Full-Physics Inverse Learning Machine for satellite remote sensing of ozone profile shapes and tropospheric columns

Kurzfassung

Ozone plays a crucial role in the Earth’s atmosphere and its chemical processes (production and destruction) are related to climate change and air pollution caused by anthropogenic emissions. Therefore, accurate information about global/regional vertical distributions of ozone in the troposphere and stratosphere turns out to be important to scientific communities. Spaceborne remote sensing of ozone information using the ultraviolet (UV) radiation has been comparatively mature, total columns have been successfully estimate by the Differential Optical Absorption Spectroscopy (DOAS) algorithm. However, characterizing ozone profile shapes from nadir-viewing satellite measurements is still known to be challenging, particularly the ozone information in the troposphere. As most atmospheric ozone resides in the stratosphere above, tropospheric ozone columns can be derived by subtracting an estimate of the stratospheric columns or by differencing total columns in cloud-free pixels from those in nearby pixels with thick/high convective clouds (the so-called CCD method). Retrieval of tropospheric ozone abundances can largely benefit in terms of representativity by obtaining reliable an ozone profile shape. Direct retrieval of tropospheric information has also been exploited and applied to the Global Ozone Monitoring Experiment (GOME) class of instruments. In general, the direct estimation of atmospheric parameters of interest from spectral measurements is treated as an ill-posed inverse problem that often requires an iterative inversion of large matrices and multiple calls to radiative transfer calculations. However, this classical inversion method is computationally expensive and reliable a priori knowledge can be vital to the retrieval outcome. Alternatively, the above-mentioned inverse problems can be solved by means of machine learning. Therefore, we propose a novel retrieval algorithm called Full-Physics Inverse Learning Machine (FP-ILM) and estimate ozone profile shapes from GOME-2 measurements on the MetOp series of satellites. Unlike traditional inversion methods, the FP-ILM algorithm formulates the retrieval of ozone profile shapes as a classification problem. The implementation of FP-ILM comprises a training phase to derive an inverse function from synthetic measurements using a radiative transfer model in conjunction with the "smart sampling" approach, and an operational phase in which the inverse function is applied to real measurements. In particular, the employed forward model symbolizes the “full-physics” feature. With the aid of machine learning techniques, FP-ILM has been proven to produce a noticeable increase in computational speed and a reasonable retrieval accuracy when dealing with satellite measurements. The comparison of retrieved ozone profiles between FP-ILM and the optimal estimation method reaches a promising agreement. This paper extends the ability of the FP-ILM retrieval to derive tropospheric ozone columns from GOME-2 measurements. Results of total and tropical tropospheric ozone columns are compared with the ones using the DOAS and CCD methods, respectively, and further comparisons of ozone profiles are conducted with ozonesonde observations. Furthermore, the FP-ILM framework will be used for the near-real-time processing of the new European Sentinel sensors with their unprecedented spectral and spatial resolution and corresponding large increases in the amount of data.