At the largest space and time scales, ECPC’s atmospheric modeling system uses the National Centers for Environmental Prediction’s (NCEP’s) medium range forecast (MRF) model or global spectral model (GSM; Kalnay et al. 1996; Roads et al. 1998). A high resolution regional spectral model (RSM; Juang and Kanamitsu, 1994, Chen et al. 1998; Anderson et al. 1998), is nested within the global model by first integrating the GSM, which provides initial and low spatial resolution model parameters as well as lateral boundary conditions for the RSM. The RSM then predicts regional variations influenced more by the higher resolution orography and other land distributions within a limited but high-resolution domain. A mesoscale spectral model (Juang 1997), which uses the nonhydrostatic equations, can then be nested within the RSM or can be initialized and bounded by the global model. All models use the same 28 terrain following sigma layers and the same comprehensive set of physical parameterization modular packages, which include land surface parameters (e.g. soil wetness, soil temperature, etc.), sensible and latent heat fluxes, radiation fluxes, cloudiness, etc.

The forecast procedure is as follows. First, NCEP makes the analysis and 3-day global forecasts and posts the analysis and 4xdaily 72-hour forecasts on their ftp site. Due to bandwidth limitations of the Internet, only the complete initial (operational analysis) and 72-hour forecasts 4 times daily (00, 06, 12, 18 UTC) for the global model are then transferred to ECPC computers. From the global 00UTC initial conditions, a global atmospheric forecast is made (every day), assuming persistent SST anomalies. After the first 7 global forecast days are completed (about 5 hours 275 MHz DEC Alpha CPU time), three regional domains (U.S., U.S. Southwest, California), which use the forecast global fields as initial and boundary conditions, are then integrated for 7 days (about 24 hours 433 MHz DEC Alpha CPU time for each regional model). These forecasts are also provided to other interested users to drive their regional models. The global forecast is then continued out to 12 weeks plus. The global and regional forecasts will soon run every day on a different work station and all the extended runs made during a particular week will then be combined for the same validation time. This will provide an ensemble of seasonal forecasts every week. Due to lack of computer time, only one member of the seasonal ensemble forecasts has so far been run on a regular basis, and regional seasonal forecasts are not yet done on a regular basis.

Several products from these global to regional and daily to seasonal forecasts are currently displayed on our Experimental Climate Prediction Center’s worldwide website. Output from the ECPC global (200 km), Pacific (200km) regions, U.S. (200 and 50km), U.S. Southwest (25 km), California (25 km). Products include a fire weather index (FWI, see Roads et al. 1991, 1997) and associated variables such as 2m temperature, relative humidity and 10m windspeed as well as precipitation and soil moisture for several regions. Separate displays of weather (up to 7 days) and climate forecasts are provided. The climate forecasts are displayed as weekly means for the first four weeks and then monthly means thereafter. For both the weather and climate forecasts, combinations (4xdaily for the weather forecasts and weekly for the 12 week forecasts) of all the latest forecasts can be seen on the ECPC web site in a virtual VCR, which can step through the forecasts one by one (4xdaily for the weather forecasts and weekly for the climate forecasts) forwards or backwards. Note that only a simplified table (Table 1) is shown for the ELLFB link.

As an example of the current displays, Fig. 1 shows the anomalous FWI forecast (from the global model) for the U.S. for the period Aug. 29-Sept. 26 and Sept. 26-Oct. 24, 1998. For the forecasts shown here, most of the U.S. has less fire weather potential than normal. Increased potential is found over the U.S. south and Mexican north and Canada. This reduced U.S. potential is due to reduced wind speed and relative humidity (which can be seen at (http://meteora.ucsd.edu/ecpc/projects/ellfb; temperature effects have less influence on FWI variations. The anomalies are calculated here with respect to the NCEP reanalysis anomalies (Kalnay et al. 1996) and could thus contain some influence of systematic forecast model bias. Eventually, the forecast bias of the model will be removed, which will provide a better forecast of anomalous values. As another example of the current display, Fig. 2 shows the anomalous soil moisture forecast for the U.S. for the same periods. The soil moisture anomalies are quite similar to the FWI anomalies. Windy and dry atmospheric conditions in the U.S. south, Mexican north, and Canada also correspond to dry soil moisture in these regions. Fig. 3 displays the precipitation anomalies, which are also similar to the forecast soil moisture and FWI in the southern U.S. and northern Mexico. Above normal precipitation is indicated in the Gulf of Mexico, Caribbean, and eastern Tropical Pacific.

