This document discusses the potential benefits of using meteorological data from the MM5 and RUC2 models as inputs for air quality modeling with CALPUFF. MM5 and RUC2 data provide higher spatial and temporal resolution than traditional weather station data, allowing CALPUFF to make more accurate simulations with less interpolation. While these prognostic meteorological models improve modeling accuracy, some regulatory agencies may not be familiar with the data formats. Overall, the document argues that MM5 and RUC2 data meet EPA modeling guidelines and can provide greatly improved air quality modeling results compared to using only surface weather observations.
Potential Benefits and Implementation of MM5 and RUC2 Data with the CALPUFF Air Modeling System
1. Environmental solutions delivered uncommonly well
Potential Benefits And Implementation
of MM5/RUC2 Data with
the CALPUFF Air Modeling System
Paper No. 04-A-314-AWMA
Prepared By:
Jeff A. DeToro - Project Manager
TRINITY CONSULTANTS
12770 Merit Dr.
Suite 900
Dallas, TX 75251
+1 (972) 661-8881
trinityconsultants.com
2. Potential Benefits and Implementation of MM5/RUC2
Data with the CALPUFF Air Modeling System
ABSTRACT
In April 2003, U.S. EPA officially revised the federal guidelines for air quality modeling
(40 CFR Part 51, Appendix W) to incorporate CALPUFF as the preferred dispersion
model to study long-range pollutant transport for regulatory purposes. At the absolute
minimum, CALMET (CALPUFF’s meteorological preprocessor) requires hourly
measurements of surface meteorological data and twice-daily upper air data soundings
(along with terrain and land-use characteristics) as input to generate three-dimensional
gridded fields of meteorology to be used in CALPUFF. These measurements come from
a variety of sources, some of which are traditional/well-known, while others are new and
emerging.
During the past decade, a new source of meteorological data suitable for dispersion
modeling applications has resulted from the emergence of next-generation weather
forecasting models. Meteorological output from prognostic mesoscale forecast systems,
such as the Penn State/NCAR Fifth-Generation Mesoscale Model (MM5) and Rapid
Update Cycle Model Version 2 (RUC2), provides superior input data for long-range
pollutant transport studies. This is due to the increased spatial and temporal resolution of
atmospheric data fields. Prognostic mesoscale forecast systems operate by assimilating
all available data and by using meteorological models to result in a consistent analysis of
the current atmospheric conditions. This initial analysis is used as the starting point for
numerical forecasting of conditions at some hourly intervals in the future. It is the
analysis of the current conditions that is the desired information to use as input to the air
quality models.
Meteorological model analysis data allow CALMET to derive the meteorological input
fields for CALPUFF much more accurately than would be possible from traditional
National Weather Service data sets. As a result, the analysis data enable CALPUFF to
provide greatly improved modeling results. CALPUFF modeling that uses
meteorological model analysis data meets the requirements of the revised federal
modeling guidelines; therefore, associated modeling results would be acceptable to
regulatory agencies.
3. INTRODUCTION
CALPUFF Modeling System
In April 2003, U.S. EPA officially revised the federal guidelines for air quality modeling
(40 CFR Part 51, Appendix W) to incorporate CALPUFF as the preferred dispersion
model to study long-range pollutant transport for regulatory purposes.1
CALPUFF is a
non steady-state, Lagrangian dispersion model that provides significant improvement in
atmospheric motion simulation, especially across large horizontal distances (i.e., 50
kilometers or more) when compared to the traditional steady-state, Gaussian dispersion
models. Although CALPUFF’s main regulatory use thus far has been limited mostly to
the analysis of pollutant impacts on federal Class I areas, advances in computer hard
drive storage capacity and processor speed now also allow the air quality community to
utilize CALPUFF for short-range modeling studies. In fact, it is likely that CALPUFF, or
another as-yet-to-be developed Langrangian model, will eventually replace Gaussian
models altogether as the preferred tool for all regulatory dispersion modeling endeavors.
