基于马尔柯夫模型的草原退化动态时空特征研究

草原植被覆盖动态时空特征是理解人类活动和自然因素影响下草原退化的关键。本研究利用2000和2010年TM影像,利用监督分类方法,对内蒙古草原退化、沙化和盐渍化时空分布特征探测和制图,同时对分类结果进行验证。检验结果表明,采用监督分类结合地面训练样本进行草原退化动态特征检测能提高分类精度,草原沙化和盐渍化分类精度分别都达到90%以上,退化草原分类精度为75%。在草原退化程度分类检测基础上,计算2000-2010年草原退化、沙化、盐渍化面积概率转移矩阵,并基于马尔柯夫模型,对未来20年间草原动态特征进行预测。研究结果表明,十年间,内蒙古草原的退化、沙化程度均呈减弱趋势。预测结果显示,在草原与生态保护建设工程持续稳定有效建设前提下,内蒙古草原呈现良性发展趋势,退化及沙化状况将持续得到改善。研究也指出,基于遥感技术与马尔柯夫模型有机结合分析草原退化动态特征是一种快速有效的途径。

A study on spatial-temporal characteristics of grassland degradation using the Markov model

The spatial and temporal features of grassland cover conversion (GCC) serve as a useful input for understanding the desertification process and degradation of grassland caused by anthropogenic activities and extreme natural events in general. Thematic Mapper data (TM 30 m) were used to detect and map degraded grassland features both spatially and temporally. Two data sets of TM 30 m data were collected from the years 2000 to 2010. Supervised classifications were developed for each of the GCC change detection of the three cases (degradation,desertification, and salinization). To address this situation, the field data were used to test the GCC detection of change results presented in this paper. The GCC change detection methods worked reasonably well and detection accuracy of deserted and salinized output was >90% although degraded output identified only 75% of the covered pixels within the ground observed perimeter polygons. The applications presented in this paper also evaluated the transition matrix between 2000 and 2010 of each of the three change detections,and predicted dynamic characteristics of grassland using the Markov model. The results showed that for the next decade, and even for a further ten years, the grassland will develop positively with a reduced trend of degradation and desertification. The research also indicated, it is credible to use remote sensing technology combined with the Markov model in analyzing the dynamic characteristics of grassland cover changes.