As discussed by Roads et al. (1998), initial skill evaluations have assumed that the NCEP MRF 0-6 hour forecasts provided the necessary validation. With respect to these analyses, there were systematic forecast biases with the temperature and relative humidity biases being most noticeable. Removing the systematic bias and evaluating the true forecast skill from a still limited number of operational forecasts is frustrating and we have to be patient in developing a robust bias and skill evaluation. Nonetheless, the forecast skill is certainly consistent with current medium range forecast model skill. The greatest skill occurs initially and then decays toward zero. However, daily or weekly or monthly forecast skill does not ever reach absolute zero and the residual weekly forecast skill, when averaged into monthly and seasonal averages, shows skill levels that are significant and comparable to other long range forecasts. For all variables, the U.S. forecast skill of individual weeks is still better than persistence forecasts at 6 weeks, which indicates that at least the 1^st and 2^nd month forecasts shown here have some skill, on the average. This skill may be partly due to the time period of the current evaluation, which included the recent large ENSO anomaly. We also believe this skill is the result of our attempt to use the observed initial state and further work to identify separate contributions of boundary and initial conditions is underway.

With the acquisition of some additional workstations, daily 12-week forecasts are about to be made every day. Combining these forecasts into expressions of uncertainty as well as ensemble averages and evaluating the additional ensemble forecast skill and error distributions will soon begin. Downscaling these global ensemble forecasts to regional ensemble forecasts will be attempted after the procedure to downscale a single global climate forecast for 3 separate regions has been evaluated. Ocean forecasts will also soon begin and be compared to available assimilated ocean products. Additional atmospheric forecast products will include net surface heating and surface wind stresses. Additional ocean products will include forecast sea surface temperature, mixed layer, and thermocline temperatures. Atmospheric forecasts forced by SSTs from this ocean model as well as forecasts from a hybrid tropical ocean atmosphere model will be evaluated as possible replacements for the current persistent SST anomalies. Also, hydrologic and land surface modules using the global and, eventually, regional climate forecasts are currently being developed. These and other forecast developments, extensions, experiments, displays, and evaluations will be reported on in future issues of ELLFB.

Anderson, B.T., J. O. Roads, S. -C. Chen, and H. -M. Huang, 1998a: Regional Modeling of the Low-level Monsoon Winds Over the Gulf of California and Southwest United States: Simulation and Validation, (submitted).

Chen, S. -C., J. O. Roads, H. H. -M. Juang, and M. Kanamitsu, 1997. California Precipitation Simulations in the Nested Spectral Model. J. Geophys. Res. (in press)

Juang, H. -M. H., and M. Kanamitsu, 1994: The NMC nested regional spectral model. Mon. Wea. Rev., 122, 3-26.

Juang, H. –M., 1998: The NCEP Mesoscale Spectral Model. (submitted)

Kalnay, E. et al., 1996: The NMC/NCAR reanalysis project, Bull. Am. Meteor. Soc., 77, 437- 471, 1996.

Roads, J. O., K. Ueyoshi, S. -C. Chen, J. Alpert, and F. Fujioka, 1991: Medium-Range Fire Weather Forecasts, Intern. Jour. of Wildland Fire, 1, 159-176.

Roads, J.O., S. -C. Chen, F. Fujioka, M. Kanamitsu and H. Juang, 1997a: Global to Regional Fire Weather Forecasts, U.N. Int’l Decade for Natural Disaster Reduction. J. Goldammer, Ed.

Roads, J.O., S. -C. Chen, F.M. Fujioka, H. Juang, and M. Kanamitsu. 1997b. Global to Regional Fire Weather Forecasts. Int. Forest Fire News No.17, 33-37.

Roads, J. O., S. -C. Chen, M. Kanamitsu and H. Juang, 1998: Surface Water Characteristics in NCEP’s Reanalysis and Global Spectral Model J. Geophys. Res.- Atmos.. GCIP special issue, (in press).

Roads, J. O. and S. -C. Chen, 1998: Evaluation of the Experimental Climate Prediction Center’s Global to Regional and Daily to Seasonal Prediction System. Proceedings of the 23^rd Annual Climate Diagnostics Meeting. Miami, Florida Oct. 20-26

Table. 1 Table entry to monthly U.S. forecasts of the fire weather index, FWI, 10m wind speed, WSP, 2m temperature, T2M, relative humidity, R2M, precipitation, PRECIP, and soil moisture, SMC, that are linked to ELLFB. Both total and anomaly fields are available for perusal. Additional products are available at the ECPC web site http://meteora.ucsd.edu/ecpc/projects