               Regional Meteorological Data Processing Protocol
                            EPA Region 5 and States
                              August 2014 - DRAFT

Purpose:    The purpose of this protocol is to recommend a methodology that can be used by Region 5 States to process National Weather Service meteorological data for use in AERMOD regulatory modeling applications.  This document highlights several aspects of meteorological data processing including:  choice of surface meteorological data, choice of upper air meteorological data, approach for processing surface characteristics, and thoughts on representativeness.   

This protocol is the result of numerous discussions held between the Environmental Protection Agency (EPA) and the States.  It is based on guidance available in the Guideline on Air Quality Models, 40 CFR Part 51, Appendix W (Guideline).  It also relies on the information in the AERMOD Implementation Guide, dated March 19, 2009 and User Guides for AERSURFACE, AERMET, and AERMINUTE.   Other references are highlighted in the text of the document.   

Contents: 

I.	Background  

II.	Surface Data   
	A.	Time period and Format
	B.	Missing data and Calms 
	C. 	Light Wind Considerations
	
III.	Upper Air Data Selection 
	A.	Time period  
	B.	Format 
	C.	Missing data  
	
IV.	Surface Characteristics  
	A.        AERSURFACE  
		1.  	Land Cover Data
		2.	Meteorological Tower Location
		3. 	Seasonal Definitions
		4.	Arid Region/Surface Moisture
		5.	Domain 
			 
					
V.	Representativeness 
I.  Background 

Over the past 20 years, meteorological (met) data processing has changed significantly.  For most of the 1990's and early 2000's, Region 5 States used ISCST3 and the meteorological data available from the Support Center for Regulatory Atmospheric Modeling (SCRAM), typically using a 5-year period from the years 1984-1992.   This data was acquired from the National Climatic Data Center (NCDC) and, for the most part, filled by EPA and ready to use in PCRAMMET.  

As AERMOD replaced ISCST3 as the recommended regulatory modeling tool, new meteorological processing procedures were implemented.  Namely, surface characteristics (i.e., surface roughness, Bowen ratio, and albedo) are used in the AERMOD preprocessor, AERMET.   This information is used in AERMET to calculate scaling parameters for atmospheric dispersion, which are used in AERMOD to determine the profiles of winds, temperature, and turbulence.   The release of AERSURFACE has provided a consistent, technically sound approach for determining appropriate surface characteristic values.  

For many years, States have provided processed meteorological data to applicants for use in regulatory applications.   The approaches used to process the meteorological data through AERMET have often varied from state to state.   This document lays out a methodology that allows Region 5 states to develop processed meteorological data in a consistent manner. 

II.	Surface Data 

A. Time Period and Format: 
      
It is getting harder to justify the use of 20-25 year old meteorological data.   Issues such as urban sprawl and other land-use changes can impact the representativeness of older data.   Additionally, the general public is increasingly questioning why old meteorological data are driving our air quality decisions.  Very recent National Weather Service data is available from NCDC in a standard hourly format (i.e., ISD) and in a 2-minute average format for ASOS sites. 

Recommendation:  SCRAM data should be replaced.  As the Guideline states, at least 5 consecutive years of recent, readily available, adequately representative data should be utilized.   Free full-archive NWS data is available at   ftp://ftp.ncdc.noaa.gov/pub/data/noaa/ .     One-minute ASOS data (DSI-6405) is available at ftp://ftp.ncdc.noaa.gov/pub/data/asos-onemin/ . The 1-minute data is actually overlapping 2-minute averages  The purpose and benefits of the 1-minute meteorological data is discussed in the next section.  



B. Missing data and calms	
      
The SCRAM meteorological data was provided with missing data filled in for the vast majority of stations (not including calms).  This was necessary because ISCST3 required a 100% complete meteorological data set (excluding calms) when run in the regulatory default mode.  While valid data retrieval requirements must still be met,  AERMOD does not require missing data to be filled.  However, Section 8.3.3.2.c of Appendix W recommends that "[A]fter valid data retrieval requirements have been met, hours in the record having missing data should be treated according to an established data substitution protocol provided that data from an adequately representative alternative site are available [emphasis added]." The EPA guidance document, Missing Data Procedures for Substituting Values for Missing NWS Meteorological Data for Use in Regulatory Air Quality Models, dated July 7, 1992, provides a technique for filling in missing data.   This approach uses either interpolation for short time frames (a few hours) or substitution from another meteorological data set determined to be reasonably representative, through examination of information such as wind roses, terrain, and surface characteristics.   The substitution procedure was primarily applicable to National Weather Service (NWS) data sets.  AERMET version 13350 and later, includes options to substitute for missing cloud cover and/or ambient temperature using linear interpolation across 1-2 hour gaps. 

