SEASONAL CLIMATE FORECASTS FOR FIREDANGER APPLICATIONS

John Roads  (SIO)

Link to NOAA Strategic Plan: NOAA Goal 2: Understand Climate Variability and Change to Enhance Society's Ability to Plan and Respond

RESEARCH OBJECTIVES AND SPECIFIC PLANS TO ACHIEVE THEM

The Scripps Experimental Climate Prediction Center (ECPC) has been making experimental, near-real-time, weekly to seasonal firedanger forecasts for the past 10 years. U.S. firedanger forecasts and validations are based on standard indices from the National Fire Danger Rating System (NFDRS), which include the: Ignition Component (IC), Energy Release component (ER), Burning Index (BI), Spread Component (SC), and the Keetch Byram drought index (KB). The Fosberg Fire Weather Index (FWI), which is a simplified form of the BI, has been previously used not only for the U.S. but also for other global regions and is thus included for comparison. As shown by Roads et al. (2005), all of these indices can be predicted well at weekly times scales and there is even skill out to seasonal time scales over many U.S. West locations. The most persistent indices (BI, ER, and KB) tend to have the greatest seasonal forecast skill. The NFDRS indices also have a weak relation to observed fire characteristics such as fire counts (CN) and acres burned (AC), especially when the validation firedanger indices are used.

The goal of our current research is to further develop these seasonal firedanger forecasts by using the recently developed seasonal NCEP global and regional CFS/RSM forecast ensembles to drive the firedanger code. From this work, we hope to significantly improve the utility of currently experimental firedanger forecasts for the USFS and other communities, which are dependent upon long-horizon forecasts for resource allocation planning but have had to previously adapt to standard climate prediction output (T and P), which is not well-suited to firedanger forecast needs. At the same time, we are hoping that this methodology can eventually be transferred to NCEP and CPC.

RESEARCH ACCOMPLISHMENTS

During the past decade seasonal forecasts have become more commonplace, although making an explicit connection to the firedanger community has been somewhat lax by the forecast community. It has been commonly assumed that forecasts of standard monthly mean temperature and precipitation could be provided and that the firedanger application community, which actually needs forecasts of other meteorological features, such as relative humidity and windspeed in addition to temperature, would somehow adapt to using only temperature and precipitation. Here we have shown that it is quite possible to make dynamical seasonal firedanger forecasts fully compatible with what the firedanger applications community need (daily to seasonal forecasts of temperature, relative humidity, wind speed, precipitation) which can then be used to drive initializing and forecast FDI.

Confirming the previous pilot experiment, it was found here that there is significant seasonal forecast skill for all of the FDI (IC, BI, ER, KB, SC) as well as the FWI, which had previously been used to forecast global firedanger forecast skill. Persistence forecasts were also evaluated and while persistence forecasts are somewhat inferior to the dynamical forecasts, they did provide a useful standard for the dynamical forecasts to exceed. It was further shown that the FDI were somewhat better related, than the input meteorological values, to fire statistics such as fire counts and acres burned, especially when the validating rather than forecast output was used.

Still, the forecast relationships are certainly weaker than the correlations with the FDI validations, especially when considering individual grid point, and further improvements in forecast skill should still be possible. We suspect that the major reason that the initializing validating fire danger had a high degree of correlation with observed fire characteristics is due to our use of observed rather than forecast precipitation. In turn, we suspect that the major decrease in forecast skill is due to the loss of skill in predicting precipitation, although the loss of skill from the forecast of other variables must also contribute.

Fig. 1 Validating and forecast fire danger indices for MJJAS

Fig. 2 Correlations of seasonal forecasts with validating analyses

Fig. 3 Correlations of seasonal forecasts with ln fire counts