Global UV Index Forecasting by the Deutscher Wetterdienst (DWD)

Henning Staiger, Deutscher Wetterdienst, Business Unit Human Biometeorology

Introduction

The UV Index is a simple means of measuring the erythemal effective UV radiation level at the Earth's surface and it is harmonised internationally. It is an indicator of the potential for skin damage and serves as an important vehicle for raising public awareness and alerting people about the need to adopt protective measures (WHO 2002). The UV Index is defined as the integral over the spectral UV irradiance on a horizontal plane, W m-2 nm1, weighted with the CIE (1987) erythemal action spectrum and multiplied by the factor 40. Harmonised methodologies for solar UVB radiation forecasting over Europe were recommended by the COST-Action 713 “UV-B Forecasting” (Bais 1999, Vanicek 2000). The DWD joined the project STREAMER funded by the European Commission and developed among others a dynamic ozone forecast. The UV Index can be derived from physical measurements or model calculations and can therefore be provided by using predictable meteorological parameters.

Methods

Dynamic prediction of ozone

The dynamic forecast is performed by using the global numerical weather prediction model GME of the DWD (Majewski et al. 2002). As the main mass of ozone is transported in the lower stratosphere, ozone can be handled as an inert tracer if only short-term predictions up to three days are performed. As the treatment of ozone as an inert tracer only requires the addition of one transport equation, an on-line coupling to GME is performed. The coupling uses a semi-Langrangian horizontal advection and a semi-implicit method for the vertical direction. The combination of these advection schemes is not mass conserving. But the differences are small and can be neglected in short-term forecasting.

The initial fields for the ozone forecast with GME are provided by the European Centre for Medium Range Weather Forecast, ECMWF, forecast period +12 h. Analysis of total column ozone initialised by ECMWF is derived from measurements of NOAA SBUV/2, of GOME-FD by KNMI/ESA up to October 2003, and of MIPAS on ENVISAT since then by four-dimensional data assimilation techniques (Jeuken et al. 1999). The ECMWF forecast +12h is interpolated to the icosahedral-hexagonal grid of GME and adapted to its topography by using a mass consistent method.

Radiation transfer calculation for clear sky conditions

The model STAR, System for Transfer of Atmospheric Radiation, from the University of Munich is applied in radiation transfer calculations (Ruggaber et al. 1994, Koepke et al. 1998). STAR is a matrix operator model based on the discrete ordinates and adding method. Radiation quantities are calculated as a function of the solar zenith angle for a model atmosphere consisting of 33 homogenous layers above a reflecting surface. Multiple molecular and aerosol scattering and absorption due to aerosol particles and gases (ozone) are taken into consideration. A neural network facilitates the calculation of the transmittance spectra of the high resolution, 150 wavelengths, from transmittance values at only seven wavelengths, STARneuro (Schwander et al. 2001). The agreement between both approaches for the UV Index calculation is in the order of ±1%. The calculations use predicted total column ozone and predicted vertical profiles between 0 and 30 km for ozone, temperature, and pressure.

Global UV Index forecasting

COST-713 recommended a centrally produced large-scale UV Index. It is intended for interpolation to the grid of high resolution models of National Meteorological Services (NMSs). It should then be adjusted to the topography of those models or to the altitude of sites the forecast is given for, to predictions of snow cover influencing changes of the albedo, to cloud conditions, and to variations of the aerosol optical depth. To accomplish the requirements, UV Index forecasting is constructed modularly, calculating in a first step a "large-scale UV Index" and then adjusting it to the topography of a given numerical weather prediction model and its predicted UV relevant results.

"Large-scale UV Index"

The "large-scale UV Index" is calculated with the following fixed values: altitude 0 m above sea level, clear sky conditions, albedo of the surface 3%, aerosol optical depth at 550 nm 0.20 for the aerosol types (Hess et al. 1998) "continental average" in the altitudes between 0 and 3 km, "free troposphere" between 3 and 12 km, and "stratosphere" above 12 km; the aerosol profile corresponds to “volcanic background”.

