The Study Of Scale Sensitivity In Land Use Change CA Modeling

Accurately mastering land use data and land use change could not only provide reliable proof for the land management department to make overall planning of land utilization, and utilize and develop land resources reasonably; but also provide scientific basis for the national economy and regional planning. Cellular automata is frequently used for modeling geographical changes, since it has the formidable spatial and temporal dynamic simulation capabilities. It can simulate complex urban development pattern through simple conversion rules; thus it is a powerful tool for modeling complex system, such as land use changes.In practical applications, it is an important issue for improving simulation accuracy of CA model. According to the literatures and existing research, the study of improving simulation accuracy of CA model focus on exploring conversion rules, it lacks disquisition on spatio-temporal scale and the sensitivity of the CA model. The effect of spatial and temporal scale on the model outcome has not been sufficiently addressed. As scale can determine the level of detail and imply the nature of local interaction among cells, it is the first stage to determine specific scale for CA modeling of land use change. The literatures show that the scale usually be defined through two ways:one is by random or on the basis of study data; the other is on the basis of experiments. Choosing scale by random is clearly not scientific. However, the approach basing on experiments cost too much time, and the results lack of generality. Consequently, it is necessary and urgent to study on the scale sensitivity of CA modeling.In this paper, a total of 125 simulation experiments were executed basing on CA_Markov model using IDRISI software, and deeply study on the relationship between the spato-temporal scale index such as cell size, neighborhood range, and time interval; and the impact of sensitivity on simulation accuracy. The results showed that:(1) there is some sensitivity of temporal-spatial scales on CA model, and the sensitivity is different under different combinations of spatial and temporal scales; (2) spatial and temporal scales are linked, and different combinations will result in different simulation accuracy; (3) according to different requirements, the best combination of scales vary for land use change in Changping District. In order to obtain the most accurate results, the cell size should be 250 square meters, neighbors range should be 5 x 5, the time interval should be 1 month; to obtain both high efficiency and precision results, cellular size should be 250 square meters, neighbors range should be 5 x 5, the time interval should be 12 months for the optimal scale combination. The scale selection method proposed by this paper is universal, and the optimal combination of parameters can be used as reference for similar studies.