Testing on the time-robustness of a landslide prediction model
Testing on the time-robustness of a landslide prediction model
复制标题
滑坡预测模型的时间鲁棒性测试
DOI：
--
复制
发表时间：
2001
期刊：
影响因子：
--
通讯作者：
C. Chung
中科院分区：
文献类型：
--
作者：
H. Kojima;C. Chung
基于输入的空间地图数据(称为“因果因子”)和过去滑坡之间的定量统计关系，构建了一个预测模型，用于识别未来可能受滑坡影响的区域。作为预测模型的有效工具，模型的时间稳健性是必不可少的组成部分。通过对模型的稳定性分析，检验了模型的时稳性。在这一贡献中，使用了使用代数和算子的模糊集合论方法来构建预测模型。哥伦比亚的里约热内卢·钦西纳被选为研究区。稳定性分析的分析步骤如下：步骤1.将以往滑坡的发生划分为两个不同的时间段。第一组包括在给定年份之前或之内发生的山体滑坡，第二组包括在该年之后发生的其余山体滑坡。步骤2.分别利用第1组和第2组的滑坡得到两个预测图，分别为Map-1和Map-2。步骤3.为了验证基于第一组滑坡的预测地图(Map-1)的预测性能，我们将Map-1的预测结果与第二组的滑坡预测结果进行了比较。比较产生了统计数据，称为“预测率”，表明了Map-1的预测能力。步骤4.为了评价Map-1和Map-2的稳定性，制作了差值图(DIF-MAP)，并计算了相应的匹配率。更高的匹配率意味着Map-1和Map-2的相似性增加，从而增加了两个Map的稳定性。根据本研究的结果，我们得出结论：模型的预测能力是由步骤3中计算的预测率来表示的。一个模型要成为一个好的预测工具，模型应该具有良好的预测能力。在步骤4中评估的基于两个时间段的两个预测图的稳定性也是一个好的预测工具的重要组成部分。特别是，如果我们要将预测地图用于山泥倾泻预防计划或土地利用规划研究，那么稳定性研究将为规划决策提供关键信息。利用模糊集合论方法建立的里约热内卢研究区预测模型确实具有较好的预测能力，但不够稳定，不能作为一个好的模型。该模型还需要进一步研究，以提高预测模型的性能。一个有效的预测模型，既需要模型的预测能力，也需要模型的稳定性。
A prediction model for identifying areas likely affected by future landslides are constructed based on the quantitative statistical relationships between the input spatial map data (termed “causal factor”) and the past landslides. The prediction model to be effective tool, the time-robustness of the model is an essential component. We test the timerobustness using the stability analysis of the model. In this contribution, a Fuzzy set theory procedure using algebraic sum operator is used for the construction of the prediction model. Rio Chincina in Colombia is selected as the study area. The analytical procedures for the stability analysis are as follows: Step 1. The occurrences of the past landslide were divided into two different time-periods. Group 1 consisted of the landslides occurred prior to or within a given year and Group 2 contained the remaining landslides, which occurred after the year. Step 2. Using the landslides in Groups 1 and 2 separately, two prediction maps, Map-1 and Map-2, respectively, were obtained. Step 3. To validate the prediction performance of the prediction map (Map-1) based on the landslides in Group 1, we compared the prediction results in Map-1 and the landslides in Group 2. The comparison produced statistics, termed "prediction rates" indicating the prediction power of Map-1. Similarly, we obtained the prediction rates for Map-2 using the landslides in Group 1. Step 4. To assess the stability of Map-1 and Map-2, the "difference map (DIF-map)" was made and the corresponding "match rate" was also computed. Higher match rate means increasing similarity of Map-1 and Map-2 and hence increasing stability of the two maps. Based on the results of this study, we concluded: The prediction power of the model is represented by the prediction rates computed in Step 3. A model to be a good prediction tool, the model should have a good prediction power. The stability of two prediction maps based on two time-periods assessed in Step 4 is also an essential component of a good prediction tool. In particular, if we were to use the prediction map for the landslide prevention plans or land use-planning study, then the stability study would provide pivotal information on the planning decision. The prediction model for Rio Chincina study area using the Fuzzy set theory procedure did have reasonably good prediction power but was not stable enough to be a good model. The model needs a further investigation to improve the performance of the prediction model. A prediction model to be effective, we need both the prediction power and the stability of the model.