ABSTRACT

In land use planning, mapping the current land use or land cover situation is an indispensable tool for determining the status quo and for identifying land use trends. Land use maps provide the basis for legal documents and create an entry point for discussions with local land users and stakeholders on improving land management practices in order to achieve sustainability. Unfortunately, land use is notorious for changing rapidly and existing maps are quickly outdated. In many cases, land use planners have to prepare new land use maps or update existing ones before actual planning can start. Using remote sensing data sources like Landsat TM and SPOT data have long been established as an effective means to prepare land use maps at scales ranging from 1:100.000 to 1:50.000. These data sources are however not suitable for the preparation of village-level maps, requiring scales of 1:25.000 to 1:10.000.

This paper illustrates the potential of the Indian panchromatic IRS-1 satellite data with a spatial resolution of 5.8 m for the purpose of large scale village-level land use planning. The technical approach applied consisted of using geometrically corrected IRS-1 image data acquired for a test area in Thanh Hoa Province. The data were enlarged and printed at a suitable scale (1:12.500), using a high quality color printer and photo quality paper. Printed satellite image maps were taken to the field and present land use/cover was mapped on transparent overlays using a simplified land use classification. The field maps were later digitized and converted into a GIS dataset.

INTRODUCTION

In many areas in southeast Asia, large-scale topographic maps and aerial photography are not available or are too old and outdated to be considered useful for the mapping of present land use. Mapping based exclusively on field surveys, on the other hand, is very time consuming and error-prone without the use of adequate base maps .

In the Forestry Sector Project in Vietnam, village-level land use maps have to be prepared at scales ranging from 1:25.000 to 1:10.000 for project areas comprising approximately 600.000 ha, located in 53 communes in four provinces. Neither Landsat TM with a spatial resolution of 30m nor SPOT Panchromatic data (10m) can be used to produce acceptable maps at these scales.

The key question facing the project team therefore was: how can village-level land use maps be prepared in large numbers (approx. 500 villages), within a reasonably short time period and at acceptable levels of accuracy and cost. The use of the recently available, high-resolution, Indian IRS-1 satellite data with a spatial resolution of 5.8 m was considered an option worth while investigating.

IRS-1 VS. LANDSAT TM AND SPOT

Satellite image data such as Landsat TM or SPOT XS have valuable characteristics due to their multi-spectral characteristics (7 bands with TM and 4 bands with SPOT), which can be used to provide a lot of quantitative information on land use and vegetation cover. On the other hand, the spatial resolution of these sensors is limited to 30m with Landsat TM and 20m with SPOT XS) which restricts their use for large scale mapping. Panchromatic (black and white) SPOT imagery has a spatial resolution of 10m, but the IRS-1 panchromatic sensor provides even higher resolution at 5.8 m (Table 1).

The higher spatial resolution of the panchromatic data permits users to distinguish details like road networks, irrigation channels and field boundaries and even individual buildings or fish ponds. The monospectral character of the data, however, restricts the possibilities to distinguish different vegetation types directly on the image. Using panchromatic image data therefore requires extensive ground surveys to determine different land use and vegetation cover classes (Figure 1).

In terms of spatial resolution, aerial photograph can prove to be better map data sources than commercial optical satellites. However, availability and access to recent aerial photographs is often limited and new aerial surveys are time consuming, costly and difficult to arrange. Another disadvantage is the limited area coverage of individual aerial photographs. While satellite image scenes cover more than 60 x 60 km, an aerial photograph at a scale of 1:25.000 covers an area of only 5 x 5 km. Thus, large area coverage need to be mosaicked from geometrically corrected and mosaicked aerial photographs, a task far more time consuming than correcting a single satellite image scene (Figure 2).

Table 1: Characteristics of Satellite Image Data Sources for Land Cover Mapping

Table 2: Cost Comparison for Satellite Image Data Sources

Note: Older imagery and multi-temporal data can often be obtained at discounted rates.

METHODOLOGY

The technical approach chosen combined a number of computer-assisted tasks with field surveys to determine the current land cover in one commune in Thanh Hoa Province. Raw IRS-1C image data were obtained and geometrically corrected using topographic map references and, at a later point, differential GPS data collected in the field. The Satellite image data were enlarged and set in a map layout at a suitable scale (1:12.500 and 1:25.000) and size (A2, 60 x 40 cm). Printing was done on photo quality paper using a high quality color inkjet printer with a resolution of 1440 x 720 dpi. Printed satellite image maps were mounted on thin plywood boards and taken to the field for present land use/cover mapping on transparent overlays. Draft maps produced in the field were finalized on location and digitized upon return to the office. Commune and village boundaries, roads, rivers and land cover classes were then converted into a GIS database.

