There have been several projects that utilize automated data import to drive the PHYGROW engine in grid computing environments. They are listed below:
- Africa Livestock Early Warning System - This system is driven by a automated system of data acquisition from the NOAA FEWS NET site where European Union weather satellite data (METEOSAT-11 x 11 km grid) is acquired each night, extracted and stored in CNRIT's weather database. The models are run each night with the newly added weather. On the 1st, 10th and 20th the model output spanning that time period is co-kriged with the EROS NDVI data to produce continuous vegetation production and deviation maps of the entire project area. The maps are converted into KML formap and displayed on Google Maps through the GLEWS portal. The countries of Kenya, Ethiopia, Uganda, Tanzania, Southern Sudan, Dijoubti and northern Somalia are currently included in the project.
- Mongolian Forage Monitoring System USAID GOBI 2 Initiative - This project uses the NOAA CMORPH weather satellite weather data integrated in an automated environment. CMORPH is a combined microwave and infrared sensor system that reports weather at a 0.1 degree grid size every 30 minutes. The data is reprocessed by the automation program into daily data and stored in CNRIT's weather database. The models are each night with the updated weather. Every 16 days the model output for that period is co-kriged with MODIS NDVI data to produce continuous vegetation production and deviation maps of the entire project area. The maps are converted into KML formap and displayed on Google Maps through the GLEWS portal.
- Burning Risk Advisory Support System (BRASS) - The BRASS system is automated much like the previous two systems, but with a few technological differences. It acquires 4km grided National Weather Service data for the continental U.S. each night, extracts the data, and stores it in the weather database. The NWS 7 day forecast at 3 hour intervals is also accquired. Instead of using the model output and NDVI data to create a coverage, it uses a predfined coverage of polygons created through a Most Similar Neighbor Analysis process. Each polygon is assigned a sampled plant community, soil type, and grazing regime. The system automatically combines this data with the lasted weather data to run the model. The outputs are then passed to the fire model which combines the PHYGROW outputs, the 7 day forecast weather data, and fuel model data to simulate fire conditions over the next seven days. All the data is output in map, line graph and text form and presented on a website with an interactive Mapserver application.
Scalability of the system is constrained only by funding. If there are few plant communities to run and the frequency of requests for analysis is monthly then a person running a single stand alone version of PHYGROW should be able to handle it. On the other extreme, a LINUX based server with a grid computing environment can be set up for automated runs where the models can acquire the weather data and increment hundreds of thousands of simulations each day. Currently we run approximately 20,000 simulations each night using 20 grid computers with a mix of dual and quad core computing power. In all, we have have 96 cores. PHYGROW and the automation systems that have been designed for it are fully scalable and only limited by the nature of the job and the level of funds available to assemble the system.