CariCOOS: Real-time data validation of high-resolution wind forecast

摘要

CariCOOS has implemented a mirror run of the operational numerical weather prediction model employed by the National Weather Service Weather Forecast Office San Juan (NWS WFO SJU), which is based on the WRF NMM model. The model configuration consists of a one-way nesting of two domains with a horizontal resolution of 6-km and 2-km, respectively. CariCOOS developed a single domain configuration of the WRF NMM model having a finer horizontal spatial resolution of 1-km. These models, in general, are discrete approximations in space and time of a continuous geophysical medium being simulated. The solutions to these models are not an exact representation of reality. The goal is to keep track of the relative ability of a model to make accurate predictions. A real-time validation MATLAB?? script, for numerical weather prediction (NWP), was developed to assist CariCOOS end-users in comparing the forecasts to actual data. The validation of these models was conducted by comparing in-situ wind observations (Oi), of CariCOOS observing network, with model forecast (Mi). Time histories plots of in-situ and modeled wind speed and direction are compared in real time; while also solving statistical algorithms with various validation parameters as outputs. The statistical parameters included in the real-time validation tool are: (i) the Pearson's moment correlation coefficient (r), (ii) mean bias (MB), (iii) mean absolute gross error (MAGE), (iv) root mean square error (RMSE), (v) Willmott index of agreement (IOA) and (vi) hit rate (HR). This useful real-time validation tool allows NWS WFO SJU forecasters and CariCOOS researchers to measure the skill of the model and quantify the error to determine future needs for modeling enhancements. The real-time validation framework developed facilitates CariCOOS researchers to set any validation date range, which can vary from days to months. The validation skill appraisal revealed noticeable improvement in the forecasting for both wind spee- and direction by the WRF-1KM when compared to the WRF-2KM and the NWS National Digital Forecast Database (NDFD) operational product. Typically the most improved locations are noticed for weather stations in regions surrounded by complex topographic features,, particularly the lee side of islands. These findings suggest that the 1-km model resolves better the local orographic forcing generated by the fine-scale topography, typical of the coastal meteorology of Caribbean islands. The improvement in forecasting 10-m winds is due to a better depiction of the 1-km resolution model of complex terrain. These findings may lead to better understanding of local convection process, which dominates the diurnal and nocturnal wind cycles of sea and land breeze, respectively.