9.12
AUTOTREND – Automated guidance for short-term aviation weather forecasts
Albert J.M. Jacobs, Royal Netherlands Meteorological Institute (KNMI), De Bilt, Netherlands; and N. Maat
Introduction
Although recent advances in Numerical Weather Prediction (NWP) modeling at KNMI have been substantial, these models have not yet reached a state where they can resolve clouds and precipitation at the spatial and temporal resolutions that are needed for weather forecasts at an airport. Aviation forecasters compensate for these deficiencies using the model data in combination with weather observations and topographical information in the airport vicinity. In particular, the quality of short-term forecasts (up to two hours and provided as TREND bulletins) highly depends on the availability of local and upstream observations.
Objectives
AUTOTREND aims at the development of methods, which objectively integrate observations and topographical information with NWP model data. The main purpose is to develop and implement the methods in an operational environment, and use them to provide detailed numerical guidance on changes in the local weather conditions, such as winds, visibility, clouds and precipitation, that are forecast to occur and that affect air traffic at civil airports in the Netherlands.
Results
a. The TREND guidance
A numerical TREND guidance has been developed, containing site-specific information on the development of clouds, visibility, significant weather, and wind within the next 6 hours. Figure 1 shows an example of forecasted cloud amounts in the TREND guidance. The TREND guidance has been integrated into the operational environment to support the aviation weather forecaster. The guidance is based on statistical and physical postprocessing of NWP model data and observations. The technique used for the statistical postprocessing is a multi-station version of traditional Model Output Statistics (MOS). A new concept in the multi-station approach is the introduction of additional advection predictors, which account for the influence of upstream observations on short-term forecasts. The inclusion of the advection predictors is a very successful new technique, which leads to an additional reduction of the variance for visibility and cloud base by 10 to 20%.
The guidance is updated every 30 minutes with model data from KNMI’s NWP model HIRLAM, and recent local and upstream observations. Encoding software has been provided that translates the guidance into the required aeronautical TREND code. A graphical user interface with an integrated code editor enables the forecaster to modify the suggested ‘first guess’ code. Figure 2 shows how the TREND guidance and TREND code have been integrated into the user interface.
TREND forecasts depend on actual observations and are required to be added to those observations, quasi-instantaneously. Actual observations are provided as half hourly Meteorological Aviation Routine Weather Reports (METARs) or Special Aerodrome Weather Reports (SPECI’s). The TREND guidance based on those observations is not yet available at that time, which forces us to add the guidance of 30 minutes previously to the METAR instead. This 30-minute delay has a large impact on the quality of the guidance, which can be demonstrated by objective verification. Verification results are presented in figure 3 in terms of Ranked Probability (skill) Scores (RPS); the TREND guidance is compared to the forecasters’ TREND code, to the persistence of the observation at issue time, and to the guidance based on this actual observation (TREND guidance+30 in figure 3). Lower RPS values represent a better forecast skill.
b. The downscaling winds
NWP model data forecasts are grid box averaged values. Locally observed meteorological parameters, however, and in particular wind, can deviate significantly from the grid box averaged value, due to local differences in land use and surface roughness. The difference between the model grid box average and the observed local value is part of the model error, which is referred to as the representation mismatch (RM). For 10 m wind speed the RM is dominated by the difference between model (grid box averaged) roughness and local roughness.
In order to reduce the RM, a high-resolution wind transformation method, called downscaling, has been developed. Downscaling NWP model wind basically increases the representativeness of local wind forecasts on spatially small scales such as airports. The downscaling method is based on a physical two-layer model of the Planetary Boundary Layer (PBL) where the upper boundary condition is provided by NWP model data from HIRLAM, and where roughness information of the surface is derived from high-resolution land-use maps. The method has been validated for the computation of the +03 hour forecast of the average 10 m wind speed and wind direction at various locations at Schiphol airport. Figure 4 shows the verification results of the downscaling method and HIRLAM for the synoptic observation location at Schiphol airport, for different atmospheric stability conditions (unstable, neutral, and stable). In the figure, the mean error (ME) and standard deviation in the error (SD) in the wind speed are presented for each wind direction.
The impact of surface winds on the aircraft depends on the angle between the wind direction and the geographical orientation of the runway. In general, aircraft cannot take off and land if the crosswind and tailwind components exceed certain threshold values. For practical use at the airport, the downscaling wind forecasts are tailored to several more runway specific products. One of these products is the crosswind and tailwind component at the touchdown positions at Schiphol airport. Figure 5 gives an example of a possible crosswind (perpendicular) and tailwind (parallel) forecast, up to +48 hours, for one of the touchdown positions (36R) at the airport.
Outlook
A guidance system consisting of postprocessing of NWP model data in combination with local and upstream observations, and topographical information of the airport terrain and its vicinity, is able to provide more detailed and accurate meteorological information on changing weather conditions at airports. By presenting this guidance information to the forecaster, aviation weather forecasts can be produced more efficiently. For short-term, visibility and cloud ceiling forecasts, the forecast skill, however, is reduced significantly when the guidance depends on observations which are too old. In order to benefit optimally from the detailed information available in the guidance, the update frequency of the guidance needs to be increased and the delay times minimized.
Figures
| Figure 1: | |
| Figure 2: | |
| Figure 3: | |
| Figure 4: | |
| Figure 5: |
Uploaded Presentation File(s):
AutoTrend-11ARAM.ppt
Supplementary URL: http://www.knmi.nl/~jacobs/en/index.htm
Session 9, Modeling and Verification
Thursday, 7 October 2004, 8:00 AM-12:00 PM
Previous paper Next paper
Extended Abstract (320K)