资源环境科技发展态势分析平台

Detecting trends in climate variables has become of great importance in the context of the rapid evolution of the Earth's climate. Among the climate variables recognized as acting on climate variability, aerosols continue to contribute one of the largest uncertainties to the total radiative forcing estimate. The present study focuses on dust aerosols, a major contributor to total aerosol loading. Observation from space offers a good opportunity to follow aerosol evolution at global scale and over long time series. In this context, infrared observations, by allowing retrieving dust optical depth (AOD) day-time and night-time, over oceans and over continents, in particular over desert, is complementary to observations in the visible on which a majority of aerosol studies are so far based. The two spectral domains are not sensitive to the same range of particle size, coarse mode particles ( > 1 m) being preferentially observed in the infrared, when both coarse and fine mode particles (0.1-1 mu m) are observed in the visible making the distinction between the two modes difficult. Starting from METOP-A/IASI-derived dust AOD (July 2007 to June 2017), this study aims at detecting and estimating, day-time and night-time, dust AOD trends over Sahara, a region where dust aerosol emissions are frequent and often intense. Detecting trends in geophysical variable time series is a difficult task due to the frequent presence of serial correlation, to holes in the data, to "outliers", or to variables not normally distributed, etc. Here, trends are determined using the non-parametric Theil-Sen slope estimator followed by the Mann-Kendall statistical test for randomness against trend, the statistical hypothesis (H-0) being that the trend is equal to zero. Resulting trends are judged to be significantly different from zero if their so-called "confidence level" is above a given percentage, usually 95% (referred to as "real" trend in the following). Then, in a second step, supposing the statistical hypothesis is now that a non-zero trend of specified magnitude exists, we can calculate the probability that this hypothesis is true. The number of years of data required to detect a "real" trend of a specified magnitude with probability 0.9 ("probability-assigned real" trend) is finally determined following (Tiao et al., 1990). This approach is then applied to the 120 months of IASI-retrieved dust AOD. Main conclusions are that the present 10 year period of IASI observations is still too limited to provide enough reliable, probability-assigned, trends over Sahara and that a period of 14-15 years would largely increase their number and spatial coverage. This is fully compatible with the already planned length of the IASI series on board METOP-A to -C, taking into account the high demonstrated stability of this instrument.