Develop of numerical forecast methods of visibility for detection of fog Gregory A. Zarochentsev ([email protected]), Victoria I. Byichkova, Roman Y. Ignatov and Konstantin G. Rubinstein Moscow State University, Hydrometeorological Research Center of Russia
MAIN ISSUE Fog is common and dangerous phenomenon in the temperate latitudes of the northern hemisphere. Annually airports delay dozens of flights because the planes can not land and take off in conditions of low visibility due to the formation of fog. A correct visibility forecast with a 30-minute lead time can reduce flight delays due to the weather by 20-35%. Existing methods of fog forecasting based on the high value of relative humidity of air shows low accuracy due to that a high value of humidity does not always lead to the formation of fog. MAIN OBJECTIVES Ø Analysis of fog observation data for 12 years. Ø Identification of several predictors of fog for its identification Ø Testing of our own algorithm and comparison with existing methods of visibility forecast
ANALYSIS OF OBSERVATIONAL DATA DEDICATED PREDICTORS The vertical temperature gradient in the 2m-925 hPa layer, oC/100м Dew point value at 2m level, oC/100м 20 80 Fog Fog Other events Other events 15 60
10 40
5 20
0 0
Frequency of occurrence, % o o -1-0 0-1 1-2 2-3 3-4 4-5 5-6 6-7 7-8 8-9 C Frequency of occurrence, % -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 C/100м Map of synoptic stations on Earth Map of aerological stations on Earth Wind speed at 10m level, oC/100м 30 The observational data were analyzed from January 1, 2004 to October 16, 2016: Fog Other events Synoptic data: Aerological data: 20 • from 10 200 stations with a 3-hour • from 852 stations with a 12-hour interval interval 10 • 3 481 091 for events • 82 714 for events 0
• 63 947 681 other weather events • 4 390 436 other weather events Frequency of occurrence, % 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 m/s
THE PROPOSED METHOD EXISTING METHODS FOR PREDICTING VISIBILITY 1 �����(�� − 97.5) 1 �����(|�| − 5.5) �� 5.5 − + , � ≤ 0.02 for 3 hours Method I - Rapid Update Cycle (RUC), USA: � = 60 exp −2.5 ∗ min (0.8; − 0.15) � = 2 � 2 � 100 � − �� 5.5, � > 0.02 for 3 hours Method II - Forecast Systems Laboratory (FSL), USA: � = 9656 �� . �� |�| � −ln (0.02)) Where is relative humidity at 2m, is a wind speed at 10m, is a number of precipitations in � = Method III - Steolinga and Warner (SW99), USA: . . . mm. This method is a discriminant function of two arguments: relative humidity and wind speed. High 144.7� + 1.1� + 163.9� + 10.4� value of humidity and low value of wind speed predicts the fog occurrence, however on the stations fog Method IV - Decision tree method (DTM), Hungary : � = −1.33 + 0.9( � − � + � − �� + � ) do not always form with such weather conditions. It means that all the necessary factors are not taken into account.
PERIOD AND DISTRICT WRF-ARW v.3.7.1 FORECATS VALIDATION OF STUDY CONFIGURATION Absolute error Dispersion Period August, October, December 2015 Forecasts lead time 48 hours Forecast date Research Europe and the European Domain 310×320 points Wind speed, m/s Humidity, % Wind speed, m/s Humidity, % district part of Russia Horizontal resolution 18 km Station count 2 200 20.08.2015 1.78 29.91 2.07 20.13 Fog events 4 909 Vertical levels 30 levels Time step 3 hours 24.10.2015 2.08 22.41 2.29 20.14 Time moments 16 moments Microphysics Thompson 19.12.2015 2.49 15.34 2.51 17.42 Cu_microphysics Grell-3 SW and LW- radiation RRTMG Estimates of the quality of the forecast relative humidity and wind speed, representing Surface and boundary layer Noah the difference in these values from the WRF-ARW model and from observations at Initial and boundary conditions 0.5° GFS meteorological stations, resulting from the use of linear interpolation on model grid.
COMPARISON OF FORECAST DATA AND OBSERVATIONAL DATA EXAMPLE OF VISIBILITY FORECAST AND COMPARISON WITH FOG OBSERVATION DATA Total Probability of Success Probability of False alarm Forecast method Accuracy, % detection, % ratio, % absence, % ratio, % Method I (RUC) 27 83 5 23 94 Method II (FSL) 64 40 6 65 94 Method III (SW99) 64 37 5 65 95 Method IV (DTM) 61 51 6 61 93 Our method 43 81 8 41 92 (without precipitation) Our method v1.2 62 61 11 61 87 (with precipitation) The results of estimations of various methods of visibility calculation for the fog forecast for the entire analysis area averaged over three observation intervals.
CONCLUSIONS Ø Database of fog observation has been created. Analysis of this data has been carried out and it shows three predictors of the fog process: high value of relative humidity, low value of wind speed, low value of vertical temperature gradient Ø Existing methods of fog forecast were evaluated. It has been obtained that the use of only the humidity and temperature values is not sufficient for the correct prediction of the visibility range Ø A proprietary method for predicting the range of visibility has been developed, assessments have shown better accuracy as compared to the methodology of other methods, but the method requires refinement in view of the large number of errors
FURTHER PERSPECTIVES Example of calculation of the visibility range for the fog forecast (gray field) for all the methods presented using WRF-ARW data and a map of stations where fog was observed Further it is planned to simulate the near-surface characteristics of the atmosphere for 15.00 October 24, 2015 using the WRF-ARW model for a region with a large number of registered fog events with a small resolution of the model grid to solve the problem of interpolation to ACKNOWLEDGE observation station points, and also to carry out a statistical analysis to identify model This work has been supported by RFBR (Russian Foundation for predictors of fog formation, whose values can not be measured with the help of Basic Research) under grants 16-35-00489, 16-05-00822 A, 14-08- equipment or detectors. 01105\16, 15-05-0239516.