NOAA Chemical Weather Forecasting – "Early Start"

Implementation Plan
Based on a Planning Meeting Held in Silver Spring on 18 April 2001
and Subsequent Discussions among
Jian-Wen Bao (ETL), Fred Fehsenfeld (AL), Georg Grell (FSL), Jim Meagher (AL), Bill Neff (ETL), Ken Schere (ARL), and Michael Trainer (AL)

Overview

NOAA is launching a major new initiative related to chemical weather forecasting. This research program, leading ultimately to an operational national air quality forecasting system, will be a collaborative effort that is built on past work and will involve other federal agencies (most notably the EPA) and the private sector. NOAA's Office of Oceanic and Atmospheric Research (OAR) has initiated a chemical weather forecasting research program, i.e. the "Early Start" research program. Additional research on air quality forecasting has been included in the NOAA/University of New Hampshire AIRMAP (Atmospheric Investigation, Regional Modeling, Analysis and Prediction) program. Research activities supported under these research programs are described in the following four tasks:

• Task 1 includes research related to the evaluation of current air quality forecast models using data collected during past field campaigns. The intent is to use the detailed data collected during these experiments to perform a diagnostic process-level evaluation of model performance, both in terms of the individual modules and of the overall modeling system. The diagnostic evaluation will go beyond a comparison of modeled and measured pollution fields (e.g. measured chemical intermediates used to test mechanisms, dynamical measurements used to test meteorological model) to identify specific causes for shortcomings in model performance. This task will provide an assessment of the forecast capabilities of current modeling technology and point to areas where improvement can be made.

• In Task 2 two photochemical grid models (MCNC and FSL) will be run in a forecast mode for four areas in the eastern U.S. Real-time and archived ozone data will be used to compare the performance of these models with statistical models used by the local air quality forecasters. This is intended as a test exercise to gain experience in providing daily ozone forecasts and identify and fix any logistical problems that may arise. In addition, the comparison of forecast and measured ozone levels will provide an initial indication of model skill in an operational mode and compliments the more detailed diagnostic evaluation planned for Task 1.

• Task 3 describes the participation of OAR laboratories in the development of chemical forecast capability within the Weather Research and Forecast (WRF) model. The work in this task builds on activities in Task 1. The diagnostic evaluation in Task 1 will test specific parameterisations and process representations and will be used to highlight areas where current chemical forecast models need improvement. This information will guide the developments proposed for the WRF. The focus of these two coupled activities will naturally fall on evaluation of current model technology (Task 1) in FY 01 with an increased emphasis on WRF development (Task 3) on the out years.

• The NOAA/AIRMAP consortium provides an opportunity to augment the feasibility study described in Task 2 and to initiate work on forecasting infrastructure needs related to data ingest and display. One of the goals of the AIRMAP program is to serve as a test bed for regional chemical forecasting by providing all the elements of a forecast "system", including the observations needed to evaluate and improve forecast skill. These activities are described briefly in the "Related Activities" task.

The planned tasks, combined with input from customer groups, provide a foundation for future NOAA work aimed at developing a chemical forecasting system. The multi-year effort will assess model capability and forecast skill while identifying areas where improvement in the forecast system is needed including those needed to extend the forecast beyond ozone.

Task 1 – "Now"

Evaluate and Assess the Forecasting Capabilities of Current Models Using Existing Data from Comprehensive Field Studies

Objectives

Using data from comprehensive field campaigns, evaluate the ability of current air quality models to reproduce the temporal and spatial distribution of secondary pollutants like ozone. First, each of the major model components will be evaluated separately including:

• Emissions model – does the emissions model properly quantify atmospheric emissions from biogenic (forests and soils), area, and point sources?

• Meteorological model – how well do the mesoscale meteorological model describe the atmospheric dynamics in terms of wind fields, temperature structure, radiation, and mixing height?

• Chemical model – is the rate of transformation and composition of chemical intermediated and end products properly characterized by the model?

Once the evaluation of the individual modules has been completed, the performance of the completely integrated chemical transport model will be evaluated.

