WRF Local Forecasts with MODIS SSTs
Numerical modeling experiments at SPoRT continue to make use of the high-resolution MODIS sea surface temperature (SST) composites (Haines et al. 2007; LaCasse et al. 2008). The primary focus has been on a numerical model initialization comparison over south Florida in which a control run included the coarser resolution NCEP Real-Time Global (RTG) SSTs while an experimental run used the MODIS composites (Case et al. 2008). The work has been done jointly with the Miami, FL (MFL) NWS WFO and the Florida Institute of Technology (FIT). This project supports SPoRT's objective of using NASA EOS datasets to help improve short-term weather forecasting by providing improved initial lower boundary information to regional mesoscale modeling.
Figure 1. SSTs in the WRF simulation initialized at 1500 UTC 21 March 2007 for (a) the 1/12° RTG SST product, and (b) the MODIS composite. (c) Difference in 12-hour forecast latent heat flux (W m-2) between the MODIS and RTG WRF simulations, valid at 0300 UTC 22 March 2007.
The figure at left shows a plot of WRF-initialized RTG SSTs, MODIS SSTs, and latent heat flux differences from a sample forecast initialized at 1500 UTC 21 March 2007. What becomes immediately apparent is the difference in the level of detail of the initial SST fields. The RTG SST shows a smoothly-varying field with ~4°C temperature increase from north to south off the west coast of Florida and only ~1°C variation off the east coast, with little variation around the shallower waters of the western Bahamas (Figure 1a). In contrast to the RTG plot, the MODIS-initialized SSTs show a very distinctive gradient of 2-3°C over a short distance on either side of the well-defined Gulf Stream current from the Florida Straits south of the Keys to the east of the Florida east coast (Figure 1b). A narrow wedge of cool SSTs is found hugging the east coast to the north of Lake Okeechobee over the Florida-Hatteras Shelf, coinciding with the location of buoy 41114, labeled in Figure 1b. Noticeably cooler MODIS SSTs are also found in the shallows of the western Bahamas. In general, the largest differences in SST are well-correlated within the regions of the shallowest ocean bottom topography (not shown).
These differences in SSTs translate directly into variations in the sensible and latent heat fluxes over the water. The difference in the 12-hour simulated latent heat flux (Figure 1c) shows as much as 100 W m-2 or more reduction in the latent heat flux over the cooler shelf waters near the Florida peninsula and western Bahamas, along with an increase in latent heat flux of comparable magnitude over the well-defined Gulf Stream region. Such variations in heat fluxes over small distances can lead to simulated mesoscale circulations that may not be resolved by predictions initialized with the much smoother RTG SST field.
Based on an SST verification at 6 marine sites, the MODIS composites improved upon the RTG errors in nearly all months (February to August 2007) for WRF simulations initialized at the 0300 and 2100 UTC, corresponding to the 1900 UTC and 1600 UTC MODIS composite times, respectively. The initial SST RMSE was reduced the most substantially in February and July, but also improved in March, April, and August (see Figure 2a and 2d below). April to June had little or no reduction in the overall RMSE.
The largest improvement in initial SST RMSE was found at buoy 41114, located within the region of cool shelf waters to the east of the central Florida east coast (Figure 1b). In every month except for May, the RMSE was reduced by as much as 1°C or more in all model initialization times (see Figure 2 below). The RMSE improvement was directly attributed to a reduction in the positive RTG bias at this station (not shown). In each model cycle, the RTG SST was too warm at buoy 41114 and the MODIS SST composite reduced this bias (sometimes too much as in the case of May and especially in the 1500 UTC forecast cycle).
Figure 2. Monthly sea surface temperature root mean square errors for all 6 marine stations in the MFL WRF domain (red lines), and buoy B1114 on the Florida east coast (blue lines) at model initialization times (a) 0300 UTC, (b) 0900 UTC, (c) 1500 UTC, and (d) 2100 UTC.
There are a few instances when the MODIS SST RMSE increased over the RTG initialization. Both the 0900 and 1500 UTC forecast cycles (which used the 0400 and 0700 UTC MODIS composites, respectively) had larger SST RMSE (see Figure 2b and 2c above) and negative biases from May to July, especially during the period from mid-June to mid-July (not shown). The possible causes of larger errors during these times and specific model initialization times include (1) Cloud contamination/latency problems in the MODIS SST compositing technique, particularly in the mid-June to mid-July time frame (Haines et al. 2007), and (2) The time difference between the MODIS composite and the model initialization. The 0700 UTC composite in particular may not be representative of the sea surface at the 1500 UTC model initialization time due to diurnal fluctuations in the SST. The enhanced SST composite being developed jointly by SPoRT and JPL should help to improve these latency issues due to cloudiness through the use of SSTs obtained from AMSR-E microwave data combined with the MODIS data.
In future work, SPoRT and FIT will complete the analysis of selected case studies, summarize the objective verification statistics, and prepare a final report of the findings. In addition, SPoRT will begin sending the 2-km MODIS SST composites to the Miami and Mobile WFOs for initializing their local WRF Environmental Modeling System (EMS) model runs. SPoRT is developing instructions and configuration files so that each office can set up their WRF EMS to incorporate the MODIS SSTs in an optimal manner for real-time WRF simulations. Once tested by Miami and Mobile, the instructions and configuration files will be provided to all of SPoRT's coastal WFO partners running their own local WRF model within the EMS. Finally, once the enhanced SPoRT/JPL SST product is developed, SPoRT will re-run selected WRF simulations during the project period for days when the latency of the MODIS product was especially large due to cloud contamination.
