Evaluating AEROCOM Models with Remote Sensing Datasets

Tuesday, 16 December 2014: 2:10 PM
Nick Schutgens1, Edward Gryspeerdt2, Natalie Weigum1, Andreas Veira3, Daniel Partridge1 and Philip Stier1, (1)University of Oxford, Oxford, United Kingdom, (2)University of Leipzig, Leipzig, Germany, (3)Max Planck Institute for Meteorology, Hamburg, Germany
We present an in-depth evaluation of AEROCOM models with a variety of remote sensing datasets: MODIS AOT (& AE over ocean), AERONET, AOT, AE & SSA and Maritime Aerosol Network (MAN) AOT & AE. Together these datasets provide extensive global and temporal coverage and measure both extensive (AOT) as well as intensive aerosol properties (AE & SSA).

Models and observations differ strongly in their spatio-temporal sampling. Model results are typical of large gridboxes (100 by 100 km), while observations are made over much smaller areas (10 by 10 km for MODIS, even smaller for AERONET and MAN). Model results are always available in contrast to observations that are intermittent due to orbital constraints, retrieval limitations and instrument failure/maintenance. We find that differences in AOT due to sampling effects can be 100% for instantaneous values and can still be 40% for monthly or yearly averages. Such differences are comparable to or larger than typical retrieval errors in the observations. We propose strategies (temporal colocation, spatial aggregation) for reducing these sampling errors

Finally, we evaluate one year of co-located AOT, AE and SSA from several AEROCOM models against MODIS, AERONET and MAN observations. Where the observational datasets overlap, they give similar results but in general they allow us to evaluate models in very different spatio-temporal domains. We show that even small datasets like MAN AOT or AERONET SSA, provide a useful standard for evaluating models thanks to temporal colocation. The models differ quite a bit from the observations and each model differs in its own way. These results are presented through global maps of yearly averaged differences, time-series of modelled and observed data, scatter plots of correlations among observables (e.g. SSA vs AE) and Taylor diagrams. In particular, we find that the AEROCOM emissions substantially underestimate wildfire emissions and that many models have aerosol that is too absorbing.