CALMET is the preprocessor component of CALPUFF that generates the necessary
three-dimensional meteorological data fields for the main model. At a minimum,
CALMET requires hourly observations of surface meteorological conditions and twice-
daily observations of upper air conditions, both of which are collected by the National
Weather Service (NWS). This type of meteorological data is commonly referred to as
“diagnostic” since the observations describe, or “diagnose”, the current meteorological
conditions. These observational meteorological data are combined with geophysical data
inputs that characterize terrain and land-use for input to CALMET.
The primary shortcoming associated with meeting only the minimum data requirements
of CALMET is that the NWS network of meteorological data collection lacks sufficient
horizontal, vertical, and temporal resolution to accurately depict atmospheric motions in a
typical CALPUFF modeling domain, which may be hundreds of kilometers in length and
width. When data resolution is lacking, CALMET must rely more heavily on
interpolation techniques to estimate wind, temperature, and turbulence fields throughout
the domain, which reduces the overall accuracy of the modeling meteorology and,
subsequently, the modeling results.
Meteorological Model Analysis Data
A second type of meteorological data that is available to the air quality community is
generated by “prognostic”, or predictive, meteorological forecast models. These models
use actual observations in conjunction with basic equations of atmospheric physics to
predict a wide variety of additional meteorological parameters that are not routinely
measured. These parameters are determined at many vertical levels over a uniformly
spaced horizontal grid that covers a wide geographic area (i.e., on a continental scale).
Prognostic mesoscale forecast systems operate by assimilating all available data and by
using meteorological models to result in a consistent analysis of the current atmospheric
conditions. This initial analysis is used as the starting point for numerical forecasting of
conditions at some hourly intervals in the future. It is the analysis of the current
conditions that is the desired information to use as input to the air quality models
4. (referred to henceforth as “meteorological model analysis data”). As might be expected,
the number of calculations needed to generate such data sets requires significant
computing resources; thus, this activity is normally accomplished in a government or
university setting. Two of the more well-known prognostic meteorological data models
are the Penn State/National Center for Atmospheric Research (NCAR) Fifth-Generation
Mesoscale Model (MM5) and the Rapid Update Cycle Model Version 2 (RUC2).
MM5 is the latest version of a forecast modeling system that was developed at Penn State
University in the early 1970s to primarily predict mesoscale atmospheric phenomena
(i.e., those that occur on the order of tens of kilometers).2
The model is currently
supported by NCAR and has been widely available for use by air quality modelers over
the past decade. MM5 can be initialized with forecast output data from any one of a
number of different weather prediction models (including RUC2, which is described
below). Many regulatory CALPUFF analyses have been performed using MM5 data in
recent years. In fact, a few regulatory agencies have become comfortable enough
working with the data set to archive entire years of MM5 meteorological data specifically
for CALPUFF modeling in their respective regions.
RUC2 (released in 1998) is the latest version of a newer forecast model prediction system
that was developed to support users who require short-range weather forecasts.3
RUC2 is
operated under the supervision of the National Centers for Environmental Prediction
(NCEP). As the name indicates, this model is rapidly updated with the latest
observations from a number of sources, including weather satellites, upper air balloons,
NWS surface stations, wind profilers, Next-Generation Weather Radar (NEXRAD),
commercial aircraft, and buoys/ships. RUC2 data can be utilized as a stand-alone data
source for atmospheric dispersion modeling, or as an initialization database for MM5
forecast output. Because it is relatively new and has significant data-handling
requirements (extremely large files), RUC2 has not yet gained widespread use in the air
quality community.
The following sections describe the main benefits of using MM5/RUC2 meteorological
data sets in a dispersion modeling context and outline certain issues that have been
identified regarding their practical implementation. This paper is not intended to be an
exhaustive analysis of specific data set features, nor an in-depth technical discussion on
appropriate CALMET settings when processing these data sets. Rather, the author
provides basic information and awareness about MM5/RUC2 to interested parties.
