Asian Dust Aerosol Models (ADAM)
Total Page:16
File Type:pdf, Size:1020Kb
Special Feature : Air Pollution Control Technologies Asian dust aerosol models (ADAM) Parameterization of dust emission reduction factors Soon-Ung Park, Anna Choe, Moon-Soo Park and Eun-Hee Lee The Asian Dust Aerosol Model (ADAM) in the Korea Meteorological Administration (KMA) was developed in 2003 as an operational forecasting model. It was first mod- ified to ADAM1 based on dust source regions in northern China. Later still it was modified to ADAM2, with enhanced ability to deliver timely and quality sand and dust storm forecastings to all Asian countries that might be affected by dust storms. ADAM2 model utilizes a Normalized Difference Vegetation Index (NDVI) obtained from spot vegetation data. This article discusses the usefulness of these data and the capability of ADAM2 as an operational dust forecast model in the Asian domain the whole year round. Introduction tions were over 1,500 ì g/m3 at most sian dust, called Hwangsa in the monitoring sites in the Republic of Ko- Republic of Korea and Kosa in rea, causing natural disasters, includ- A Japan, is a typical example of ing temporary closing of most airports mineral aerosol frequently originating and elementary schools in the country.3 Soon-Ung Park in the Gobi desert, other sand deserts, In 2002, the Asian Dust Aerosol Anna Choe the loess plateau and the barren mixed Model (ADAM) was developed on the Moon-Soo Park soil in northern China and Mongolia basis of a statistical method of dust emis- Center for Atmospheric and during the spring season.1, 2, 3 Dust sion conditions, using the World Mete- Environmental Modelling, storms occurring in East Asian desert orological Organization’s 3-hourly syn- Seoul National University Research regions tend to cause major aerosol optic reporting data for seven years from Park, Rm 515, San 4-2, Bongcheon- events well beyond the Asian conti- 1996-2002 in the source region.2, 5 The dong, Kwanak-gu nent.4 The occurrences of strong dust ADAM model successfully simulated the Seoul, 151-818, Republic of Korea storms are often associated with cata- temporal and spatial distribution of dust Tel: (+82-2) 885 6712 strophic consequences to humans and concentration and the starting and end- Fax: (+82-2) 885 6715 their environment. Indeed, very severe ing times of Asian dust events observed E-mail: [email protected] dust storms were observed in the Re- in the Republic of Korea during 21-22 Eun-Hee Lee public of Korea from 21 to 23 March and March and 7-9 April 2002.2, 3 Seoul National University from 7 to 9 April 2002.3 During these An ADAM1 model has been de- Seoul 151-741, Republic of Korea periods the observed PM10 concentra- veloped, based on the ADAM model, TECH MONITOR z Nov-Dec 2008 24 . Special Feature : Air Pollution Control Technologies by taking into account observed Figure 1: The domain of the model and the distribution of WMO synoptic monitoring tower data at Duolun in In- stations. The inner rectangle is the domain of ADAM and ADAM1 ner Mongolia and the changes in land- use types in the source regions.6 This Model Domain and WMO Synoptic Stations ( ) 60 model has been able to simulate quite reasonably the Asian dust events ob- 55 served in the Republic of Korea since 50 2003.6 45 Both ADAM and ADAM1 use the 40 total dust emission flux parameterized 35 for Saharan dust outbreaks,7, 8 where sandy soil predominates. However, the 30 25 Asian dust source regions are com- Latitude (N) posed of gobi, sand, loess and mixed 20 1, 2, 3, 4, 5 soils; so one type of soil may not 15 be relevant in estimating the volume of 10 dust emission. To take into account dif- ferent soil types, saltation fluxes9 have 5 0 been used for the parameterization of 70 75 80 85 90 95 100 105 110 115 120 125 130 135 140 145 150 155 160 165 170 175 180 the total dust emission flux from differ- Longtude (E) ent source regions10 with the observed clay content in each soil type. The new- tion factors in the source regions, esti- spatial resolution of 1×1 km2 in the ly parameterized total dust emission mated by USGS land-use types, even Asian domain (Figure 1) can be ob- flux was capable of simulating the start- though emission reduction factors can tained from the site http://free.vgt.vito.be. ing and ending times of dust events be changed along with changes in sur- The Spot 4 satellite has four veg- and the maximum dust concentrations face properties, caused by the growth etation spectral bands: the blue (BLU: observed in the Republic of Korea of grasses and trees and the expan- 0.43-0.47 μm ), the red (RED: 0.61- slightly better than those results of the sion of cultivated land in the source 0.68 ), the near infrared (NIR: 0.78- ADAM1 model. regions. 