White Evan, Shephard Mark W, Cady-Pereira Karen E, Kharol Shailesh K, Ford Sean, Dammers Enrico, Chow Evan, Thiessen Nikolai, Tobin David, Quinn Greg, O'Brien Jason, Bash Jesse
Environment and Climate Change Canada, Toronto, ON M3H 5T4, Canada.
Faculty of Science, University of Waterloo, Waterloo, ON N2L 3G1, Canada.
Remote Sens (Basel). 2023 May 17;15(10):2610. doi: 10.3390/rs15102610.
Presented is a methodology to explicitly identify and account for cloud-free satellite measurements below a sensor's measurement detection level. These low signals can often be found in satellite observations of minor atmospheric species with weak spectral signals (e.g., ammonia (NH)). Not accounting for these non-detects can high-bias averaged measurements in locations that exhibit conditions below the detection limit of the sensor. The approach taken here is to utilize the information content from the satellite signal to explicitly identify non-detects and then account for them with a consistent approach. The methodology is applied to the CrIS Fast Physical Retrieval (CFPR) ammonia product and results in a more realistic averaged dataset under conditions where there are a significant number of non-detects. These results show that in larger emission source regions (i.e., surface values > 7.5 ppbv) the non-detects occur less than 5% of the time and have a relatively small impact (decreases by less than 5%) on the gridded averaged values (e.g., annual ammonia source regions). However, in regions that have low ammonia concentration amounts (i.e., surface values < 1 ppbv) the fraction of non-detects can be greater than 70%, and accounting for these values can decrease annual gridded averaged values by over 50% and make the distributions closer to what is expected based on surface station observations.
本文提出了一种方法，用于明确识别并考虑传感器测量检测水平以下的无云卫星测量数据。这些低信号通常出现在对光谱信号较弱的次要大气成分（如氨（NH））的卫星观测中。在传感器检测极限以下的条件下，如果不考虑这些未检测到的数据，会导致平均测量值出现高偏差。这里采用的方法是利用卫星信号中的信息内容来明确识别未检测到的数据，然后用一致的方法对其进行处理。该方法应用于CrIS快速物理反演（CFPR）氨产品，在存在大量未检测到数据的条件下，得到了更符合实际的平均数据集。这些结果表明，在较大的排放源区域（即地面值>7.5 ppbv），未检测到数据的情况发生频率低于5%，对网格化平均值（如年度氨源区域）的影响相对较小（降低幅度小于5%）。然而，在氨浓度较低的区域（即地面值<1 ppbv），未检测到数据的比例可能大于70%，考虑这些值会使年度网格化平均值降低超过50%，并使分布更接近基于地面站观测预期的结果。