MM5 BENEFITS
When compared to traditional meteorological data, MM5 offers many benefits to the air
quality modeler, especially greatly increased horizontal resolution of meteorological data
points. For example, North Dakota has 10 NWS surface weather-observing stations that
collect meteorological data suitable for dispersion modeling and only one NWS upper-air
observing station. In traditional modeling efforts, data from these stations would be used
in an attempt to characterize atmospheric conditions over the entire 69,000 square-mile
area of the state. Obviously, there would be many important weather phenomena that
would be “missed” by this sparse set of observations. Because MM5 is designed to
5. simulate mesoscale meteorological phenomena, users have the ability to generate
meteorological data from MM5 for input to CALMET at gridpoints spaced only 36
kilometers (km) apart, and as small as 1.33 km apart over smaller areas. This would
allow one to set up a modeling domain over North Dakota with more than 100 surface
meteorological data points (with 36-km spacing) covering the study area. In this
example, MM5 produces at least a ten-fold increase in resolution over NWS data for
CALMET to derive the necessary meteorological data fields. Increased resolution
produces better air quality analyses.
A second consideration is vertical resolution. While NWS balloon soundings record
basic meteorological variables (i.e., temperature, pressure, winds, and humidity) at
approximately 20 vertical levels above ground with every sounding, each MM5 data
point can include over 50 vertical levels of meteorological data, which is more than
double the number of readings from the soundings. This increased vertical resolution
allows for much better characterization of the atmosphere’s vertical structure at any given
time, especially in the boundary layer, which is where the majority of pollutant dispersion
occurs. The MM5 data sets also allow modelers to overcome a significant shortcoming
that has been associated with NWS soundings for many years, which is the forced
extrapolation of hourly vertical conditions from soundings that are only conducted twice
a day (usually near sunrise and near sunset when the atmosphere is transitioning the
most). Because MM5 data contain vertical profiles for each hour, the simulation of the
atmosphere during the hours that occur in-between NWS soundings can be greatly
improved when utilizing this data source in CALMET.
Another factor is the breadth of meteorological parameters available to the air quality
community. Traditional meteorological observations are somewhat limited with respect
to winds, temperature, turbulence, and moisture measurements. Since weather forecast
models are run by supercomputing systems, the algorithms contained within the models
can be significantly more complex and robust than those used in typical dispersion
models. These complex algorithms require a wide range of weather parameters as inputs,
many of which are incorporated into MM5 data sets, including (but not limited to):
• Vertical Heat Fluxes
• Precipitation Types/Amounts
• Available Moisture
• Wind Vector Components
• Atmospheric Instability Parameters
It is important to note that the MM5 data also include all of the parameters collected by
NWS stations.
To summarize, it should be again noted that the MM5 system was originally designed
with an emphasis on mesoscale atmospheric simulation. As such, the output from this
model is uniquely suited to characterizing pollutant dispersion in air quality studies. All
of the benefits of MM5 outlined above would primarily allow CALPUFF to greatly
6. reduce the amount of interpolation required when deriving meteorological fields, thus
minimizing errors in the calculations.4
MM5 IMPLEMENTATION
As a practical matter, MM5 data sets are currently being utilized in a number of
regulatory modeling studies with CALPUFF. Some regulatory agencies (in the Pacific
Northwest, for example) even offer MM5 data sets to the air quality community, which is
a step that had been taken earlier with meteorological data files for older models (such as
ISCST3). Modelers can easily customize the MM5 data set (i.e., use data pertaining to a
specific forecast model initialization and a specific modeling domain region) depending
on the requirements of a particular application. Other than the time investment needed to
become familiar with MM5 data and learn how to properly manipulate the data for input
to the dispersion models, there is generally little/no cost to obtain and process the data.
