2003 Seasonal Consensus Climate Forecasts for Wildland Fire Management
contributed by Timothy J. Brown1, Anthony Barnston2, John O. Roads3, Russell Martin4, and Klaus E. Wolter5
1Desert Research Institute, Reno, Nevada
2International Research Institute for Climate Prediction, Palisades, New York
3Scripps Institution of Oceanography, La Jolla, California
4NOAA/NCEP/NWS Climate Prediction Center, Camp Springs, Maryland
5NOAA/CIRES Climate Diagnostics Center, Boulder, Colorado
Seasonal forecasts of two-category probabilistic temperature and precipitation anomalies were produced for the contiguous United States and Alaska as significant input into a wildland fire seasonal outlook. Forecast consensus was reached by combining several monthly and seasonal forecasts produced at the International Research Institute for Climate Prediction (IRI), the Scripps Institution of Oceanography Experimental Climate Prediction Center (ECPC), the NOAA/NCEP/NWS Climate Prediction Center (CPC), and the NOAA/CIRES Climate Diagnostics Center (CDC). The process was facilitated as part of a national seasonal assessment workshop for wildfire and climate held on February 25-28, 2003 in Mesa, Arizona and organized by the University of Arizona, National Interagency Coordination Center, and the Desert Research Institute. The primary purpose of the consensus forecast was three-fold; 1) to produce seasonal climate forecasts for use in developing a national seasonal wildfire outlook; 2) to determine whether or not additional probabilistic information could be provided for areas where individual forecasts showed little confidence; and 3) to directly integrate climate forecast information into specific stakeholder decision-making.
The forecast periods were March-May (MAM) and June-August (JJA) 2003. A combination of dynamical and statistical models from the respective organizations, and forecaster judgment were incorporated in producing the forecasts. Specifically, the IRI contribution was their most current seasonal forecasts based on the CCM3.2 (NCAR), ECHAM4.5 (MPI), NCEP (MRF9), COLA2.x, and NSIPP (NASA) models and SST predictions (Mason et al. 1999). The ECPC contribution included current monthly forecasts from two versions of the Global Spectral model as well as the Regional Spectral Model (Roads et al. 2001, 2003; Kanamitsu et al. 2002). The CPC contribution was the current seasonal long-lead outlooks based on a dynamical model, a statistical model, and long-term trend (http://www.cpc.ncep.noaa.gov/products/predict ions/90day). The CDC contribution was based on a newly developed statistical model and analysis for precipitation forecasts in the southwest U.S. (http://www.cdc.noaa.gov/~kew/SWcasts/index.html). In addition, it was based on ENSO composites for MAM and JJA during rapidly declining El Niħo phases. These objective forecasts were then combined with forecaster judgment including model forecast skill, temperature versus precipitation correlations, and current ENSO opinions.
The forecasts were produced via a round-table forum during the workshop. Forecast discussion lead to determining regions of warm/cool and dry/wet, and assigning a consensus probability. Since the forecasts were comprised of only two categories, the probabilities simply represent the chance of above or below normal. A 7.5% probability change for the median is roughly the same as a 5% change in tercile probabilities (e.g., 65% is similar to +10% for terciles). Given the current state of art for climate prediction, we equate probabilities of 65% and larger to fairly high confidence, whereas 55% represents only a slight hedge.
Figures 1 - 4 show the 2003 seasonal U.S. consensus forecasts for MAM temperature, MAM precipitation, JJA temperature and JJA precipitation, respectively. The primary highlights of these maps are above normal temperature for large portions of the West during MAM and JJA, and above normal precipitation for the Southwest during MAM. The seasonal outlook of wildfire potential, which was developed in part from these figures, is available at http://www.nifc.gov/news/intell_predserv_forms/season_outlook.html.
Since this is the second effort to produce a consensus forecast by combining forecasts from different organizations (see Brown et al. 2002), quantitative skill results cannot be offered at this time. However, the skill has been established for most of the inputs, and it is likely that the consensus forecast skill would be equal to or slightly larger than individual forecasts, depending on the region and the number of "ensemble" members that were in agreement. Skill results related to some of the individual forecasts can be found in (Barnston et al. 2001; Roads et al. 2001; Hartmann et al. 2002).
