论文部分内容阅读
对滑坡危险度制图进行精度评价具有重要意义。现有对滑坡危险度制图进行评价的研究多集中于评价指标的建立和应用,而评价指标能否有效地度量滑坡危险度图的精度、区分不同质量的滑坡危险度图尚值得研究。针对只使用正样本的评价指标,选择其中3种常用的非阈值依赖性评价指标——预报效率曲线及曲线下面积、P/E曲线及连续Boyce指数和空间聚集度作为精度评价标准,以甘肃省油坊沟流域为研究区,采用正样本和不同的负样本,基于随机森林模型推测生成不同质量的滑坡危险度图,应用3个评价指标对其进行精度评价,验证了各评价指标的有效性。结果发现,预报效率曲线与P/E曲线的分布可以体现不同质量的滑坡危险度图的差异,但是预报效率曲线下的面积只适宜于滑坡危险度高估时的制图精度评价,而连续Boyce指数与空间聚集度只适宜于滑坡危险度低估时的制图精度评价。
It is significant to evaluate the accuracy of landslide hazard mapping. The existing researches on the landslide hazard mapping mostly focus on the establishment and application of the evaluation indexes. However, whether the evaluation indexes can effectively measure the landslide hazard map and distinguish the landslide hazard maps with different qualities is still worth studying. According to the evaluation index which only uses the positive sample, three commonly used non-threshold-dependent evaluation indexes-prediction efficiency curve and area under the curve, P / E curve, continuous Boyce index and spatial aggregation degree were selected as the accuracy evaluation criteria. Using the positive samples and the negative samples, the landslide hazard map of different quality was estimated based on the random forest model, and the accuracy of the landslide hazard assessment was evaluated by using three evaluation indexes to validate the validity of each evaluation index . The results show that the distribution of forecast efficiency curve and P / E curve can reflect the difference of landslide risk map of different quality, but the area under the forecast efficiency curve is only suitable for the mapping accuracy evaluation when the landslide risk is overestimated, while the continuous Boyce index And the degree of spatial aggregation is only suitable for the accuracy of the mapping when the risk of landslide is underestimated.