There are a few items to keep in mind when considering MM5 data as a potential source
of meteorological data for dispersion modeling. The data sets can be significantly larger
(typically on the order of several megabytes) than traditional meteorological data files;
thus, increased computing resources may be needed to run, store, and manage MM5 data.
One must also be aware of the increased preparation/processing time and CALMET
execution time that may be required. Although MM5 has been used in regulatory
applications for the past few years, many regulatory agencies have not yet worked with
MM5 data, so they would not likely have the computational resources to properly review
a CALPUFF analysis that uses MM5 data. It is also important to note that, although a
great deal of interpolation is eliminated when using MM5 in conjunction with CALMET,
some data interpolation may still be required if the underlying forecast model used to
initialize the MM5 data set does not generate hourly output.
Although dispersion modeling with MM5 is relatively new, U.S. EPA has documented
some basic guidance pertaining to meteorological model analysis data use in regulatory
applications. These include (but are not limited to):
• Only forecast model initial analyses (i.e., 0-hour data) should be used to derive
meteorological data sets for modeling, not forecast predictions.5
• It is not recommended that individual meteorological model data points be
represented as “observations” or “pseudo-stations” in CALMET without also
including traditional observations. However, U.S. EPA will allow certain
CALMET adjustments to be made (i.e., use small radius of influence values to
adjust diagnostic winds in Step 2 to preserve prognostic winds) as a measure of
compensation. 5
• Modelers utilizing meteorological model analysis data (such as MM5) may be
allowed to conduct their regulatory air quality studies with less than the
“traditional” five years of meteorological data that is normally required when
modeling solely with NWS data.
7. RUC2 BENEFITS
As with MM5 data sets that are initialized using output data from other forecast models,
RUC2 (either alone or in combination with MM5) offers a number of the same benefits
over traditional meteorological data. Since RUC2 is specifically designed to provide
accurate, short-term weather forecasts, RUC2 has the added advantage of more frequent
(hourly) output than most forecast models, which may offer output on only a 3- or 6-hour
basis. Thus, the RUC2 hourly representations of atmospheric conditions further reduce
the amount of interpolation required by CALMET to derive the hourly meteorological
fields for CALPUFF. RUC2 can also provide higher horizontal data resolution over most
MM5 applications since its meteorological grid can be derived with spacing as small as
20 km.
Another consideration is the number of actual meteorological observations that are
assimilated into RUC2. RUC2 includes many observations that may not be incorporated
into other forecast models, including (but not limited to):
• NEXRAD-measured Winds
• Wind profiles from VHF Wind Profilers
• Satellite Data
• Temperature and wind soundings from commercial Aircraft
• Observations from buoys and ships
This feature allows RUC2 to provide higher-quality estimates of certain meteorological
parameters than other models. For example, NWS surface winds are measured by
anemometers that have a starting threshold (generally between 0.5 and 1.5 meters per
second) below which the observations are subject to inaccuracies.5
Ironically, these light
or calm wind conditions often result in worst-case pollutant impacts due to the lack of
atmospheric dispersion. Since NEXRAD Doppler radar is a remote-sensing platform
(i.e., it does not involve physical instrumentation), NEXRAD wind observations are
typically more accurate at lower speeds. RUC2 is also able to better characterize upper
air conditions since its underlying observations (e.g., 6 or 10-minute NEXRAD-measured
winds) are much more frequent than the twice-daily NWS soundings used in other
models. As a result, transient boundary-layer phenomena such as nocturnal low-level jets
are more likely to be incorporated into RUC2 data, and subsequently CALMET, during
processing. These boundary-layer phenomena can be important in atmospheric
dispersion.
These additional benefits of RUC2 over MM5 data sets that are initialized using other
forecast model output allow CALMET to further reduce the amount of interpolation
required when deriving meteorological fields, even further minimizing errors in the
calculations.6
RUC2 IMPLEMENTATION
To summarize, the main operational advantage of using RUC2 in dispersion modeling is
that it incorporates more “actual” observations than any other meteorological data source.