Although the STARneuro version with minimised computing time is used, complete radiation transfer calculations must be limited to ±3 h around the period +36 h for European sites to achieve a reasonable production time. Calculations for UV Index predictions for all hours and a global coverage with UV Index predictions become feasible based on lookup tables. Uncertainties of UV Index calculations under cloud- and snow-free conditions due to deviations from actual height profiles from monthly averages are <3% for ozone profiles, <1% for extinction profiles, and <2% for temperature profiles (Schwander et al.1997, Reuder and Schwander 1999). Monthly ozone height profiles were taken from the UGAMP ozone climatology (Li and Shine 1995), the temperature profiles from ISO 5878 (1982/1990), and the humidity profiles from the AFCLR data McClatchey et al. (1972). A symmetric distribution to the equator was assumed for temperature and humidity profiles. The profiles are then interpolated to the standard profile of 33 homogenous layers recommended for STARneuro. To determine the UV Index clear sky at mean sea level for the above standard, radiation transfer calculations were performed for the 15th of each month, for each of the climatological regions tropic, sub-tropic, mid-latitudes, sub-arctic, arctic of northern and southern hemisphere for sun zenith angles between 0 and 90° with a resolution of 1° and for total column ozone between 90 and 700 DU with a resolution of 10 DU. The monthly calculations include the varying distance earth - sun.

Adjusting for varying aerosol optical depth, altitude, snow albedo, and cloudiness

The UV Index calculations could enclose spatial and seasonal variations of aerosol optical depth by global fields of monthly averages. The MODIS sensors onboard the NASA "Earth Observing System" satellites "Terra" and "Aqua" provide inter alia monthly means of aerosol optical depth at 550 nm, AOD550, with unprecedented accuracy (Kaufman et al. 2002, Ichoku et al. 2002). A cloud mask ensures that only cloud free pixels are selected. Snow and ice covered surfaces are excluded as well as land surfaces with a high reflectivity at 2.1 µm, e.g. especially the deserts, to minimise the error. Nevertheless the cloud free pixels may still be partially contaminated by sub-pixel clouds and snow/ice resulting in an increased AOD550. Considering that increased small meshed variability, monthly averages with an increased standard deviation are reduced by a statistical approach. Especially affected are areas bordering the polar regions with missing data. These polar regions are filled with background values of AOD550 from the Global Aerosol Data Set, GADS (Koepke et al. 1997). Seasonal changes due to arctic haze accounted for (Barrie 1986, Herber et al. 2002). The remaining gaps over the deserts are filled by monthly AOD550 averages between 1979 and 2001 of the NASA Total Ozone Mapping Spectrometer, TOMS (Torres et al. 2002). The same months of the accessible period March, 2000 to September, 2003 are averaged and the resulting global fields slightly smoothed. The required global fields of the single scattering albedo are taken from GADS.

Matrices of coefficients have been developed to adjust the "large scale UV Index" to variable aerosol optical depth. They consider the sun zenith angle, the aerosol optical depth, and the aerosol type. The maximal error of parameterisation compared to modelled values is 0.21 UV Index, i.e. 1.8%, for an urban aerosol.

Matrices of coefficients allow to adjust UV Indices valid for sea level to altitudes of up to 5000 m and more. Modelled values were parameterised depending from AOD550 and are calculated for steps in altitude of 500 m, in solar zenith angles of 5°, and for the aerosol types "continental average" and "urban". The maximal deviation of parameterised versus modelled values is 0.21 UV Index, i.e. 1.37%, for an urban aerosol, an adjustment to 5000 m a.s.l, an AOD550 of 0.05, and a solar zenith angle of 0°.

Following, a UV Index clear sky is calculated accounting for an enhanced albedo due to snow cover. In Antarctica and for the Greenland Inland Ice it can be assumed that the terrain is homogeneously covered with snow or ice. Reported albedo values for UV radiation are higher than 0.9 and can increase the UV Index up to 45%. Outside these areas a regional albedo is calculated according to Schwander et al. (1999) and Lapeta et al. (2001) resulting in an increase of the UV Index between 6 and 16% depending of snow type and water content of the snow cover.

In agreement with COST-713 the UV Index clear sky is modified to a UV Index cloudy by empirical factors for a forecasted cloud cover in the ceilings low, medium high, and high (Staiger et al. 1998). The lowest cloud modification factor of the forecasted cloud cover in the three levels is selected for the adjustment.