Approach

During the period 1994 – 2000 NOAA participated with the Southern Oxidants Study (SOS) in three major field campaigns designed to improve our understanding of the atmospheric processes that control the formation and distribution of ozone and fine particles in the atmosphere. The campaigns were:

1. SOS Nashville 1994/1995

2. SOS Nashville 1999

3. TEXAQS2000

Each of these studies included: (1) an extensive network of ground-based chemical and meteorological measurements; (2) remote sensing and balloon-born sensors to provide profiles of meteorological and chemical parameters; and (3) instrumented aircraft equipped with state of the art sensors for chemical and meteorological measurements. This extensive array of measurement capability allowed the physical and chemical characteristics of the atmosphere to be measured to a level of detail that had not hitherto been possible.

These rich data sets will serve as the basis of the proposed evaluation of current air quality models. The spatial and temporal coverage coupled with the process-level information will insure that the sources of any model errors can be identified. Models to be evaluated include:

• ARL/EPA MODELS3 System

• FSL coupled chemistry/meteorology model

• AL regional chemistry model

Each of these models is built around a common synoptic meteorological model (the NCAR/Penn State MM5 model). However, the model structure and chemical module are different in each model.

The evaluation process has already begun with existing funding from EPA (ARL) and NOAA (FSL, ETL, and AL). The Nashville99 data set is currently being used to evaluate the MM5 performance. The "Early Start" funds will be used to accelerate and expand the evaluation and to provide a closer integration of the research being conducted in North Carolina (ARL) and Colorado (FSL, ETL, and AL).

Initial work will focus on an evaluation of the mesoscale meteorological model, MM5, used in the air quality forecast models. The quality of meteorological forecast is critical to that of air quality forecast. There are many atmospheric processes that control or strongly affect the evolution of emissions, both gases and aerosols. The process parameterisations used in current meteorological models have been optimised for weather forecast skill. The combination of parameterisations that provide acceptable weather forecast skill (e.g., quantitative precipitation forecasts) may result in poor air quality forecast skill. Therefore, it is very important to provide information on the errors associated with different parameterisations, individually and in combination, and ways to correct them. Such information will lead to recommendations for meteorological parameterisations to be used in the real-time forecasts (Task 2), and what improvement needs to be incorporated into the future chemical forecasting models (Task 3).

There is a direct and obvious coupling between this task and Task 3, WRF development. It is anticipated that, as the model evaluation in Task 1 proceeds, shortcomings will be identified in the existing air quality models. This information will, in turn, guide the development efforts in Task 3. The same data sets described above will prove invaluable in the development of improved process modules and approaches for inclusion in the WRF.

Deliverables

* Proposed peer-reviewed publication

Task 2 - “Next”

Develop and Evaluate Prototype Forecasting and Information Transfer Capabilities and Tools

Objectives

Using several existing air quality simulation modeling systems, prototype ozone forecasts will be made for four geographic areas each day for the summer of 2001. The goals of this exercise are to test and demonstrate the capabilities of current and emerging air quality modeling systems for use in short-term forecasting on a regular basis, to develop protocols and apply them for demonstration of air quality forecast skill, and to test information transfer/delivery systems of the forecast information in a timely manner and useful form for potential users.

Approach

NOAA’s Forecast Systems Laboratory (FSL) and MCNC (RTP, NC) have air quality modeling systems capable of simulating ambient ozone concentrations over regional to urban scales. . The FSL system includes the MM5 meteorological model with “on-line” chemical capability. In this system the chemical kinetic mechanism is embedded within the meteorological model structure, and thus the chemical-transport is performed as part of MM5. Biogenic emissions are also integrated on-line since they are strongly modulated by meteorology. The MCNC system, on the other hand, includes separate models for meteorology (MM5), emissions (SMOKE emissions model), and chemical-transport (MAQSIP-RT air quality model). These separate models are all loosely coupled in an on-line manner. The principal difference between these systems is the degree of integration of the components. Plans include the use of these two systems to provide twice-daily 36-hr forecasts of maximum 1-h and maximum 8-h ozone concentrations during the period 1 July through September 30, 2001 over the eastern United States, with specific local-scale emphasis over Texas, North Carolina, East Tennessee and New Hampshire.

