Aarhus Universitets segl

No. 143: Meteorological data series for the OML model

Løfstrøm, P. 2019. Meteorologiske dataserier til OML modellen. Aarhus Universitet, DCE – Nationalt Center for Miljø og Energi, 73 s. - Teknisk rapport nr. 143. http://dce2.au.dk/pub/TR143.pdf

Summary

The OML model is an atmospheric local scale dispersion model that is used to document that Danish installations observe air pollution limit values (C-values) as stated in the Industrial Air Pollution Control Guidelines from the Danish EPA. The C-value of a substance is a limit value for how much each installation may contribute to air pollution. OML is also used for regulation of odour. OML calculates among other statistics the maximum monthly 99 percentile for one year, and the value should comply with the C-value and limit values for odour.

Meteorological data is a central part of the data used in dispersion calculations. At present data from Copenhagen Airport from 1976 (Kastrup 1976) serves as a standard. However, odour from animal housing is regulated with data from a 10 years period from Aalborg Airport 1974-83 (Aalborg 1974-83).

As requested by the Danish EPA this report analyses which other data could possibly replace Kastrup 1976 and the corresponding consequences for the immission concentrations.

Meteorological data for OML can be constructed from measurements or from weather forecast models. Measurements consist of observation from meteorological masts typically 10 m high and radio soundings (vertical balloon profiles of temperature). The meteorological OML pre-processor uses the measurements to construct the special micro-meteorological data necessary for OML. The pre-processor includes statistical relations between e.g. cloud cover, type of cloud and radiation. The radio soundings performed with 12 hours interval also require use of models for estimation of the boundary layer height for hours in between measurements. Kastrup and Aalborg data are constructed in this way (OML data). In a number of airports in Denmark, surface observations are available for many years back in time, but radio soundings stopped in 2006.

Special masts with advanced micro-meteorological measurements for almost direct use in OML have been operated in Denmark. However, the masts are few and covers only short periods of time that differs between masts.

Changes over time of some meteorological parameters are observed during the last many decades (Danish Meteorological Institute). Meteorological data from the weather forecast model WRF (WRF data) operated by Institute of Environmental Research at Aarhus University are available for 39 years, 1979-2017 and covers Denmark in a 5.6 km grid. Due to this and the very few available radio soundings and mast measurements, the further analyses are aimed at the use of WRF data for longer periods.

WRF-data was compared with one year of ultrasonic measurements and 5 years of OML data, 1979-83. The evaluation consisted of comparing the statistical distribution of the values of available parameters. WRF data was in agreement with both measurements and OML data. However, the frequency of low (<150 m) boundary layer heights in WRF data were too high compared with OML data and was adjusted to comply with OML data statistics.

OML dispersion calculations were also compared for WRF and OML data for the 5 years for two regions. In total four types of calculations. For this purpose, seven different types of point sources with different stack and building heights were employed. The maximum monthly 99 percentile (c99) and the maximum average (avg) was calculated for each year.

The average of the c99’s for all sources in a dataset differ from -3 % to +16 % compared to data from Kastrup 1976. For a single year and a single source, the difference range from -21 % to +97 %. The average of the yearly averages for all sources in a dataset differ between +29 % to +40 % compared to data from Kastrup 1976. For a single year and a single source, the differences range from -27 % to +128 %. For sources of the same type, the direction towards both c99 and avg differs between the years and the datasets. It is not possible to tell which of the two datasets that is the most accurate, because both datasets use models.

Comparison of meteorological data from WRF for 10 years (2008-17) show no major differences among 17 Danish locations representing near-shore and inland locations, and different regions. The associated OML calculations of c99 show no systematic variation either. The un-systematic variations occur no matter if the whole area of calculation or certain distances are concerned. For an area, the direction to the maximum also varies but the distances are more alike.

The maximum of 10-year averages show some systematic behaviour. The values are lowest in near-shore locations. Values are highest for inland locations with relative high frequency of a certain wind direction.

Because WRF data represents a square of about 30 km2 that should exclude very local variations, some systematic geographical variation of c99 and in particular avg was expected. The un-systematic variations could be real and be caused by natural meteorological differences between areas. However, it could also be due to limitation in WRF data.

If WRF data express the true variations then the question is to which degree of geographical detail the WRF data should be available for OML calculations for regulatory purpose. More than 1,000 WRF points cover Denmark and this could give some practical difficulties in the everyday regulatory work. Reducing the number of WRF data to one set for each of the 98 municipalities will still have the problem with differences between near-shore and inland locations.

The c99 calculated for 10 years of WRF data for an area is in general higher than the values for Kastrup 1976 (relevant for C-values). This is due to the longer time series, which increase the likelihood for encountering disadvantageous meteorological situations and because WRF data on a yearly basis gives higher c99 values. In average for all sources and for the 17 locations, the c99 for 10 years of WRF data is 26 % higher compared to Kastrup 1976 and varies from -1 % to +91 % for individual sources and locations.

Sharp directional assessment of c99 using WRF data compared to non-directional assessment using data from Kastrup 1976 gives both higher and lower values depending on the direction.

The avg for 10 years of WRF data is in general higher than the values calculated using data from Kastrup 1976. In average for all sources and the 17 locations, the value is 21 % higher and varies from -43 % to +102 % for individual sources and locations. An important reason is that avg for Kastrup 1976 is the year with the lowest value calculated for the period 1974-83 and the values for the seven sources are 18-26 % lower than the average for all 10 years.

OML calculations with WRF data for 1979-2017 do not show trends in the yearly values of c99 or avg.

Calculations of c99 based on 10 years WRF data are in general higher than calculations with Kastrup 1976 data. In order to obtain statistical values that would be more in line with c99 values for Kastrup 1976, e.g. the fourth highest monthly 99 percentile might be applied instead. This might also result in a smooth and explainable geographical variation of the values as for the averages.