Image manipulation and SIM preparation included the following tasks:

• Geometric correction
  GCPs were collected on the topographic maps. Verification of reference points with DGPS showed that better corrections can be obtained using DGPS.
• Mosaic
  Has to be applied in the case where one commune spreads into more than one image scene. It these cases image enhancement becomes more difficult.
• Enhancement
  As a rule, linear stretch is the best
• SIM
  To create a satellite image maps from the enhanced image, some additional information should be added to the map layout before printing, such as:

In the field, after determining the observation position, mapping was done by using a combination of field surveying techniques and image interpretation. The image was used to determine reference points in the field (houses, bridges, roads, rivers, etc.) and to identify recognizable boundaries between land classes. The most important key in image interpretation was the gray tone of the objects and their texture. Field observations are needed to identify land classes undistinguishable in the SIM and to map their location.

The following challenges were noted by field user who were using the SIMs for the first time:

• Positioning, without using topographic maps
• Identifying different land classes with similar visual representation on the SIM (like rice field and lake; sugarcane and bamboo)
• Identifying object of the same land class looking different on the SIM (like clear and muddy water, wooden and cement houses).
• Topographical features are difficult to recognize

Possible Solutions:

• For positioning: use GPS
• For interpretation: conduct more intensive initial training
• For identification of topographic features: Print contour lines into SIM or use overlay

Figure 3: Flow chart of work methodology

RESULT AND DISCUSSION

As result of the methodology applied, a detailed land cover map of the commune and villages of Xuan Cao was produced and area totals per land class and village were calculated. In addition, the SIM can be used as base maps for other project activities, like infrastructure planning, site matching for forestry activities etc. The field work needed to map present land cover and demarcate village boundaries amounted to approximately one week, implying two field teams and some additional time requirements for explanations and limited initial training.

At present, we do not feel confident to make a judgment about the time requirements for this type of land cover mapping in comparison with traditional mapping methods practiced in Vietnam using (enlarged) topographic maps. Most people involved, however, were convinced that the mapping accuracy was considerably better using the SIMs. We will conduct further studies to come to a clearer understanding of the time required for both methods and the quality that can be achieved respectively.

The question that remains to be answered is the cost involved in preparing and using SIM maps. The cost of using SIMs depends to a large extent on the size of the area to be mapped within a satellite image scene. Provided that approx. 50% of a full scene are used, the cost data cost will not exceed one US dollar per km², or one cent per ha at 10% use the cost is 0.1 US-$ per ha. Additional cost for data processing and map printing depend to a large extent on the level of in-house data processing and manipulation possible versus the amount of services that would need to be contracted out. Total data processing and SIM preparation can be estimated to cost roughly the same as raw satellite data. For the purpose of orientation, one can estimate costs at approximately two US dollars per km² for using 50% of a satellite scene.

CONCLUSION AND RECOMMENDATIONS

IRS 1-C data can be applied successfully for village-level land use planning at scales ranging from 1:10,000 to 1:25.000, due to their high spatial resolution. However, land cover mapping with IRS1-C SIMs requires extensive field surveys and cannot be understood as an image-based interpretation. The SIM provide a highly accurate base map which can be used to identify important features like roads, rivers, building and field boundaries. The land units identified on the image provide a reference system for the field work and can prove a cost-efficient option that yields accurate map results.

Using IRS-1C data successfully for field mapping of village land use and vegetation cover would yield better results if field staff are trained to understand the specific characteristics of the panchromatic image data and have access to additional data sources from GPS and topographic maps to facilitate field orientation.

Figure 1. Comparison of IRS-1 with Landsat TM

Figure 1: Comparison of IRS-1 with LandsatTM

a. IRS-1, 4 June 1998, original resolution
b. TM, 15 November 1997, original resolution
c. TM, 15 November 1997, enlarged

Figure 2. Comparison of IRS-1 image with aerial photo

A: Aerial photo of Xuan Cao commune, 1978; B: IRS –1 image of Xuan Cao commune, 1998

Figure 4. SIM of Xuan Cao commune

Figure 5. Land Use map of Xuan Cao