makes it difficult to draw conclusions from comparison between the two forms of analysis. However, compared to the road/land cover data, Agriculture/ Residential land is higher (4.77 % for 1997) in the area within two kilometers along roads than in areas within two kilometers of rivers. This suggests that the relationship between roads and active agriculture is even more pronounced than river/agriculture relationships.

Alternatively, Wood/Shrubland covers relatively less area in zones near roads compared to zones near rivers. This could indicate a pattern of more intense cultivation near roads, with less land left as fallow in these areas. This is supported by the lower percentage of Wood and Shrubland in areas between zero and four kilometers from roads, compared to areas between four and eight kilometers from roads. This difference is not noticeable in the river/land cover data.

CONCLUSION

The Lao section of the Ca River Basin shows a clear relationship between distance from roads, forest cover and land use. It shows areas close to roads and rivers display a higher percentage of agricultural activity than other areas, while forest cover increases progressively away from roads or rivers. The forest cover is fragmented and scattered slightly along the roads and rivers. Most large and dense forest areas are located in the steep and difficult areas that are hard to access, so land use is limited.

REFERENCES

Brunner and Nielsen: Draft "Ca River Basin Forest Cover Analysis: Preliminary Results" WRI, Washington. This paper which was written in 1998 presents preliminary findings of forest cover and analysis of a time series of Landsat TM Images of the Ca River Basin.

FAO (1998): Shifting Cultivation Stabilization project. Report No. 97/097, Vientiane.

MRC/GTZ, 1997 Land cover maps from 1992 and 1997 of Xiengkhuang and Huaphan Provinces.

Prida, T., (1991): Surface Water Evaluation in Northeast Thailand: A pilot Project Using Satellite Remote Sensing, Final Report, Asian Institute of Technology, BKK, Thailand.

Roy, P. S., Kaul, R. N., Sharama Roy, M. R., and Garbyal, S. G., (1985): Forest type Stratification and delineation of Shifting Cultivation areas in Eastern Part of Amanachal Paradesh Using Landsat MSS data, International Journal Remote Sensing, Vol. 6,411-418.

Jadhave, R. N., Spivastaver, V. K., Kandya, A.K. Sarat BuBu, (1987): Large Scale Forest Type Mapping Using Satellite Data, International Journal Remote Sensing, Vol. 6, 10-17.

Unni N. V. M., Roy P. S. and Parthasara, (1985): Evaluation of Landsat and Airborne Multispectral Data and Aerial Photos for Mapping Forest Feature and Phenomena in Part of Godavari Basin, International Journal Remote Sensing, Vol. 6, 419-431.