It is of interest to examine last year's forecasts (Brown et al. 2002) at least qualitatively in relation to observed anomalies. Figures 5 - 8 show maps of the 2002 probabilistic forecasts and associated anomalies produced at CPC. The MAM temperature forecast was generally good in the Southwest, but did not validate in the Great Basin and northern Rockies. The MAM precipitation forecast was quite good for portions of the West and Southeast, but did not validate in Colorado and Wyoming. The JJA temperature forecast was good for portions of the West, Southwest and East regions, but did not validate in the Southeast and the northern Rockies. Though the JJA precipitation forecast was quite conservative overall, it was generally good in the Southwest and northern Rockies, but did not validate in portions of the Northwest, central Rockies and western Plains. For all forecast seasons, there were regions with substantial observed anomalies for which no forecast was made, highlighting an overall desire and need to improve forecast skill and confidence for all regions.
Acknowledgements:
Several agencies cooperated to facilitate the workshop, including the National Interagency Coordination Center (NICC; representing agencies of the Department of Interior and the USDA Forest Service), the University of Arizona Climate Assessment for the Southwest (CLIMAS; a NOAA funded Regional Integrated Sciences and Assessments project), and the program for Climate, Ecosystem and Fire Applications (CEFA) of the Desert Research Institute. The USDA Forest Service, the NOAA Office of Global Programs, and the University of Arizona Institute for the Study of Planet Earth provided funding for the workshop. ECPC is funded by the NOAA/OGP/CDEP/ARCs program NOAA-NA77RJ0453.
References:
Barnston, A.G., L. Goddard and S. J. Mason, 2001: Verification of IRI's Seasonal Climate Forecasts. Proceedings for the 26th Annual Climate Diagnostics and Prediction Workshop, La Jolla, California, October 22-26, 2001.
Brown, T.J., A.G. Barnston, J.O. Roads, R. Tinker, and K.E. Wolter, 2002: 2002 Seasonal Consensus Climate Forecasts for Wildland Fire Management. Experimental Long-Lead Forecast Bulletin, March 2002, Vol. 11, 4 pp.
Hartmann, H.C., T.C. Pagano, S. Sorooshian, and R. Bales, 2002: Confidence builders: Evaluating seasonal climate forecasts from user perspectives, Bull. Amer. Meteor. Soc., 83, 683-698.
Kanamitsu, M., A. Kumar, H.-M. H. Juang, W. Wang, F. Yang, J. Schemm, S.-Y. Hong, P. Peng, W. Chen and M. Ji., 2002: NCEP Dynamical Seasonal Forecast System 2000. Bull. Amer. Met. Soc., 83, 1019-1037.
Mason et al., 1999: The IRI seasonal climate prediction system and the 1997/98 El Nino event. Bull. Amer. Meteor. Soc., 80, 1953-1873.
Roads, J.O., S-C. Chen and F. Fujioka, 2001: ECPC's Weekly to Seasonal Global Forecasts. Bull. Amer. Meteor. Soc, 82, 639-658.
Roads, J.O., S-C. Chen, J. Ritchie, 2003: ECPC's March 2003 Forecasts. Experimental Long-Lead Forecast Bulletin, March 2003, Vol. 12, 6 pp.
Figure captions
Fig 1. MAM temperature forecast. Red shaded areas indicate above normal temperature, blue shaded areas indicate below normal precipitation and white areas indicate a no confidence forecast region. Forecast probabilities are indicated by the percent value; areas without a value imply a 50% probability.
Fig 2. MAM precipitation forecast. Green shaded areas indicate above normal precipitation, yellow shaded areas indicate below normal precipitation and white areas indicate a no confidence forecast region. Forecast probabilities same as for Fig. 1.
Fig 3. Same as Fig. 1 except for JJA temperature.
Fig 4. Same as Fig. 2 except for JJA precipitation.
Fig 5. MAM 2002 observed temperature anomalies (departure from average) and 2002 season 2-category consensus forecast probabilities given as above and below (letters 'A' and 'B', respectively); areas without a value imply a 50% probability.
Fig 6. MAM 2002 observed precipitation anomalies (departure from average) and 2002 season 2-category consensus forecast probabilities given as dry and wet (letters 'D' and 'W', respectively); areas without a value imply a 50% probability.