8. RUC2 data sets are particularly well suited to simulate rapid changes in atmospheric
motion over short time scales due to its design as a short-term forecast model.
Although it appears that RUC2 is poised to become the future standard in meteorology
for air pollution modeling, it has some limitations. Due to the extremely large amount of
data that are assimilated into RUC2 output, the computing resources required to run,
store, and manage these data sets are currently cost-prohibitive for most users in the air
quality community. At the time of this writing, only one vendor (Software Solutions and
Environmental Services Company [SSESCO] of St. Paul, MN) currently offers
CALMET-ready RUC2 at a premium price. Unlike most forecast models, RUC2 is
presently a limited area model (i.e., its domain does not cover the entire planet);
therefore, data may not be available for certain geographical areas. It is also important to
note that RUC2 was designed to forecast synoptic (as opposed to mesoscale) weather
conditions; therefore, its boundary layer characterization may not be as accurate as MM5
in some cases.
See the bulleted list in the “MM5 Implementation” section of this paper for a summary of
basic U.S. EPA guidance when utilizing RUC2. Since it is a meteorological model
analysis data set, this same collection of guidelines applies to RUC2.
CONCLUSIONS
The recent updates to the federal guideline for air quality modeling (40 CFR Part 51,
Appendix W) specifically included a number of references regarding the use of
prognostic meteorological data for long-range transport and complex wind situations.1
In
addition, the revised guideline emphasizes that the ability of a meteorological data set to
accurately construct boundary layer profiles/three-dimensional fields and minimize
reducible uncertainty in modeled results are key criteria when selecting meteorology for
dispersion modeling. U.S. EPA indicates a clear preference in the revised guideline
towards the use of mesoscale meteorological data for regulatory dispersion modeling by
stipulating that a shorter time period can be modeled (minimum of three years, not
necessarily consecutive) versus traditional meteorological data (five consecutive years) to
derive accurate results.1
It is clear that meteorological model analysis data sets such as
MM5 and RUC2 represent the best available means to satisfy these criteria in a
regulatory modeling setting. As these data sets evolve and become more widespread, it is
expected that official, detailed guidance will be developed by the air quality community
to ensure that meteorological model analysis data sets are properly used by CALPUFF
modelers.
ACKNOWLEDGMENTS
The author would like to thank a number of Trinity Consultants peer reviewers (Bruce
Turner, Dr. Erwin Prater, Ryan Gesser, and Tony Schroeder) who reviewed of the draft
manuscript. The author would also like to acknowledge the efforts of the Lignite Vision
21 Program, Barr Engineering Company, Russell Lee, and Gale Biggs and Associates in
providing the basis for this study and helping to develop much of the information
presented within.
9. REFERENCES
1. Code of Federal Regulations, Title 40, Part 51, Appendix W (Guideline on Air Quality
Models), July 1, 2003.
2. MM5 Community Model Homepage (www.mmm.ucar.edu/mm5/overview.html).
3. SSESCO Weather Data Archives, RUC2 Overview
(www.ssesco.com/weatherdata.html).
4. Memorandum to Jeff Burgess (Lignite Vision 21 Program) from Eric Edwalds (Barr
Engineering Company), Russell Lee, Gale Biggs (Gale Biggs and Associates), and
Jeff DeToro (Trinity Consultants), MM5 Meteorological Data for CALPUFF
Modeling, March 25, 2003.
5. Report to the North Dakota Department of Health (NDDH) from ENSR International,
Response to NDDH Comments on ENSR’s Initial CALPUFF Submittal, March 2003.
6. Memorandum to Jeff Burgess (Lignite Vision 21 Program) from Eric Edwalds (Barr
Engineering Company) and Jeff DeToro (Trinity Consultants), RUC2 Data
Application for CALPUFF Modeling, May 22, 2003.