Beginning in the early 1990's NWS surface data has been collected using the Automated Surface Observation System (ASOS) method.  In addition, the Meteorological Terminal Aviation Routine (METAR) system for coding weather observations was adopted in July 1996, which imposed a strict wind speed threshold of 3 knots (i.e., any wind speed less than 3 knots is reported as a calm.)  Additionally, wind speeds up to 6 knots associated with wind directions varying more than 60 degrees within the 2 minute observation period is reported as variable (i.e., missing wind direction for purposes of air quality modeling). The new reporting system has resulted in an increased number of calm and missing hours in the ASOS meteorological files.     

Meteorological data consisting of 2-minute average wind information, rolled every minute, is available on the NCDC website at ftp://ftp.ncdc.noaa.gov/pub/data/asos-onemin/.  These data sets are titled 6405 (2-minute data) and 6406 (5-minute data).  The rolling 2-minute average wind data currently being used in AERMET is 6405.  This data is available for most NWS stations and can be used to generate 5-year meteorological data sets that contain significantly fewer calm and missing hours compared to standard hourly ASOS data.   A methodology for processing the 2-minute data has been developed by the Office of Air Quality Planning and Standards (OAQPS), called AERMINUTE and can be applied to the 2-minute average data to generate hourly averaged wind data.   Recent versions of AERMET (beginning with version 11059) can incorporate the AERMINUTE output file.  In addition to addressing the calm and missing hour issue, the meteorological data files generated using the 2-minute wind data can improve the representativeness of airport data by providing true hourly averages rather than using a single 2-minute observation to represent the hour.   

An EPA Office of Air Quality Planning and Standards memorandum was released on March 8, 2013 entitled Use of ASOS Meteorological Data in AERMOD Dispersion Modeling.  This memorandum provides a complete discussion of the implementation of ASOS data and the sensitivity of AERMOD results to automated rather than human based observations, the impact of the METAR reporting system, and the use of AERMINUTE.


Recommendation:    Two-minute average wind data should be used for regulatory modeling applications.  AERMINUTE should be used to generate the hourly average observations from the 2-minute data.  (See AERMINUTE and AERMET User's Guides for more information).   AERMET will process AERMINUTE output to produce the surface meteorological data file for use in AERMOD.   Standard hourly ASOS data should be included using the SURFDATA keyword in the AERMINUTE control file.   The Quality Control tools available in AERMINUTE should be used to examine the usefulness of flagged data and to validate against quality controlled data from the standard observations.    Missing cloud cover and/or temperature data should be substituted using the linear interpolation options available in AERMET.  

C.  Light Wind Considerations

The AERMET processor, beginning with Version 12345, contains a keyword (THRESH_1MIN) that allows the user to specify a wind speed below which winds in the 2-minute ASOS data will be considered calm.     There is no default value, however a warning will be given if the value chosen is greater than 0.5 m/s.  The rationale behind the setting the 0.5 m/s threshold is to make that value consistent with the minimum wind speed value set under current meteorological monitoring guidance for site-specific towers.   This minimum wind-speed threshold in AERMET is applicable for applications using 2-minute average data. 
    
AERMOD and AERMET (Version 12345 and later) contain several other options designed to address concerns regarding AERMOD model performance in light wind conditions.  These options are described in Model Change Bulletin 8 as follows:  "Incorporated two new BETA (non-Default) options to address concerns regarding model performance under low wind speed conditions.  BETA options are not available for standard regulatory use.   Prior to use, they need to be justified in accordance with the provisions of Appendix W and approved by EPA in order to be used in a regulatory modeling context.  

