A flexible and robust IASI-NH3 retrieval algorithm
Abstract
In recent years, infrared sounders on board satellites have demonstrated their capabilities to detect and
measure atmospheric ammonia (NH3). The retrieval of NH3 total columns from satellite-based
measurements remains, however, challenging due to the large variability both in terms of NH3
columns and measurements sensitivity. We present here a new flexible and robust NH3 retrieval
algorithm from the measurements of IASI. The method is an extension of the method presented
recently in Van Damme et al. (2014), based on the calculation of a spectral index (HRI) from the
level1C radiances. The difference lies in the conversion of the HRI to a NH3 column. Indeed, instead
of using two-dimensional look-up tables (LUT), the conversion of the HRI relies now on a neural
network (NN), offering therefore a lot more flexibility since a neural network can easily cope with
hundreds of input parameters. We describe the major improvements of the NN-based method over the
other retrieval methods developed so-far. We next derive the first global distributions and compare the
impact of different averaging procedures on these distributions. We assess the impact of the use of
variable NH3 vertical distribution on the retrieved total column (this is a unique feature of the new
retrieval method) on a global scale. We will show with different example applications and comparison
with models how the NN-based HRI method will provide a further step in a better assessment of the
NH3 atmospheric budget, its spatial distributions and long-term trends.