We evaluated predictions of hourly PM2.5surface concentrations produced by the experimental BlueSky Gateway air quality modeling system during two wildfire episodes in southern California (Case 1) and northern California (Case 2). In southern California, the prediction performance was dominated by the prevailing synoptic weather patterns, which differentiated the smoke plumes into two types: narrow and highly concentrated during an offshore flow, and diluted and well‐mixed during a light onshore flow. For the northern California fires, the prediction performance was dominated by terrain and the limitations of predicting concentrations in a narrow valley, rather than by the synoptic pattern, which did not differ much throughout the wildfire episode. There was an over‐prediction bias for the maximum values during this episode. When the predicted values were compared to observed values, the best performance results were for the onshore flow during the southern California fires, indicating that the coarse grid used by BlueSky Gateway appropriately represented these well‐mixed conditions. Overall, the southern California fire predictions were biased low and the model did not reproduce the high hourly concentrations (>240μg/m3) observed by the monitors. The predicted results performed well against the observations for the northern California fires, with a large number of predicted values within acceptable range of the observed values.

For large wildfires and wildland fire use (WFU) fir es for which there is a federal response, Incident Command Summary reports (known a s ICS-209 reports) are created on a near-daily basis. ICS-209 reports contain useful i nformation about particular fires or fire complexes from the incident command team on the gro und, such as descriptions of the fuel loading, growth potential, and type of fire. Howev er, ICS-209 reports also have several limitations. Daily estimates of actively burning a re s are required, but ICS-209 reports provide only the ignition point of the fire and an estimate of the total area burned over the lifetime of the fire. For large fires, active flame fronts can mov e dozens of kilometers from the original ignition point of the burn. More importantly, ICS-209 repor ts are only created for a small subset of fires. Fires that are not tracked with ICS-209 reports inc lude prescribed burns, agricultural burns, and wildfires for which there is no federal response. Taken together, these missing fires represent a large fraction of the total area burned and resulti ng smoke emissions. The National Interagency Fire Center (NIFC) reports that at least 9000 km 2 of prescribed burning has been accomplished each year since 2001 in the US, representing up to 40% of the total area burned (http://www.nifc.gov/fire_info/fire_stats.htm).

This project addresses the need for a collaborative architecture for scientific modeling that allows various scientific models to easily interact. The need for such a system has been documented by recent studies such as the JFSP Smoke Roundtables and the JFSP review of tools done by the Software Engineering Institute. This project addresses these needs by modifying the BlueSky Modeling Framework so that it can better serve as a collaborative architecture, and then utilizing this architecture to create an advanced application that could not otherwise be created. The BlueSky framework was modified for this purpose, and all changes integrated into all versions of BlueSky from 3.1.0 forward. BlueSky now contains a command line option that will automatically start it as a web-service provider, allowing it to be used by remote clients. When the web-service option is used, all models contained within BlueSky are automatically converted into web-service accessible modules, without need for a specialized web-service enabled version. Simple examples and documentation scripts designed to show a website or user interface creator how to access these models via web-service function calls were created. In addition, a more advanced website interface was created to show some of the advantages of web-service based scientific modeling. This tool, called BlueSky Playground, provides a single user interface into 10 models of fuels, consumption, emissions, plume rise, and smoke dispersion. A user can walk step-by-step through all of the model steps in the framework from fire information to smoke impact maps. At each step the user can choose the model they want to use and alter the modeled information before continuing on, allowing for a game-playing exploratory mode of interaction. Both the ability to access so many models through a single interface as well as the capability to obtain on-the-fly smoke dispersion calculations are novel to this tool. This application will be highlighted in 2010 through RX-410 classes as a way for users to learn about the various component models. It will also serve as a training tool for managers needing to run multiple scenarios and understand the implications of various choices. The web-service oriented architecture utilized in the project offers many potential advantages to scientific research done with the goal of decision support. Separation of the scientific computing portion of such work from the user interface allows scientists to focus on creating the best models and web designers to focus on creating the best interfaces. Remote functioning of the models through the web means that local installation of the model is no longer required solving distribution issues, and allows an Internet user to run a model that requires resources not available to them locally (such as large datasets or fast processors). Modularity allows for “mash-ups” where models are combined in ways not originally foreseen to meet emerging needs, and provides choices to be made on exact modeling pathways at the user or institutional level.