The LOWWIND1 option increases the minimum value of sigma-v from 0.2 to 0.5 m/s and "turns off" the horizontal meander component. The LOWWIND2 option increases the minimum value of sigma-v from 0.2 to 0.3 m/s, and incorporates the meander component, with some adjustments to the algorithm, including an upper limit on the meander factor (FRAN) of 0.95. A new LOW_WIND keyword has been added to the CO pathway that allows users to adjust the minimum sigma-v value (within a range of 0.01 to 1.0 m/s), and the minimum wind speed value (within a range from 0.01 to 1.0 m/s), with a default value of 0.2828 m/s, consistent with the default applied in previous versions based on SQRT(2*SVmin*SVmin) with SVmin=0.2. The new LOW_WIND keyword also allows users to adjust the maximum value for the meander factor (FRAN) within a range of 0.50 to 1.0, inclusive, when the LOWWIND2 option is used.  Both of the new LowWind BETA options also modify the the adjustment of vector mean wind speeds (based on Eq. 112, p. 79, of the AERMOD Model Formulation Document) to use the original values of sigma-v before they are adjusted based on SVmin. The new LOWWIND1 and LOWWIND2 options are mutually exclusive and the model will issue a fatal warning message if both options are specified."

Changes to the surface friction velocity value in AERMET are applied by use of the BETA option, ADJ_U*.     The User's Guide describes the change as follows: "Beginning with version 12345, the AERMET program includes a non-Default BETA option in Stage 3 processing to adjust the surface friction velocity (u* or ustar) for low wind speed stable conditions, based on Qian and Venkatram (2011)....The ADJ_U* "BETA" option is considered to be a non-Default option and is therefore subject to the alternative model provisions in Section 3.2 of Appendix W (40 CFR Part 51).  Users should coordinate with the appropriate reviewing authority regarding the procedures and requirements for approval of this BETA option for regulatory modeling applications."

Recommendation:   The minimum wind speed threshold option should be used in AERMET when processing 2-minute average data (DSI-6405) by applying a minimum speed of 0.5 m/s.    (See March 8, 2013 EPA memorandum entitled Use of ASOS meteorological data in AERMOD dispersion modeling.)

The use of the Beta LOWWIND and ADJ_U* options in AERMOD and AERMET represent alternative forms of the models and therefore must be adequately justified by the applicant in accordance with Appendix W Section 3.2.2 before they would be accepted for use on a case-by-case basis.  For ADJ_U*, some evaluation of the option has already been conducted by OAQPS and may be useful to justify its use on a case-by-case basis.   



III.	Upper Air Data

A.	Time Period Recommendation:   Use time period concurrent with surface data. 

B.	Format Recommendation:  	Upper Air data is available for free from http://esrl.noaa.gov/raobs/.  Users should select:  Hours of access --  All times;  Data levels  -  All levels; Wind units  -  tenths of meters/second.   

C.	Missing data recommendation:  As with surface data, filling upper air missing data is not required.  However, changes to AERMET, starting with Version 12345, allow for substitution of missing UA soundings with other adequately representative UA stations.   Filling in missing upper air data when adequately representative alternate station data is available is recommended.   

IV.   	Surface Characteristics 

A. AERSURFACE  	 
   
As mentioned above, AERSURFACE is an EPA tool which implements a recommended methodology for processing surface characteristics.    The AERSURFACE User's Guide, combined with the AERMOD Implementation Guide, provides substantial information on how to implement AERSURFACE.   This protocol will address some decision points not explicitly outlined in the Guides.  The information in this document will apply to those using AERSURFACE as well as those using other spatial tools such as Geographic Information Systems (GIS).  

1.  	Land cover data 

National Land Cover Database (NLCD) data is available from the USGS Land Cover Institute website:   http://www.mrlc.gov/ .  Hit the Find Data tab to access 1992 land cover data.   1992 Land Cover data by state can also be found at http://edcftp.cr.usgs.gov/pub/data/landcover/states/ .   AERSURFACE is currently designed to read the 1992 Land Cover data.   A draft version of AERSURFACE has been developed which can read 2001 and later land cover data (supplemented by Impervious and Canopy data), but has not yet been released for general use.  Some concerns about the 2001 and later land cover categories have emerged. However, some of these concerns are mitigated by the availability of Impervious and Canopy data.  Furthermore, the temporal representativeness of 1992 NLCD for the meteorological data periods being processed will likely become a more significant issue as more recent data are processed.  

