Estimated sea ice thickness and concentration for the Weddell Sea during autumn and austral winter in 1992 to 2009 using passive microwave SSM/I (Special Sensor Microwave Imager) data

We estimate the sea ice thickness for the Weddell Sea during autumn and austral winter using passive microwave SSM/I (Special Sensor Microwave Imager) data. First, we mapped pixels with a sea ice concentration above 90% from fraction images calculated from the spectral linear mixing model using the brightness temperature in the channels and polarizations 19H, 19V, and 37V. Multiple linear regression was performed from 37V, 85V, and GR3719 data (Gradient Ratio 37V and 19V) using sea ice thickness local observations produced by the Antarctic Sea Ice Processes and Climate (ASPeCt) program. We compared our model validation with Cloud, and Land Elevation Satellite 1 (ICESat-1) satellite data. We were considering p < 0.05, the regression model shows an R² = 0.62 limited to sea ice up to 1.5 m thick and, in the validation, we identified a bias between the ICESat-1 (samples with a thickness < 1.5 m) and the SSM/I ranging between 0.02 m to 0.19 m and Root Mean Square Deviation ranging from 0.13 m to 0.46 m.本研究利用被动微波SSM/I（专用传感器微波成像仪，Special Sensor Microwave Imager）数据，估算了威德尔海秋季及南半球冬季的海冰厚度。首先，我们借助19H、19V与37V通道及极化方式下的亮温数据，通过光谱线性混合模型计算得到组分分数图像，从中筛选出海冰浓度高于90%的像素点。随后，我们依托南极海冰过程与气候（Antarctic Sea Ice Processes and Climate, ASPeCt）计划获取的局地海冰厚度观测数据，以37V、85V以及GR3719（37V与19V的梯度比）数据为自变量构建多元线性回归模型。我们将模型验证结果与云与陆地高程卫星1号（Cloud and Land Elevation Satellite 1, ICESat-1）的卫星观测数据进行了对比。在显著性水平p<0.05的条件下，仅针对厚度不超过1.5米的海冰，该回归模型的决定系数R²为0.62；验证结果显示，ICESat-1数据（对应厚度小于1.5米的样本）与SSM/I数据之间的偏差介于0.02米至0.19米之间，均方根误差则介于0.13米至0.46米之间。