Simulating forest
degradation: the application of a GIS-based Area Production Model in Kali
Konto, area in East Java, Indonesia Alfred de Gier, Johan Bode
and Yousif Ali Hussin AbstractInternational Institute for Aerospace Survey and Earth Sciences (ITC), Enschede, The Netherlands Fax: +3153874399 E-mail :mailto:degier@itc.nl" The spatial. version of the Area production. Model (APM) would permt , the simulation of future.land use change in response to changes in response to change in population, gross domestic product, and in agricultural productivity. Land transfers from forest to agriculture, for example, can be simulated in this way. The model was developed for the Food and Agriculture true Organization (FAO) of the United Nations by Nilsson in 1984. The model is used for long-term scenario development and analysis. In the Kali Konto area forest degradation into scrub, takes the place of ' transfer. of forest and to agriculture, due to strict law enforcement. The growing populat1on sells the forest products and uses the proceeds to purchase food products. It was found ear11er that a good relation ship exists between the area of agricultural expansion, as calculated by the APM, and the actually observed area of degraded forest. Forest policy makers can thus use the APM to simulate the effect of changes in the above factors, and decide on appropriate actions. Recently, an GIS-based APM was developed, by integrating it within the ILWIS-GIS. This permitted the inclusion and analysis of key operators, such as slope, distances, population pressure and population growth. This paper describes the results of the GIS-based APM in the Kali Konto area. Introduction One of the major functions of the Area Production Model (APM) is its capacity to simulate long-term land use changes, by varying several input variables, especially: population and population growth, gross domestic product and its rate of change, and land use and the rates of change in agricultural productivity. For a period of up to 50 years ,the APM calculates the required primary yields from agricultural and forest lands and matches this by appropriate changes in agricultural land area. The model thus simulates the future need for agricultural land. The model's demand and supply scenarios for agricultural products and land are generated primarily by the growth rates of population and GDP, and by changes in land productivity. The model is comprehensive, but does not have excessive data requirements. Its output in in the form of tables and graphs; it is thus a numerical model. The model uses three different agricultural classes:
De Gier and Hussin (1993) linked a spatial component to the numerical : APM. Recently an GIS-based version of the APM was developed, where its .main components were integrated into the Integrated Land and Water Information System, ILWIS, a GIS with built-in processing capabilities for digital remote sensing images. It was developed by the ITC in The Netherlands. Hussin and Bode (1995) describe the structure of the GIS based version. This paper describes the experiences of the GIS-based version of the APM in the Kali Konto area in East Java, Indonesia. In the Kali Konto area strict law enforcement prevents the state-owned forest land to be converted into agricultural land. The remaining land is already brought under agriculture or is used as housing land. The growing population is therefore facing a severe constraint to satisfy .its demand for food and other basic needs, since agricultural expansion is not possible. It is felt that a growing number of land less people in particular, engage in illegal felling of trees and sell them for fuelwood and timber. The proceeds are then used to purchase food and other goods. As a result, the forest is then more and more degraded, finally reaching a state of scrub. Because the actual development of scrub is known from multitemporal data over the period 1979-1993, the overall aim of the research project is to identify the level to which the APM can be enhanced for properly simulating the ongoing forest degradation. Under the assumption that the landless people are .mainly responsible for forest degradation, the class "land for subsistence crops" is determining in the APM. It follows that population figures and their rates of change are relevant, and not GDP. This paper describes the results of applying the recently developed GIS-based Area Production Model to data of the Kali Konto area over the period 1979-1993, and compare the outcome with the real observations of scrub development over the same period. Method In an earlier paper, De Gier et a] (1995) describe the performance of the spatial component that was linked to the APM, using multitemporal data for the period 1979-1993. The population growth was assumed to be the same for the entire area. One of the conclusions was that although the spatial component performs well in general, more precision in the specific locations of scrub development was desirable. For example, it was noted that scrub development did not develop uniformly along the fringes of the forest. It was hypothesized that these locations were not only dependent on the factors mentioned earlier (population, GDP,crop production, and their rates of change), but also on specific locational factors. In the GIS-based model the so-called village land (ie the non-forest land) of Kali Konto, was first subdivided into individual villages, each one with its known area, population and population growth. In addition, the GIS-based model included six spatial factors: Slope percent, Distance from the village land, Priority of land transfer, population density, population growth. Each pixel was accordingly labeled. The underlying assumptions are that a larger slope percent means a slower forest degradation and a higher the friction value; a greater distance results in a slower the forest degradation and a ghigher the friction value; a higher priority value (ie low priority!) means a slower forest degradation and a higher friction value; a higher population growth results in a faster forest degradation and higher starter value; a higher population density means a faster , forest degradation. It was assumed that one or more of these factors , would explain most of the variation of the scrub land development. Ct Contrary to the numerical APM, it was further assumed that different types of forest, such as plantation forest, protection forest, etc. could be degraded at the same time, as is normally observed. In order to avoid unreasonably large values for the factor slope*distance, these values were reduced to within the range 1-10. The demand for future agricultural (subsistence) land is calculated according to: where. Na demand for agricultural land in year n Pa present amount of agricultural land p growth percent of the population The equivalent amount of degraded forest land or scrub (Ns) at year n is calculated according to (Hargyono, 1993) : where Na is defined as above. Results To verify the outcome of the GIS-based APM, first a numerical com- parison was made with the observed data from 1979 to 1993. Table 1 shows that the differences were 5% or less.