Recommendation: 	For the state supplied meteorological data, examine the temporal representativeness of the 1992 land cover data for use in AERSURFACE.   If land use has not changed significantly at the meteorological tower site since 1992,  1992 data should be used.  If 1992 NLCD is not representative for a particular application, contact the EPA Region 5 Office to discuss possible processing using the draft AERSURFACE tool for later NLCD information.   Any alternative approach to determining surface roughness, Bowen ratio and albedo values should be discussed, justified, and documented prior to use.   

The Gust Factor Tool is available on a case-by-case basis to help determine an appropriate surface roughness value if the NLCD information is questionable.  The Gust Factor Tool has not been released yet but is expected to be available soon.  Talk to EPA Region 5 about potential use of this tool.  The Gust Factor Tool may also be useful for identifying possible temporal representativeness issues with the land cover data and/or errors in the location of the meteorological tower.  

NLCD can be downloaded in two formats: complete state files or a user defined geographic area.  Either approach can be accessed through http://www.mrlc.gov/ .    State files are available at http://edcftp.cr.usgs.gov/pub/data/landcover/states/ .  
Domain-specific National Map data is preferable for sites near state boundaries because it will provide complete coverage on both sides of the state line.  

2.  Meteorological Tower Location

The AERSURFACE program asks the user  "Is this site at an airport?"   The answer impacts how AERSURFACE will apply the Commercial/Industrial/Transportation category (Cat. 23).   If the site is at an airport, 95% of the land cover is considered to be "transportation" (low roughness) and 5% "commercial/industrial" (high roughness).   For sites not considered to be at an airport, 20% of the land cover is "transportation" and 80% is "commercial/industrial".    The answer to this question may not be obvious for some tower locations. 

Recommendation:  While there may be a few cases where the tower location would not be correctly represented by the "airport" land use described above, in general, if the tower is located on airport property (away from buildings), the answer to the question above should be "yes".    Tower location can be reported either as UTM or LAT/LON coordinates.   If using LAT/LON, the data should be reported to at least the ten-thousandth digit (e.g., 39.8453, -89.9675).   The reported meteorological tower location should always be verified.   This is an important point worth repeating.   The reported meteorological tower location should always be verified and documented.  

3.  Seasonal Definitions  

Snow Cover:   AERSURFACE defines seasons by asking whether the area experiences continuous snow cover, either for at least one month or for most of the winter.   Because snow cover inhibits the amount of convection in the atmosphere, AERSURFACE users should caution not to overestimate the amount of snow cover assumed in the model.  
  	
Recommendation:   Local climatological data (LCD) reports can help determine an answer to this question for specific seasons and months.  They are available free for .gov, .edu, .mil, .k12 domains online at  http://cdo.ncdc.noaa.gov/qclcd/QCLCD.   Daily snow/ice cover totals are reported.   For months with days with snow cover (1 inch or more), AERSURFACE should be run twice; once with continuous snowcover and once without snowcover.   The values are then weighted based on the snow cover days / no snow cover days ratio.  The application can be processed on a seasonal or monthly basis.  An example below shows the calculation for an albedo monthly application, which may be the most common approach.   The following approach should be used for Bowen ratio and surface roughness also.  The five-year period of meteorological data should be examined when answering this question.  
Example:  For calendar year 2006 at station XY, the following information is gathered from LCD reports;

January - 	10 days with 1 inch or more or more of snow cover

Number of days in January   -   31 

Two separate AERSURFACE runs provide the following albedo information

Run without continuous snow cover -  0.15
Run with continuous snow cover  -	0.33

To determine the value to be included in AERMET Stage 3 for January albedo values use:

(21 * (0.15) + 10 * (0.33))     =    0.21 
                  31 

Where 21 is the number of days without snow cover, 10 is the number of days with snow cover,  

4.  Arid Region/Surface Moisture	

AERSURFACE will prompt the user to answer the question "Is this site in an arid region?"    If continuous snow cover is assumed for the area, this question will not be asked.   Region 5 does not have arid regions so AERSURFACE will ask the user to characterize the surface moisture conditions at the site relative to climatological normals.  The moisture conditions impact the Bowen ratio value generated by ASF.   Bowen ratio, which is the ratio of sensible heat to latent heat, can enhance or inhibit convective conditions in the atmosphere.  Wet conditions will result in a lower Bowen ratio, which will dampen the convective conditions in AERMOD.  Similar to the snow cover recommendation discussed on page 5, users should caution not to overestimate wet conditions for use in ASF.  

