Application of Spectral Decomposition Techniques in the Assessment and Intercomparison of Models and Observations

Abstract:

In the assessment of models using observations, or the intercomparison between different observational datasets, it is necessary to examine the coherency in the spatial and temporal variability present in different datasets. Meanwhile, global datasets are always high dimensional, therefore efficient comparison is not an easy task. In this study, we apply several spectral decomposition techniques, namely Combined Principal Component Analysis (CPCA) and Combined Maximum Covariance Analysis (CMCA), as effective means to reduce data dimension and extract the dominant variability. More importantly, these methods find the common modes of variability in different datasets, therefore allowing parallel comparison and evaluation. These methods were applied to the AOD fields from fifteen CMIP5 models and three observational datasets: MODIS, MISR and AERONET. We focus on large-scale features including the spatial distribution, seasonality and long term trends. Results show that while models qualitatively agree with observations, significant regional differences still exist, especially in regions with mixed aerosol types such as the Sahel, North India and East Asia. Compared with observations, models in general lack interannual variability. Moreover, all models indicate consistent AOD trends with increases over East Asia and decreases over East US and Europe. However, the AOD trends over these regions are not very significant in the observations Instead, a significant increase in dust concentrations over the Arabian Peninsula and a significant decrease over the biomass burning regions of South America are found in MODIS and MISR. The aerosol composition for the regions with largest disagreement is also examined.

Figure caption: The dominant mode of CMCA analysis using fifteen CMIP5 models and MODIS, MISR and AERONET. The color of the circles indicate the signal of AERONET. This mode is associated with a summer-winter seasonal cycle and models agree qualitatively with observations.