3种土壤冻融判别算法在青藏高原的分类精度评价

本文比较了基于AMSR-E被动微波数据的3种土壤冻融判别算法在青藏高原相关地区的分类精度。3种算法分别是：双指标算法、决策树算法、判别函数算法。本文选取了来自青藏高原那曲、玛曲、阿里3个地区土壤温湿度观测网的地表温度数据,并结合AMSR-E被动亮温数据,对上述算法在以上地区的分类精度分别进行了比较评价。结果表明：不论是白天还是夜间,相较于干旱区微波信号来自深层土壤的难以准确探测,在青藏高原半湿润半干旱区算法可取得相对较好的判别准确率;双指标算法相较于其他2种算法,在观测区具有较高的分类精度,且夜间分类精度高于白天;实测数据存在资料代表性不普遍即网格所包含站点信息量不够的问题,这也是后续工作中提高分类精度值得关注的着手点。

Evaluation of Classification Accuracy in Tibetan Plateau of Three Soil Freeze/Thaw Discrimination Algorithms

The thesis paper is based on AMSR-E passive microwave brightness temperature data, chosen three soil surface freeze-thaw discrimination algorithms and compares the soil freeze-thaw status classification accuracy of the ground surface on Tibetan plateau area seperatly. Three algorithms are consisted of Dual Index Algorithm, Decision Tree Algorithm and Discriminant Function Algorithm. This thesis collects the soil surface temperature data from three soil temperature and humidity observational networks on Tibetan plateau that includes Naqu, Maqu and Ngari district, and combines with AMSR-E passive microwave brightness temperature data, then separately compares and evaluates the classification accuracy of the mentioned algorithms in the previous three observational areas. The results illustrate that whether during daytime or nighttime, in arid area like Ngari network district the microwave signal comes from deep layer soil, on the one hand, it’s kind of hard to probe in aid area and get an accurate detection as well, on the other hand, for all three discriminat algorithms that leads to the low discriminant accuracy in arid area at the same time. On the contrary each algorithm has relatively high discriminant accuracy in semi-arid area like Naqu network and subhumid area like Maqu network on Tibetan plateau; Dual Index Algorithm, compared with two other algorithm, has the highest classification accuracy among three discriminat algorithms, and has the best entirety classification accuracy in three observational networks on Tibetan plateau as well. The classificition accuracy at nighttime is higher than the dayttime due to the spatial heterogeneity of the air temperature between day and night; Except that the actual measured data also has the problem of incomprehensive representativeness, which means the inadequacy of information from all sites in the networks, and this is what needs to pay attention in the follow-up work to improve the algorithm classification accuracy.