Verifications

The GME forecasts 00 UTC +36h of the total column ozone are compared to the ECMWF initialised analysis on a monthly basis. The root mean square error in percent of the assimilated monthly pixel average is calculated and the forecast skill is reported as a reduction of the variance related to the mean square error of persistence of the analysis two days previously. The root mean square error varies between 2 and 4%, i.e. 10 - 14 Dobson Units (DU), for the moderate and high latitudes, and between 4 and 7%, i.e. 10 - 18 DU, for the tropics. There are seasonal variations with the best results in the summer months of the hemisphere and lowest in winter and beginning spring. The variance against the persistence forecast is reduced significantly by the forecasts up to 70%, especially in the moderate and also in parts of the sub-arctic latitudes. There is no reduction of the variance in the tropics due to the nearly unchanged ozone levels from day to day. A bias of 3 to 5% is found for the tropics. This may possibly caused by the GME, which uses 10 hPa (31 km) as the highest model level. But in the tropics 4 to 5% of total column ozone are found above this level. This bias will result in a slightly increased UV Index clear sky, which is on the safe side seen from the UV Index applications. Furthermore the total ozone forecasts are validated against the KNMI/ESA assimilated GOME-FD ozone fields and covering the satellite equator crossings– the reassembled GME hourly forecasts +25 to +48h against the global fields of TOMS. The root mean square error rises especially due to the systematic differences in detected column ozone between the two instruments (Corlett and Monks 2001) and the functions ECMWF applies in the initialisation to correct the sun zenith dependent GOME ozone detection (Dethof and Holm 2003). For the northern summer the agreement with TOMS is better for the moderate and arctic latitudes and with GOME-FD in the subtropics, tropics, and the southern hemisphere. A seasonal change can be anticipated for the northern winter.

Measured values of the UV Index were available at the COST-713 UV Index data base for 12 European locations between 67° N and 32° N. The daily maximum of the UV Index, May to September 2003, was retrieved from the measurements of a location and compared to the forecasted UV Index cloudy of the next clock hour. 80% of the forecasted UV Index cloudy are in the range of ±1 UV Index of the measurement. The predominant error can be attributed to cloudiness which increases the root mean square error. The forecasts show a bias of +0.27 UV Index. The TOMS total column ozone is taken as an estimate for the measured ozone and compared to the forecasted ozone. The root mean square error of 5.7% includes the systematic difference between TOMS and the instruments and methods used by the ECMWF ozone assimilation. The statistical results differ considerably the different locations. The global forecasts can not account for all local specifics. The site of Florence e.g. shows a bias of +1.35 UV Index which in essential parts is the result of the urban aerosol not accounted for in the applied wide area single scattering albedo.

The quality of the UV Index lookup tables can be examined against the complete radiation transfer calculations +33 to +39h for European pixels, 620,576 spectra of the period April to September 2003. For UV-Index values lower than 8.5 the absolute mean deviations are lower than 0.1 UV Index with a 95% confidence interval lower than 0.01. The deviations decrease rapidly with a decreasing UV Index. The maximal deviation is seen with -0.32 UV Index for a UV Index of 11. These high UV Indices result from pixels situated around 40° N. At 40° N the climatological profiles of the lookup tables are changed from "moderate latitudes" to "sub-tropic". But it can be concluded from the results that differences between the actual and climatological profiles are contributing at the most up to 3% of the variations of the UV Index clear sky.

Acknowledgements

The development of the forecast methods was supported by the European Commission through the COST Action 713 and the STREAMER project (FP4: CT97-0756, DG XII). The ECMWF provides ozone predictions. The MODIS MOD08_M3 data ( came from the NASA Goddard Space Flight Centre.We also owe thanks to the Finnish Meteorological Institute for hosting the UV Index database and to the contributors to that database which provided the measurements for comparisons. We would like to thank the Royal Netherlands Meteorological Institute for allowing the access to the assimilated GOME-FD data by an automated ftp routine. Special thanks go to Dr. Peter Koepke and Dr. Harry Schwander from the University of Munich for their expert advice in radiation transfer calculations and to Dr. Stefan Kinne, Max Plank Institute for Meteorology Hamburg, and by him to Dr. Omar Torres, NASA Goddard Space Flight Centre, for providing the TOMS mean monthly aerosol optical depths.

References

Bais AF (1999): "UV-B Forecasting (COST Action on Meteorology)", presented at the Weather, Climate and Health Symposium, Lisbon, 23 March 1999.

Barrie LA (1986): Arctic Air Pollution: An Overview of Current Knowledge. Atmospheric Environment, 20:643-663.

CIE (1987): A reference action spectrum for ultraviolet induced erythema in human skin. CIE Reseach Note. CIE Journal 6: 17-22.

Corlett GK, Monks PS (2001): A Comparison of Total Column Ozone Values Derived from the Global Ozone Monitoring Experiment (GOME), the Tiros Vertical Sounder (TOVS), and the Total Ozone Mapping Spectrometer (TOMS). J. Atm. Schi 58: 1103-1116.

Dethof A, Holm EV (2003): Ozone assimilation at ECMWF. Submitted to Q. J. R. Meteorol. Soc., pp. 15.

Herber A, Thomasson LW, Gernandt H, Leiterer U, Nagel D, Schulz K-H, Kaptur J, Albrecht T, Notholt J (2002): Continuous day and night aerosol optical depth observations in the Arctic between 1991 and 1999. Journal of Geophysical Research 107, NO. D10: AAC 6-1 - 6-13.