When comparing similar values for a period up to 50 years, differences do no exceed 7% .Unfortunately, no comparison was made for all years of whJ.ch observed values existed. They were: 5757 ha (1979) , 6062 ha (1982), 6275 ha (1984) and 6991 ha (1993) .The data suggest that the ; GIS-based APM performs better than the numerical APM. : Next, the spatial data were inserted: a raster map of the villages was available. The population and population growth data for each village were obtained from Nibbering (1980) and Hargyono (1993) .The year 1979 was the starting year. A detailed land cover map, based on an inter pretation of panchromatic aerial photographs of 1979, scale 1:20000,was available. A second land cover map, also based on airphoto inter pretation, and using the same criteria as used for the 1979 map, was available for 1984. A map was constructed, showing the scrub develop ments that took place between 1979 and 1984. Applying the GIS-based APM on the data, and comparing this with the number and location of the scrub pixels as they developed in reality from 1979 to 1984, revealed that: forest plantations should be considered as absolute barriers (ie they hardly degrade), and new scrub .develops from existing scrub land. Altogether, only 36% of the scrub pixels were spatially predicted correctly; numerically, as much as 77% of the pixels was classified correctly, highlighting the difficulties ,. of spatial accuracy. A further analysis was made with regards to which factor was the most determining one in scrub development. These factors were: Slope percent, population Density, population Growth, PriorityTransfer (ofland), and the combined factors Slope Distance (= slope *distance) and VillageFactor (population densi ty*population growth) .Two weights were .used (0.1 and 10) he analysis refers to numerically and spatially correct number of pixels. In decreasing order of importance, the following sequence of factors emerged: Village Factor (36%), PriorityTransfer (31%), Slope Distance (31%), Slope (30%), population Growth (29%) and population Density (29%) .The percentages refer to the percentage of correctly classified pixels for the factor concerned, when setting the other factors at default. It was also found that a low weighing of the factors, ie 0.1 gave better results. Commentary The study showed that better results were obtained when plantation forest was considered an absolute barrier, ie not degradable. However, field observations and multitemporal airphoto interpretation show that change into scrub can take place. It is not known, however, whether this is due to failed plantations, or to deliberate actions by people .Most of the forest plantations are located just inside the forest boundary, to serve as a barrier for forest penetration. Also, they are generally well protected by forest guards. The numerical result of the GIS-based APM is satisfactory (77% correct), but the spatial accuracy is low, not exceeding 36%. To further increase this percentage, a further reduction of weighing should be tested, especially to the factors village Factor and Slope Distance .Taking into consideration the location of roads and especially the tracks might further improve the model, because it is along such ways of access that people enter ..the forest, and transport the forest products. Many tracks can be seen on the 1:15000 and 1:20000 scale aerial photographs, but not on the current generation of satellite images. It is to be remarked that up to 1993 no cloud free SPOT image was available. Finally, the period of verification was from 1979 to 1984. Using a longer time period, eg from 1979 to 1993 for which verification data are available, might remove short-time variations, and yield a more accurate result. References
|