论文部分内容阅读
土壤水分是陆面生态系统和能量循环的核心变量之一,利用微波遥感技术获得的土壤水分产品的时间分辨率一般是2-3 d,因此精确地获得具有较高时间分辨率的土壤水分成了人们关注的焦点。本文尝试将SMAP(the Soil Moisture Passive and Active)土壤水分和MODIS光学数据相结合,利用广义回归神经网络进行全球36 km土壤水分的估算,提升SMAP土壤水分的时间分辨率。结果显示,广义回归神经网络估算土壤水分与SMAP保持了高相关性(r=0.7528),但其却保留了较高的误差(rmse=0.0914 m3/m3)。尽管如此,估算的土壤水分能够很好地保持SMAP土壤水分的整体空间变化,并且提升了土壤水分的时间分辨率(1 d)。此处,本文研究了SMAP土壤水分与MODIS光学数据之间的关系,这对今后利用机器学习进行SMAP土壤水分降尺度研究提供了重要的参考价值。
Soil moisture is one of the core variables of terrestrial ecosystem and energy cycle. The temporal resolution of soil moisture products obtained by microwave remote sensing technology is generally 2-3 d, so soil moisture with high temporal resolution is accurately obtained The focus of people’s attention. This paper attempts to combine the soil moisture of SMAP (Soil Moisture Passive and Active) with MODIS optical data to estimate the global 36 km soil moisture using generalized regression neural network and improve the temporal resolution of soil moisture in SMAP. The results showed that the generalized regression neural network estimated that the soil moisture was highly correlated with SMAP (r = 0.7528), but retained a higher error (rmse = 0.0914 m3 / m3). Nevertheless, the estimated soil moisture can well maintain the overall spatial variation of soil moisture in SMAP and enhance the temporal resolution of soil moisture (1 d). Here, the relationship between SMAP soil moisture and MODIS optical data is studied in this paper, which provides an important reference value for SMAP soil moisture reduction using machine learning in the future.