HWRF Sensitivity to Cumulus Parameterizations
HWRF Sensitivity to Cumulus Paramet. Executive Summary
• The DTC conducted tests to document the sensitivity of HWRF towards different cumulus schemes (the HWRF operational SAS, a different implementation of SAS, Kain Fritsch and Tiedtke)..
• This test was motivated by discussions that followed the 2011 physics workshop of the HFIP Regional Modeling Team.
• Almost 250 forecast cases for the 2011 season were run for the Atlantic and Northern Eastern Pacific basins for four configurations of the model in which only the cumulus schemes were varied. The sample size for the Pacific is small and results were only interpreted for the first 72 h.
• The code employed was a developmental version of HWRF used in pre-implementation testing during February 2012.
• The configuration using the HWRF SAS cumulus parameterization, HPHY, provided Statistically Significant (SS) better track forecasts for the Atlantic, whereas for the Pacific the results are tied.
• The configuration using the Tiedtke scheme had the highest track errors for the Atlantic basin at longer lead times. The degradations are SS, and are the result of higher along-track errors.
• In both basins the along-track errors are near zero in the first few days of the forecast. After that, they remain negative for all configurations in EP, but are positive for the Tiedke configuration in the AL.
• The intensity mean absolute error results are comparable for all configurations and there is no clear superior scheme for either basin.
• In the Atlantic basin, for most configurations and forecast lead times the intensity bias is not SS different from zero. The exception is that HPHY and HNSA have positive intensity bias at the four- and five-day lead times, and HKF1 has negative bias in days two and three. In the Pacific basin, all configurations display SS negative biases in the first three days of forecasting.
• In the Atlantic, the forecast storms are too large for all lead times, wind radii, quadrants, and configurations. The two configurations using the SAS scheme further exaggerate the storm size at the longer lead times. In the Pacific, the storms are initialized too large but decrease to a near-zero size bias after one to two days.
• A relationship between intensity error and storm size error was detected. For all configurations, especially for the 50- and 64-kt radii, storms that are too large are also too strong.
• The wind-pressure relationship for the different schemes shows a good subjective match with observations in the Atlantic basin. In the Pacific, the overall wind-pressure relationship was good but there were some instances in which the initial MSLP was too low.
• No clear relationship between track and intensity errors was identified in this study. It is not possible to say that errors in forecasting the storm location lead to worse intensity forecasts.
• This test was motivated by discussions that followed the 2011 physics workshop of the HFIP Regional Modeling Team.
• Almost 250 forecast cases for the 2011 season were run for the Atlantic and Northern Eastern Pacific basins for four configurations of the model in which only the cumulus schemes were varied. The sample size for the Pacific is small and results were only interpreted for the first 72 h.
• The code employed was a developmental version of HWRF used in pre-implementation testing during February 2012.
• The configuration using the HWRF SAS cumulus parameterization, HPHY, provided Statistically Significant (SS) better track forecasts for the Atlantic, whereas for the Pacific the results are tied.
• The configuration using the Tiedtke scheme had the highest track errors for the Atlantic basin at longer lead times. The degradations are SS, and are the result of higher along-track errors.
• In both basins the along-track errors are near zero in the first few days of the forecast. After that, they remain negative for all configurations in EP, but are positive for the Tiedke configuration in the AL.
• The intensity mean absolute error results are comparable for all configurations and there is no clear superior scheme for either basin.
• In the Atlantic basin, for most configurations and forecast lead times the intensity bias is not SS different from zero. The exception is that HPHY and HNSA have positive intensity bias at the four- and five-day lead times, and HKF1 has negative bias in days two and three. In the Pacific basin, all configurations display SS negative biases in the first three days of forecasting.
• In the Atlantic, the forecast storms are too large for all lead times, wind radii, quadrants, and configurations. The two configurations using the SAS scheme further exaggerate the storm size at the longer lead times. In the Pacific, the storms are initialized too large but decrease to a near-zero size bias after one to two days.
• A relationship between intensity error and storm size error was detected. For all configurations, especially for the 50- and 64-kt radii, storms that are too large are also too strong.
• The wind-pressure relationship for the different schemes shows a good subjective match with observations in the Atlantic basin. In the Pacific, the overall wind-pressure relationship was good but there were some instances in which the initial MSLP was too low.
• No clear relationship between track and intensity errors was identified in this study. It is not possible to say that errors in forecasting the storm location lead to worse intensity forecasts.