Recommendation: 	Start with the default assumption that moisture conditions are average.   If the precipitation information and/or soil moisture data indicate an exceptionally "dry" or "wet" season or year, that time period should be processed as wet or dry in ASF.   There is considerable discretion in categorizing the moisture conditions for an area.   

Useful information can be found in the AERSURFACE User's Guide, which recommends surface moisture condition can be determined by comparing precipitation for the period of data to be processed to the 30-year climate record, selecting "wet" conditions if precipitation is in the upper 30th-percentile, "dry" conditions if precipitation is in the lower 30th-percentile, and "average" conditions if precipitation is in the middle 40th-percentile.   The 5-year period precipitation, judged on a year-to-year annual basis, should be examined and compared to the area climate record.   The LCD reports can help answer this question.    

Additionally, thirty-year monthly and yearly climate data on soil moisture can also provide useful information.   For example, the website http://www.cpc.ncep.noaa.gov/soilmst/w.shtml is a National Weather Service product which contains climatological soil moisture normals on a monthly basis that can be compared to specific monthly values.   Drought index archive maps are available from  http://www.cpc.ncep.noaa.gov/products/monitoring_and_data/drought.shtml  and http://www.ncdc.noaa.gov/temp-and-precip/drought/historical-palmers.php.  

5.  Domain Issues	

The AERSURFACE methodology uses a 1 km default domain for determining surface roughness values around the tower location.  (A beta version of AERSURFACE has been developed that bases the radius for determining surface roughness on internal boundary layer considerations.  This tool has not yet been released.)  As the AERSURFACE User's Guide notes, significant discrepancies have been identified between station coordinates reported by NCDC and the actual location of the meteorological tower for NWS sites.  It is very important to verify coordinates of the meteorological towers.   The default domain for Bowen ratio and albedo used in the AERSURFACE methodology is a 10km by 10km box centered on the tower site.   

The surface roughness domain can be subdivided into at most 12 sectors (i.e., 30 degrees each).   For many areas, fewer sectors may suffice depending on the homogeneity of the 1km area surrounding the tower.  Other factors such as nearby water bodies, can influence the number and orientation of the sectors.   The AERMOD Implementation Guide document available on SCRAM provides a substantial discussion of the methods recommended for determining land use characteristics.  

Recommendation:  	A site specific evaluation should be conducted to determine the most appropriate number of sectors and their orientation.  As a default, the 1 km domain with twelve 30 degree sectors should be adequate for the surface roughness domain.  Twelve sectors may be more than needed for many NWS tower sites.   If the site encompasses a distinct land use variation, such as a coastline, additional guidance on buffer zone recommendations is available in the AERMET User's Guide. 

The 10km by 10km albedo and Bowen ratio domain should be centered on the tower location for most applications.  However, for applications primarily involving tall stacks in flat terrain, it may be appropriate to consider using a domain representative of the application site for Bowen ratio and albedo (see Section 3.1.2 of the AERMOD Implementation Guide).  If an application site is used as the center of the 10km by 10km domain, the rationale should be documented and justified.   

Lastly, it could be useful to compare the NLCD categories and the AERSURFACE model results to aerial photos, or other site-specific information, to determine if the results are reasonable.    

V.  Representativeness
   
Many Region 5 states have opted to provide pre-processed met data for use in regulatory AERMOD modeling.   While this provides a useful service to applicants, and to a large degree, simplifies the review process for the reviewing authority, it is important to review representativeness of the pre-processed data when being used in AERMOD.   For most areas in the Midwest, terrain influences are minor and historically, the nearest NWS site has been determined to be the best representative station available.   Because surface characteristics can vary significantly even within areas of similar terrain, it's important to examine whether the surface characteristics of the source are reasonably representative of   the surface characteristics of the meteorological site.   Surface stations located further away from a source, or even in a neighboring state, may be more representative than the nearest NWS station.         
 
