Multi-scale Modeling Of Land Use Change A Case Study In Maotiao River Basin, Guizhou

Abstract: Maotiao River Basin in Guizhou province is a typical karst area in southwestern China. In this dissertation, we simulated the land use changes of this area at different scales and with different subject combining top-down and bottom-up approaches, aiming at constructing a multi-scale framework of land use changes. Results and conclusions are stated below.Farm households are the smallest subject and basic unit that influence land use change. On the other hand, land use change observed on landscapes is a reflection of aggregated land use decisions at the household level. Thus, studying the characteristics of farm households'decision-making will improve our understanding on the basic mechanism of land use change. In the bottom-up approaches, we started with analysis at a farm household level, examining thoroughly the economic status and land use decision making of individual household. Based on this information, the spatial distribution of the households was modeled using downscaling analysis of population density, while the population decision making was deduced from individual information through household classification. At last, the agent-based modeling was used to simulate the land use change as a product of farm households' reclamation decision making.1) Farm household analysis at small scale. Based on the survey carried out in this region, we analyzed the net income and expenditure of the farm households and the relationship of income and household properties. Using multiple regression analysis, cropland area, off-farm working experience and number of off-farm labor were identified as the main factors influencing households' incomes. Next, we studied the household's decision-making and its relationship with farm household properties. In particular, correlation analysis was used on the reclamation decision-making to determine the primary contributors. Finally, farm households' perceptions and responds to policy were explored using binary Logistic regression. The factors affect farm households' decisions on cropland conversion include household income per person, percentage of cropland conversion compensation in household's net income, cropland area per person and the percentage of converted cropland in household's total cropland area.2) Down-scaling analysis of population density. Through analyzing main spatial influencing factors of population density distribution, we converted the village-scale population data to a smaller grid unit scale, and generated the population density distribution map in the basin based on multiple regression analysis. The results showed that, at small basin scale in southwestern China karst area, the primary factors influenced population density distribution were build-up land index, cropland index and distance from the road. The population density map generated from the multiple regression modeling was effective representation of the basin population.3) Household classification and spatial distribution modeling. There are both diversity and similarity in farm household decision making. The similarity is presented as group decision making. To model the influence of household decision making on regional land use, we need to divide individuals into groups according to their internal factors that influence the decision-making. Based on the results of farm households analysis, we classified the households into five types in accordance with reclamation decision-making and characterized the properties of each type, realized the conversion of individual farm households to population decision-making. Then the five types of households were distributed to the basin map according to the spatial distribution algorithm generated from the population density distribution and other spatial information. The modeling result is highly comparable to the actual distribution map, no matter on the quantity or spatial pattern. This work further supported the ABM modeling in the next step.4) Agent-based modeling of land use change. We applied the multi-agent modeling concept into our study, simulated farm households' possible reclamation behaviors after ending the cropland-to-forest/grass conversion policy at the basin scale. The results showed that, the possible reclamation behavior would have a great impact on the land use patterns. The reclamation decisions were variable among different household types. We thus designed a policy scenario for the households with larger reclamation area and simulated the land use change under the scenario, proving that the appropriate land use policy plays an important role in land use change.Regional development and object is the primary concern of government who is the stakeholder at this scale. In the top-down approaches, we used ANN-CA (artificial neural network and cellular automata), SOFM (Self -Organizing Feature Maps) and MOLP (multi-objective linear programming) model, from the aspect of land use pattern, land use regionalization and optimal allocation of land use structure, comprehensively modeled the land use change at the basin scale, in order to provide scientific basis for sustainable development and land use management.1) ANN-CA model based land use change modeling. In this part, we used ANN-CA coupled model, based on the modeling of historical land use pattern change, to predict the land use pattern in 2014. The model precisely simulated the land use change and the predicted result is consistent with historical development. The ecological environment is continuing becoming better, with 109.89km~2 decrease of cropland, 22.23km~2decrease of grass and 103.23km~2 increase of forest.2) SOFM model based land use regionalization. By using SOFM model, 45 villages and towns in the basin was classified and then regionalized according to the land use regionalization assessment index based on the natural ecology and socio-economy properties. The result was comparable with the one generated from traditional cluster analysis, while the SOFM model was more objective and applicable. We divided the basin into 3 regions, and gave out the direction and suggestions of land use development to each region respectively.3) MOLP model based optimal allocation of land use structure. We used economic benefits, ecological benefits and integrated ecological-economic benefits as the optimization goals, setting 11 constraints according to the facts of the study area and got the optimized land use quantitative structure with MOLP model. The results showed that the economy and ecological benefits should be taken into consideration comprehensively, in order to achieve an optimal land use structure. As each area is different in many aspects, the policy should be made according to individual properties to meet the need of sustainable development in different areas.