Hess M, Koepke P, Schult I (1998): Optical Properties of Aerosol and Clouds: The software package OPAC. Bull. Am. Met. Soc. 79, No.5: 831-844.

Ichoku C, Chu DA, Mattoo S, Kaufman YJ, Remer LA, Tanré D, Slutsker I, Holben BN (2002): A spatio-temporal approach for global validation and analysis of MODIS aerosol products. Geophysical Research Letters 20: MOD 1-1,4.

ISO (1982/1990): Reference atmosphere for aerospace use. First Edition 1982-04-15 + amendment 1 1990-12-01. International Standard 5878, Ref. No. ISO 5878-1982 (E). International Organization for Standardization.

Jeuken ABM, Eskes HJ, van Velthoven PFJ, Kelder HM, Hólm EV (1999): Assimilation of total ozone satellite measurements in a three-dimensional tracer transport model. J. Geophys. Res. 104: 5551-5563.

Kaufman YJ, Tanré D, Boucher O (2002): A satellite view of aerosols in the climate system. Nature 419:215-223.

Koepke, P, Hess M, Schult I, Shettle EP (1997): Global Aerosol Data Set. Max-Planck-Institut für Meteorologie, Report No. 243, pp. 44.

Koepke P, Bais AF, Balis D, Buchwitz M, de Backer H, de Cabo X, Eckert P, Eriksen P, Gillotay D, Koskela T, Lapeta V, Litynska Z, Lorente J, Mayer B, Renauld A, Ruggaber A, Schauberger G, Seckmeyer G, Seifert P, Schmalwieser A, Schwander H, Vanicek K, Weber M (1998): Comparison of Models used for UV Index Calculations. Photochem. Photobiol. 67: 657-662.

Lapeta B, Schwander H, Koepke P (2001): Adaption of Radiative Transfer Modelling to Simulate UV Radiation on Top of a Mountain, in IRS 2000: Current Problems on Atmospheric Radiation, Deepack publ. Hampton Virginia, pp. 373-376.

Li D, Shine KP (1995): The UGAMP ozone climatology and associated software. Department of Meteorology, University of Reading.

Majewski D, Liermann D, Prohl P, Ritter B, Buchhold M, Hanisch T, Paul G, Wergen W, Baumgardner J (2002): The Operational Global Icosahedral-Hexagonal Gridpoint Model GME: Description and High-Resolution Tests. Monthly Weather Review 130: 319-338.

McClatchey RA, Fenn RW, Selby JEA, Volz FE, Garing JS (1972): Optical Properties of the Atmosphere (Third Edition). Air Force Cambridge Research Laboratories. AFCRL-72-0497, Environmental Research Papers, No. 411.

Reuder J, Schwander H (1999): Aerosol effects on UV radiation in nonurban regions. J. Geophys. Res. 104, No. D4: 4065-4077.

Ruggaber A, Dlugi R, Nakajima (1994): Modelling radiation quantities and photolysis frequencies in the troposphere. J. Atmos. Chem. 18:171-210.

Schwander H, Kaifel A, Ruggaber A, Koepke P (2001): Spectral radiative-transfer modeling with minimized computation time by use of a neural-network technique. Appl. Optics 40: 331-335.

Schwander H, Mayer B, Ruggaber A, Albold A, Seckmeyer G (1999): Method to determine snow albedo values in the ultraviolet for radiative transfer modeling. Appl. Optics 38: 3869-3875.

Schwander H, Koepke P, Ruggaber A (1997): Uncertainties in modeled UV irradiance due to limited accuracy and availability of input data. J. Geophys. Res. 102, No. D8: 9419-9429.

Staiger H , Vogel G, Schubert U, Kirchner R, Lux G, Jendritzky G (1998): UV Index Calculation by the Deutscher Wetterdienst and Dissemination of UV Index Products, in WMO 1998, pp. 89-92.

Torres O, Bhartia, PK, Herman JR, Sinyuk A, Ginoux P, Holbein B (2002): A Long-Term Record of Aerosol Optical Depth from TOMS Observations and Comparison to AERONET Measurements. J Atmospheric Sciences 59:398-413.

Vanicek K (2000): "UV-Index for the Public" - an output from the COST-713 Project. Quadrennial Ozone Symposium, Sapporo, 3-8 July, 2000.

WHO (2002): Global solar UV index: a practical guide. A joint recommendation of the World Health Organization, World Meteorological Organization, United Nations Environmental Programme, and the International Commission on Non-Ionizing Radiation Protection. World Health Organization, Geneva. ISBN 92 4 159007 6.