A protocol for evaluation of the short-term chemical forecasts using real-time ozone monitoring data from national/state/local monitoring networks (e.g., AIRS) and on-going NOAA ozone studies (e.g., AIRMAP, ETOS) will be developed using a skill-score or equivalent approach. The short-term meteorological forecasts will also be evaluated independently, since meteorological forcing has a primary influence on air quality. Parameters of interest include wind speed and direction, temperature, clouds/radiation, precipitation, boundary layer depth, among others. Real-time data from NWS will be used, as well as other relevant NOAA measurements available in real-time. An assessment will be made of the ability of the models to simulate the ozone metrics, as compared to observational data, as well as their ability to meet or exceed the forecast skill of local-scale methods (e.g., statistical techniques, persistence) used by air quality managers in the four target regions.

Forecast products will be placed on a password-protected NOAA research website, and will be made accessible to project scientists and collaborators in NOAA and the participating state agencies (TX, NC, and NH) for evaluation of their clarity, utility, and relevance.

FSL(1) plans include setup of semi-operational system to forecast weather/air-quality in real time, using an “online” MM5/chemistry model

• domain coverage including TX, NC, TN and NH

• horizontal resolution of 27 km

• twice daily 36 hr forecasts

• 3DVAR RUC analysis for FDDA and initial conditions

• timely availability of 1-hr and 8-hr max ozone values

• if needed, 3d v5d files will be made available

A second run will be made once a day that either uses different anthropogenic emissions input data (run in same configuration as above), or use a better resolution over one or two of the three forecast areas. Preparations will be made for very high resolution forecast runs (dx = 1-2 km) that will take place in FY02 over NH. This will include testing of real-time forecast runs at that resolution over selected summer periods during FY01. Assistance will also be provided to ingest real-time air-quality model forecast output into FX-Net.

MCNC(2) plans include the use of the MM5V3.4 mesoscale meteorological model, the SMOKEV1.3 emissions model and processing system, and the MAQSIP-RT multiscale Eulerian photochemical model. MM5 will be run using a CONUS+ coarse domain at 45 km grid spacing, and an inner 15 km domain covering most of the eastern 2/3 of the US and southeastern Canada. . Inside the 15 km grid will be several 5 km urban-scale nests: one centered on Houston, Texas and another on Portsmouth, NH. SMOKE and MAQSIP will make use of MM5 meteorological fields and be deployed using a 45 km grid covering the eastern 2/3 of the US and southern Canada. Inside this grid, two 15 km grids will be implemented, one covering much of the southern US south of the Mason-Dixon line, and another covering much of the northeastern US. Inside the southern 15 km grid will be a 5 km grid covering the Houston metro-area. Inside the NE 15 km grid will be a 5 km grid centered on Portsmouth, NH and encompassing Boston, NE Massachusetts, and parts of NH, Vermont, and Maine. Forecasts for NC will be based on the 15 km results as no additional 5 km grid is planned for NC.

Data to drive MM5 will be obtained from the objective analysis and boundary conditions produced by one of the NCEP operational models, most usually the ETA. As a backup, the AVN will be used when the ETA data are not available. MAQSIP will be self-cycled with surface ozone data- assimilation on the 00z run using data from about 900 surface monitors across the eastern 2/3 of the US. The emissions database will be based on the NET-96 inventory released by EPA. In addition, several modifications particular to the states of Texas and NC will be implemented. Also, BEIS-3 will be used for biogenic emissions.

Two forecasts will be run per day, one at 00Z and one at 12Z. Forecast products should be available from these runs by 3AM EDT and 1 PM EDT, respectively. The principal forecast products delivered will be peak 1-hour average and peak 8-hour average ozone depicted graphically. In addition, animated ozone and precursor plots and animated MM5 forecast plots will also be available.

Forecast evaluation protocol development, evaluations, and database development will be conducted at ARL-RTP(3).

Deliverables

Task 3 – "After Next"

WRF Development

Objectives

• Develop air-pollution prediction system based on WRF, the next generation Weather Research and Forecast (WRF) model, capable of simultaneously forecasting weather and air-quality

• Develop NOAA leadership role in WRF-chemistry model development

Approach

Since December 2000, a prototype meteorological version of WRF exists and is being run in real time to test its weather forecasting abilities. This prototype version of the model is in flux, with fundamental changes still occurring constantly.

