Using Multiple Satellites with a New Forward Variance Maximization and Coupled Inverse Filter Method to Quantify the Emissions of Biomass Burning and Urban Sources which have Changed Dramatically Over the Past 2 Decades

Since 2000 there have been two significant changes impacting loadings of aerosols and trace gasses on the troposphere. First, there has been a rapid expansion of urbanization and access to energy sources, coupled with significant deforestation, all leading to a rapid increase in emissions and a change in its distribution in space and time. Secondly, we now have access to multiple daily to weekly measurements of aerosols and related trace gasses on a global scale. Combining the data from these different remotely sensed platforms in space and on the ground, coupled with an understanding of the basic physical and chemical differences of different sources and substances should allow us to understand and begin to quantify how the emissions have changed over time. However, we have serious issues when it comes to analyzing changes which are rapid in either space or time, with traditional Kalman filters and 3D/4D variance techniques tending to smooth out such changes.

The approach uses the rate in the change of the difference of the variance  of the loadings of NO₂ (from OMI) which is short-lived, CO (from MOPITT) which is long-lived, and AOD (from MISR) which is short-lived in the presence of rain, and intermediate-lived under dry conditions. This combination is used to generate new a priori, which in turn have a significantly different spatial, temporal resolution than currently existing emission datasets. The magnitudes are then scaled by using a simple forward-inverse modeling framework based on an approximation of an EnKF approach, using measurements not used in the a priori fitting: AOD from AERONET and MODIS, surface measurements of trace gasses from various national and international projects, and other sources.

Our results of this new approach demonstrate that these rapidly varying sources in space and time can contribute from an additional 10% to up to 500% of emissions over these various rapidly changing regions, as compared with existing present-day inventories. The results seem to be robust for changes occurring over time scales from a week to two months, and spatial scales of 25km x 25km and larger. The technique is able to capture significant single events, inter-annual and intra-annual variation. In specific, we observe clear decreases in sources from urban North America and urban Western Europe, both increases and decreases over East Asia, and significant increase in biomass burning sources from North America, and both biomass burning and urban sources from Southeast Asia, Africa, and regions of South America.

Finally, weaknesses in the model assumptions associated with vertical transport, mis-characterized removal and in-situ processing, remotely sensed measurement biases (i.e. cloud cover), and the mathematics of sampling of the differences of the variance are discussed. In some cases, uncertainties in emissions can be expanded to cover these observations, and in other cases are highlighted for future work.