Ground Truth Survey for the Development of Current
Land use map, the western seaboard Regional
development Master plan study, Thailand
Ground truth survey preparation and operation
Before conducting a ground truth survey, researchers need to investigate thoroughly details of the images of the study area in order to identify interesting points or areas where ground truthing is needed. All images will be covered transparencies to protect images. After identification, all selected points or polygons will be delineated for further interpretation and classification.
The next step is to locate the points selected by comparing the image and the topo-sheets. When location is found, researchers need to plan the survey route (s) according to the distance, the order of places, their accessibility and available timing. To travel to the point (s) previously selected, researchers need to consult the GPS and topo maps, and in many cases, the local people. Once the location is found, the GPSs will be used to establish the coordinates-latitude and longitude, and UTM- of each survey point and record them on the interpretation key sheet. To establish the ground truth data, color (s), tone, shape, size, and texture of that point in the image will be identified and recorded, and compared to the true land cover. Color photographs showing the real land cover of the location will be taken. Explanation for each land use/land cover category will be provided on the key sheet according to its ordering survey point numbers. The survey point numbers will also given on the topographic map sheet. These numbers will correspond to the survey points.
Appropriate symbols describing land cover categories will be marked on the transparency for areas having same land use class. Interviewing farmers/owners/workers of the survey area for more details is necessary especially in area where land use activities are not clear to researchers. This information includes month as the image, and other related activities. During field survey, many errors on the topo map are always found. The main reason is that most data on the topo maps are not updated, therefore, they have been corrected and updated by the field survey and the images.
Each interpretation key sheet consists of descriptions of land cover; location; path, row and date of the Lnadsat-TM; UTM coordinates and/or latitude and longitude values
Taken by the GPSs; the survey point number; details of the topo map sheet; data of the ground truth survey; a photograph of the selected location taken from the field showing the real land cover and a photograph of the topographic map sheet showing the selected point; and comments. An example of the interpretation key sheet is shown in Figure 1.
Figure 1 The Final Key for Survey Point Numbers 11 and 12.
To read the final key, one need to have a print of the Landsat-TM image of the selected area with the identification of the survey point numbers and identification of the land use classes that shown in the interpretation key sheet. For example, Figure 1 shows key sheets of the two survey points namely 011 and 012. These two points can be identified from the copied topo map sheet and the Landsat image sheet as number 11 and 12. From the comments, color photos of the area and the topo map sheet, and the Landsat image interpretation which firstly interpreted by the survey team, one can easily interpret the satellite images. Interpretation key sheets have later been prepared and combined the satellite images. Interpretation key sheet have later been prepared and combined into 6 volumes represented 6 provinces in the western seaboard region.
5. Development of Land Use Map
The visual interpretation was firstly simply interpreted by the DOG as resulted in the 6 volumes of the interpretation key.s. GAC has further interpreted the whole images of the area to develop a land use map. Digital image processing was firstly applied to Kanchanaburi province using ERDAS software. The results of the supervised classification in comparison to visual interpretation have low accuracy due to complexity ofland use classes. For example, bare soil and buildings cannot be separated from each others. More difficulties are to separate the upland crops from forest plantation, and to distinguish forest types. Even paddy fields may be irrigated and non-irrigated, and some part appears in the images as open space when it is harvested. In this case the digital image processing cannot separate the real open spaces from the dried harvested paddy fields due to the same value of spectral responses. Finally, visual interpretation is done in integrating with the GIS database and the field survey data which provide large raw for this purpose.
Other GIS data such as the existing land use map at the scale of 1:250,000 and the topo map sheets at the scale of 1:50,000 have been used during the visual interpretation of the interpretation of the entire area. When the interpretation step is over, all results are joined for the preparation of current land use map as shown in Figure 2.
Figure 2 Current Land Use Map of Kanchanaburi Province Derived from Landsat-TM Data, 1995
6. Recommendations
Firstly, it is recommend that the satellite imageries should have been geo-corrected and overlaid by UTM and/or latitude and longitude values prior to conducting field survey. This will assure the survey team members in getting the right location of the real world and the exact spot on the image. Secondly, it is recommended that a powerful notebook computer with a CD-Rom drive can be utilized in the field survey computer while raw data from ground truthing can be directly can by directly input in it.
