论文部分内容阅读
0cm土壤温度是冻土模型的上边界条件,连续的、高质量的青藏高原0cm土壤温度数据是进行准确冻土模拟的必要条件.然而受复杂下垫面的影响,遥感手段无法获取可靠的0cm土壤温度.利用自适应网络模糊推理系统(ANFIS)结合青藏高原实测资料建立遥感地表温度产品(LST)与0cm土壤温度的关系,以实现通过LST估算青藏高原逐日0cm土壤温度.研究了ANFIS的各种参数组合,发现筛选合适的小波函数、小波窗口、小波层数建立起来的Wavelet-ANFIS模型能较准确实现估算0cm土壤温度的目的.验证表明,估算结果与气象站点实测0cm土壤温度绝对误差在2K以下,相关系数0.98以上.考虑到原始MODIS LST误差在0～2K之间,该方法可以获取较为理想的0cm土壤温度,为冻土模型提供准确的上边界输入.
0cm soil temperature is the upper boundary condition of permafrost model, continuous and high-quality 0cm soil temperature data of Qinghai-Tibet Plateau is a necessary condition for accurate permafrost simulation.However, affected by complex underlying surface, remote sensing can not obtain reliable 0cm Soil temperature.The relationship between remote sensing surface temperature product (LST) and soil temperature 0 cm was established by using adaptive network fuzzy inference system (ANFIS) and the observed data of Qinghai-Tibet Plateau so as to realize daily daily 0 cm soil temperature by LST.Analyses of each ANFIS Wavelet-ANFIS model established by screening appropriate wavelet function, wavelet window and wavelet layer can achieve the goal of estimating 0cm soil temperature accurately.Experimental results show that the absolute error between 0cm soil temperature and the estimated results is in the range of Under 2K, the correlation coefficient is above 0.98. Considering that the LST error of the original MODIS is between 0 ~ 2K, this method can obtain the ideal 0cm soil temperature and provide the accurate upper boundary input for the frozen soil model.