基于CA-Markov和Geomod模型橡胶林变化预测比较

Comparison of CA- Markov and Geomod Models for Rubber Plantation Prediction

近年来,学者们发展了一系列的模型对土地利用/覆盖变化进行预测模拟研究,CA-Markov和Geomod模型是其中应用较为广泛的两种模型。分别利用CA-Markov和Geomod模型模拟云南景洪地区2010年橡胶林LUCC状况,旨在通过两种模型的模拟对比,探讨两种模型的差异和优劣,选取一种更加合理、可靠的模型来对该地区橡胶林变化趋势进行预测分析。利用1995年、2003年、2010年3个时期的影像进行解译,将1995年和2003年的解译结果利用两种模型预测2010年景洪地区内橡胶林分布格局,并且运用2010年ALOS影像解译的橡胶林空间分布结果图来检验模型的预测模拟结果。研究发现:1.两个模型模拟结果人工目视检验差异明显,Geomod模型的预测模拟结果与影像分类的结果更加接近;2.两个模型ROC分析精度较高,都大于0.65,且AUC值基本一致;3.利用模糊矩阵精度检验结果表明,Geomod模型从栅格象元的数量、匹配的准确性以及总体精度高于CA-Markov模型的预测模拟结果。

In recent years,researchers have developed a series of models to predict the land use / cover change where the Geomod and CA- Markov model is one of the two models which is widely used. In the paper CA- Markov and Geomod models are used to simulate the LUCC of rubber plantation in 2010,Jinghong of Yunnan,China comparing the results of simulation of two models,aims to explore the differences and pros / cons of two models,and select the better one which is more reasonable and reliable to analyze and predict the LUCC situation in the region.Using the interpreted results of image of TM in 1995 and 2003 to simulate the distribution of rubber in the region,2010 and validate the performance of two models by using the interpreted results of image of ALOS in 2010. The results showed that： 1. the differences are significance in visual inspection between the two models,result of simulation of Geomod matches the true map better than another; 2. both of two models which AUC value 0. 65 have high precision which is basically the same in ROC analyze; 3. Geomod model performances better than CA- Markov model on the number of pixels,matching accuracy,overall accuracy in the validation of fuzzy matrix.