Before implementation of chemistry or coupled chemistry/meteorology modules can begin, questions concerning the software design of WRF will have to be answered. Examples are software design issues related to map projections, model I/O, computational efficiency, accuracy, and precision. A meeting has been arranged between NOAA scientists, WRF-chemistry working group members as well as other WRF-team members, to address some of these issues and decide which further modifications are necessary on the meteorological prototype version of the WRF model, before chemistry implementation can take place. The modifications that are deemed necessary, should then be made by a collaboration of scientists from NOAA, NCAR, and other involved institutes.

Using this finalized model framework, a WRF-chemistry prototype version for ozone forecasting can then be prepared. This will require the inclusion of coupled meteorology/chemistry modules for subgrid-scale transport (such as convection or turbulence), a module for dry-deposition, as well as one for biogenic emissions. This first prototype will also require the inclusion of at least one module for the chemical mechanism (such as RADM2, CB4, RACM,…). We would expect this first prototype to be ready for testing by the end of FY02.

A workshop is planned for the latter part of 2002 to introduce the prototype model, to define the further directions for research and development of the modeling system, and also to define and discuss guidelines for evaluation and verification.

Deliverables (depend on planning meeting in May 2001, subject to change)

Objectives

Some of the activities to be conducted under the NOAA/AIRMAP cooperative agreement in FY 01 are directly linked to the research planned under the OAR "Early Start" research4. These tasks have the following objectives:

1. Provide ozone forecasts for New Hampshire during the summer of 2001.

2. Expand FX-Net data ingest and display capabilities to include real-time air quality data and forecast model output

Approach

The NOAA/AIRMAP cooperative agreement includes participation by two OAR laboratories (Aeronomy and Forecast Systems) and the University of New Hampshire, Plymouth State College, Mount Washington research foundation, and the New Hampshire Department of the Environment. The AIRMAP consortium in New Hampshire operates three research monitoring stations in the state:

1. Hyland Farms – located just outside Durham New Hampshire this site is mainly impacted by pollution that comes from nearby Portsmouth/Kittery and interstate 95 as well as long-range transport (presumably from NY/NJ and cities along the mid Atlantic corridor) off the Gulf of Maine.

2. Castle in the Clouds – a rural site in the middle of the state selected to characterize the rural background in New England.

3. Mount Washington – the site on the summit is often impacted from pollution transported from the industrial Mid West and southern Canada. The data also suggests that this site is occasionally in the Free Troposphere.

Ozone forecasting5 – The Forecast Systems Laboratory's high-resolution coupled meteorology/chemistry model will be used to provide 24- and 48-hr ozone forecasts for New Hampshire. The model forecasts will be compared against ozone observations from the AIRMAP and state regulatory networks to provide an operational evaluation of model skill. The measurements performed at the Hyland Farm site and as part of the 01 summer sea-breeze study will be used to perform a diagnostic evaluation of forecast performance for both meteorological and chemical parameters. These evaluations will be performed on a series of 3-5 case studies (3-5 days each), selected jointly by the FSL modeling team and the AIRMAP scientists involved in the sea-breeze study.

Forecasts for the 01 summer period will be performed using a reasonably coarse grid (dx = 10-30 km) to determine the impact of larger scale transport. Initial work will also be done in FY 01 to prepare for high resolution (dx = 1-2 km) simulations in FY 02 to support air quality forecasts and the planned ground-based field experiment.

FX-Net developments6 – FX-Net is a web-based version of the AWIPS workstation used by NWS forecasters throughout the U.S. A fully operational national chemical forecasting system will require specialized data ingest and display tools similar to those available to weather forecasters. The capabilities of FX-Net will be expanded to accommodate the ingest of real-time air quality data and chemical model output as a first step in developing this needed capability. Since the FX-Net systems have already been implemented under AIRMAP this program provides an ideal test bed for the prototype system. Three following activities are planned for the 01-02 timeframe:

• Local air quality data – Add local air quality parameters to the FX-Net data ingest as well as to the FX-Net display system (parameters to be determined after discussion with Robert Talbot)

• National air quality – Develop data ingest and display capability in FX-Net for visibility data from the 900+ station ASOS network. These data are acquired in real-time (every 20 minutes for many stations) at FSL providing a useful test of future national air quality data ingest and display capability, needs, and applications.

• Forecast model display – The FSL ozone forecast model output will be added to the FX-Net data ingest system and FX-Net client display system to support mesoscale forecasting and air pollution forecasting.

Deliverables