0.89 ) and MIR (1.58-1.75 ). ADAM1 as an operational model Recently, remote sensing has The normalized difference vege- to forecast dust events in the Korea been applied to monitorμm the desertifi- tation index (NDVI) from Spot 4/vege- Meteorological Administration (KMA) cation and to access the amount of tation satellite datasets is defined as: has been used mainly to forecast dust vegetation in arid and semi-arid re- events over the Republic of Korean gions.11, 12 Satellite vegetation monitor- peninsula during the spring season, ing using red and near infrared chan- when the frequency of observed Asian nels has been one of the most widely with the pixel brightness count in the dust events in the Republic of Korea is used indices of change of vegetation sea area being set to 0.18 the highest. The model therefore uses coverage.13 The Normalized Difference o a rather small domain (20-60 N and 95- Vegetation Index (NDVI) is known to WMO synoptic data 150oE) (Figure 1) with fixed land use correlate highly with green biomass Three-hourly reports of present weath- types. However, many Asian countries 14, 15, 16 and the leaf area index. Also, the er, weather codes of 07 (blowing sand) that do not have their own dust predic- number of dust storm days is known to and 08 (dust whirl) and wind speed at tion models want to use the results of have a correlation with NDVI in spring, the WMO regular reporting stations in the ADAM model. This calls for an ex- especially in central and east Inner the domain of 70-180oE and Equator- pansion of the model domain (Figure Mongolia in their inter-annual varia- 60oN (Figure 1) for nine years from 1) and for a model that can be run the 17 tion. 1998 to 2006 have been analyzed to whole year round, since Asian dust This article discusses the poten- identify occurrence locations and fre- occurs all year round somewhere or tial of the ADAM2 model for Asian dust quencies of Asian dust storms and dust the other in Asian dust source regions. forecasting with a time-dependent dust rises. There are 821 regular reporting Therefore, some parameters in the emission reduction factor due to vege- stations in the study domain (Figure 1). ADAM model need to be changed to tation in the dust source region with yield more accurate results in predict- the use of Spot/Vegetation NDVI data. ing other Asian dust events. Model description One of the most important control- The meteorological model used in this ling factors for dust emission and in- Normalized difference study is a fifth-generation mesoscale tensity is the reduction of vegetation in vegetation index (NDVI) model of a non-hydrostatic version the source region over time. However, The spot/vegetation product of Maxi- (MM5, Pennsylvania State University/ all the ADAM models mentioned above mum Value Composite Syntheses National Center for Atmospheric Re- use time-independent emission reduc- (MVC) acquired in a ten-day period in a search) defined in the x, y and ó coordi TECH MONITOR z Nov-Dec 2008 25 Special Feature : Air Pollution Control Technologies Figure 2: Spatial distribution of surface soil types (1: Gobi, 2: Sand, 3: Loess, ratio of the occurrence frequency of the 4: Mixed) in the Asian dust source region dust-rise to that of the wind speed in various soil-type regions; for each Model Domain and Surface Soil Type 60 month of a period of nine years (1998 to 2006). The threshold winds speed 55 was defined as the wind speed at the 50 normalized cumulative ratio of the dust 45 occurrence frequency being 3.5 per 40 cent.2 Very few dust-rise occurrence fre- 35 quencies was observed in the Loess region for the period of July to October. 30 25 Latitude (N) Monthly variation of the upper limit 20 Soil Type value of NDVI in the Asian dust 15 Gobi source region Sand 10 Loess The ADAM2 model uses the NDVI val- Mixed 5 ues obtained from spot/vegetation 0 product of the maximum value compos- 70 75 80 85 90 95 100 105 110 115 120 125 130 135 140 145 150 155 160 165 170 175 180 ite syntheses (MVC) for a ten-day peri- Longtude (E) od in a spatial resolution of 1×1 km2 (http://free.vgt.vito.be/) in the Asian dust nate.19, 20. The model domain (Figure 1) 60 oN) so that the models can be used source region (Figure 2). From these has a horizontal resolution of 30 km and as a forecasting model in the whole data 9-grid cells of data centred at each 25 vertical layers including major Asian Asian domain.