2012 Progress Report: Investigation of the Effects of Changing Climate on Fires and the Consequences for U.S. Air Quality, Phase 2.EPA Grant Number: R834282
Title: Investigation of the Effects of Changing Climate on Fires and the Consequences for U.S. Air Quality, Phase 2.
Investigators: Logan, Jennifer A. , Mickley, Loretta J. , Rind, David
Institution: Harvard University , NASA Goddard Institute for Space Studies (GISS)
EPA Project Officer: Chung, Serena
Project Period: November 1, 2009 through October 31, 2012 (Extended to October 31, 2013)
Project Period Covered by this Report: November 1, 2011 through October 31,2012
Project Amount: $599,366
RFA: Adaptation for Future Air Quality Analysis and Decision Support Tools in Light of Global Change Impacts and Mitigation (2008) RFA Text | Recipients Lists
Research Category: Air Quality and Air Toxics , Global Climate Change , Climate Change , Air
This project continues our investigation of the impact of changing climate on wildfires,and the consequences for air quality over the United States funded under EPA 2004-STAR-L1. We will improve the prediction tools of wildfire activity over the western United States by taking into account new and critical factors. To reduce uncertainties in our predictions, we will perform ensemble projections for future wildfire activity with new fire models driven by data from multiple scenarios and climate models. We will then apply the calculated fire emissions to the chemistry-aerosol transport model GEOS-Chem to estimate the fire-induced changes in carbonaceous (black carbon and organic carbon) aerosol concentrations. For these simulations, we will drive GEOS-Chem with meteorological fields archived from the GISS GCM3.
As a part of this project, we focused on predicting future wildfire activity in Southern California at the midcentury, following our work that predicts future wildfire over the western United States carried out in the first year. We compiled new gridded area burned data on 0.5°×0.5° grids for Southern California during 1980-2009 based on the interagency fire reports. We improved the prediction tools of wildfire activity over Southern California by taking into account local geographic factors (e.g., topography, population, and fuel load) and the Santa Ana wind (SAW). We applied these fire models with simulated meteorological variables under the IPCC A1B scenario from 14 climate models to project future area burned in Southern California. We used the median results from these GCMs to reduce the uncertainties from using only 1-2 models.
We extended our fire projections to Alaska and Canada, where emissions from large wildfire are transported to the contiguous United States and affect ozone air quality there. We developed relationships between area burned and meteorological variables using linear forward stepwise regression for 12 boreal ecoregions. The area burned data for 1980-2009 are compiled using interagency fire reports. After evaluating the regression fits, we applied them with simulated climate under the Intergovernmental Panel on Climate Change (IPCC) A1B scenario from 13 GCMs to project future area burned. In combination with the projected future area burned in the western United States from our previous work, we estimated wildfire emissions for ozone precursor gases in North America. With the GEOS-Chem model, we quantified the impact of the changes in fire emissions on ozone air quality in North America at the mid-century.
We developed both regression models and a parameterization model for fire prediction over the western United States. The regression models build relationships between area burned and meteorological factors with stepwise regression based on ecoregions; they explain 25%-60% of the variance in area burned in the six ecoregions. The parameterization model determines daily area burned for each grid point using an empirical function composed of temperature, relative humidity, and precipitation. We used simulated present-day and future daily meteorological variables under the A1B scenario from 14 IPCC models and the NASA/GISS GCM3 to drive both fire models. The regression models project that the annual area burned will increase by 25%-125% and parameterization models by 35%-169% in the six ecoregions; both fire models predict a significant increase in area burned for forest ecoregions. The length of fire season is 3 weeks longer in the warmer and drier climate. With the GEOS-Chem model, we estimate that the average surface concentrations of organic carbon aerosol in summer over the western United States will increase by 46%-70% and black carbon will increase by 20-27% at midcentury, due to the increased wildfire emission. A paper on this work in in review [Yue et al., 2013a].
In the above work, California was treated as one region, and our approaches for predicting wildfires were least successful there. In Year 2, we focused on improving the predictions for California. We compiled new gridded area burned data on a grid of 0.5°×0.5° for southern California during 1980-2009 based on over 55,000 fire reports from the Fire and Aviation Management Web Applications (FAMWEB). With the new area burned data, we developed and evaluated both the regression and parameterization models for three sub-regions in southern California. The regression fits, which use sitebased meteorological variables from the FAMWEB, explain 40-46% of the variance in area burned in the sub-regions, a large improvement over our previous result, 25%. The parameterization is driven with the North American Regional Reanalysis (NARR) on 0.5°×0.5° grids. We improve the parameterization approach over southern California by taking into account local geographic factors (e.g., topography, population, and fuel load) and Santa Ana wind events. It is most successful in southwest California, explaining 64% of the variance in area burned. In addition, the model captures well the seasonality of wildfires in the three regions. Projections with the parameterization show that 7 out of 14 GCMs successfully capture the maximum area burned in October induced by Santa Ana events in southwest California during 1981-2000. With these 7 GCMs, the regression models projected median increases of 103%, 22%, and 32% in area burned over three regions by the mid-century. With the parameterization, the projections show median increases of 43% and 46% in southwest California and the Sierra Nevada. The seasonality of area burned also shows some changes, with possible increases of the area burned in September and October and more large fires in November by midcentury. A paper on this work will be submitted early in 2013 [Yue et al., 2013b].
We extended our fire projections to Alaska and Canada in Year 3. We divided the North American boreal forests into 12 ecoregions. In each ecoregion, we used the same regression method as in Yue et al. [2013a] to build relationships between area burned and meteorological variables. We aggregated area burned from the interagency fire reports from FAMWEB and fire point data from the Canadian National Fire Database (CNFDB) as the predictand. The regressions explain 34-75% of the variance in area burned except for Eastern Taiga Shield. We applied these regressions with simulated climate from multiple GCMs and project that the median area burned increases by 150-390% in Alaska and western Canada at mid-century, due to the higher temperature and geopotential height relative to present day. We predict median increases of 20-90% for central and southern Canadian ecoregions but a decrease up to 50% in northern Canada due to the increase in GCM precipitation. Using the GEOS-Chem model driven by GISS GCM meteorology, we found that changes in wildfire emissions alone lead to increases in summer surface ozone level by 4 ppbv for Alaska, 2 ppbv for Canada, and 1 ppbv for the western United States by mid-century. For northwestern U.S. states, the increase of local wildfire emissions enhances surface ozone by an average of 1 ppbv and the transport of pollutants from boreal fires worsens it by an additional 0.5 ppbv. The projected changes in wildfire increase the surface ozone above the 95th percentile by 1 ppbv in northwestern U.S., 5 ppbv in high latitudes of Canada, and 15 ppbv in Alaska, suggesting a greater frequency of pollution episodes in the future atmosphere.