Accurate, dense, and continuous observation of water vapor is important to realize good initial condition of weather forecast and to monitor and understand global hydrological cycle. Both ground-based and space-based GPS measurement of water vapor (or refractivity) is considered as one of the key observations of water vapor because GPS observation technique is calibration free and it works under all weather condition, not like as other passive remote sensing technique.

A new ground-based network of GPS stations has been installed in the Caribbean (Caribbean network) since 2007, and we can now use near real-time and continuous GPS precipitable water vapor (PWV) data as one of the observation for weather forecast. Because the GPS stations are well distributed evenly near ocean coast and in islands, it is anticipated that GPS PWV from the network helps to improve precipitation and hurricane forecasts. A GPS occultation observation system, Constellation Observing System for Meteorology, Ionosphere, and Climate (COSMIC) was successfully launched in April, 2006, and refractivity profile over the globe has been available with near real-time. The combination of these two GPS observation is expected to show positive impact on tropical cyclone forecast.

The Weather Research and Forecasting (WRF) model and its 3DVAR data assimilation system are used for the study. The domain size is 361x261 (12 km) x 38 (vertical), centered in the north of Caribbean Sea. The analysis field of NCEP Global Forecast System (GFS) is used as initial and lateral boundary conditions for the experiment. We performed hourly cycling 3DVAR for 6 hour and 24 hour time window of data assimilation.

GPS data from IGS (International GNSS Service) stations and Caribbean network has been processed every 30 minutes, and PWV is retrieved from zenith total delay (ZTD) from the GPS processing and surface meteorological observation. The GPS PWV has been used for hurricane forecasts for 2007 and 2008 hurricane seasons, and the current forecast information is available from http://www.cosmic.ucar.edu/~iwabuchi/caribbean/.

The significant impact of ground-based GPS PWV on hurricane intensity forecast is observed in hurricane Dean case in 2007 (see figure, difference of minimum pressure of about 22 hPa is observed in 24 hour forecast) because one GPS station locates very close to the hurricane center). We thus focused on the Dean case and performed various forecast and analysis. We confirmed that just one point and one time GPS PWV data affects hurricane intensity with one-point data assimilation experiment. The phase change of water from vapor to liquid releases latent heat, and hurricane Dean becomes strong. We present detail how GPS PWV improves hurricane intensity with three-dimensional analysis of forecast output.

We have also performed OSE (Observation System Experiment) for the Dean hurricane case. The reference field comes from real-time WRF forecast operated by NCAR/MMM with higher horizontal grid resolution of 4 km. The observation data (with noises) generated for the same locations of the current GPS network was assimilated in the forecast system. The forecast result was consistent with that from true data assimilation.

Not only ground-based GPS PWV, but also COSMIC refractivity profile showed positive impact on hurricane Dean forecast. However, number of the current observation are still not enough to improve forecast of other hurricanes because we cannot choose locations of observation after genesis of tropical cyclone. We thus present OSSE (Observing System Simulation Experiment) results for the Dean and other tropical cyclone storm cases for future deployment of GPS observation network such as buoy- and ship- based GPS networks and adjusted COSMIC refractivity profiles for hurricane center.