Title: Real-Time Applications of a CMAQ-Related Photochemical Model Community Modeling and Analysis System Workshop, US EPA, Research Triangle Park, NC Oct 21-23, 2002 by John N. McHenry, Jeff Vukovich, Carlie Coats, Ted Smith, Don Olerud, and Aaron Sims
1Real-Time Applications of a CMAQ-Related
Photochemical ModelCommunity Modeling and
Analysis System Workshop, US EPA, Research
Triangle Park, NCOct 21-23, 2002byJohn N.
McHenry, Jeff Vukovich, Carlie Coats, Ted Smith,
Don Olerud, and Aaron Sims The Environmental
Modeling Center at MCNC_at_ the North Carolina
Supercomputing CenterResearch Triangle Park, NC
27709-2889Contact mchenry_at_emc.mcnc.org
2Outline
- Abstract
- History of Air Quality Prediction at MCNC
- Operational Model Overview
- Computational Forecast Domains
- Forecast Product Overview
- New England Air Quality Study 2002
- Preliminary Results for the SE US
- Highlights from Operational MM5 Evaluation
- Conclusions
3Abstract
- For the 5th year, MCNCs Environmental Modeling
Center (EMC) has run an operational numerical air
quality prediction (NAQP) system in real-time. - The forecast system uses the Multiscale Air
Quality Simulation PlatformReal Time (MAQSIP-RT)
air quality model. - MAQSIP-RT derives from the original CMAQ
prototype co-developed by MCNC and EPA/NOAA-ORD
in the 1990s. - Forecasts were produced in support of NOAA-based
research toward an operational air quality
forecast system the New England Air Quality
Study (NEAQS). - Forecasts were also produced for a variety of
state-based operational forecast clients (North
Carolina, Alabama, Arkansas, Oklahoma, Texas) - Preliminary results are encouraging and suggest
that MAQSIP-RT is an important new tool in
understanding the evolution of real-time air
quality.
4A Little History Environmental Modeling at MCNC
- MCNC began as the Microelectronics Center of
North Carolina in Research Triangle Park, NC - MCNC is a private, non-profit institution
bridging the worlds of science and commerce,
providing concepts and transferring research
ideas into the marketplace - Environmental Modeling Center (EMC) staff of 25
meteorologists, atmospheric chemists, and HPCC
experts - The EMC is housed at the North Carolina
Supercomputing Center, which is on the MCNC
campus - The EMC historically maintains a close working
relationship with EPA RD/NOAA-ORD, RTP Office,
especially in the area of photochemical model
development, application, analysis, and
evaluation (e.g. CMAQ and MAQSIP)
5A Little History EMC at MCNC (cont)
- Environmental Modeling Center Staff and Hardware
located in the Supercomputing Building
6A Little History EMC at MCNC (cont)
- Supercomputing Building (North Carolina
Supercomputing Center)
7Operational Model Overview
- Models used
- MAQSIP-RT Photochemical Model
- New in 2002 Corrected earlier tendency toward
ozone over-prediction on marginal air quality
days, where partly cloudy skies have a
significant influence, by improving treatment of
cloud attenuation effects. - New in 2002 Activated AIRNOW ozone monitor data
assimilation system to initialize MAQSIP-RT - Sparse-Matrix Operator Kernel Emissions (SMOKE)
- BEIS-3,NET-99 Point/Area, Mobile-5
- All emissions online including point-source
specific plume rise - PSU/NCAR MM5V3.4 initialized with NCEP/Eta
Analysis - Computing dedicated 24p O3800 expansion planned
in Spring 2003
8Operational Model Overview
- How MAQSIP-RT differs from CMAQ System Aspects
- Ozone data-assimilation using AIRNOW monitor data
- Robust/redundant model initialization in
real-time - Automated generation of analysis/forecast
products - Fast enough to meet operational forecast
deadlines - Automated error-recovery
- High-temporal resolution input of meteorological
forecast data using the MCPL output module (see
Coats, Section 7, Paper 4 MCPL An I/O API
Output Module for MM5, Wednesday, October 23,
930AM) - Highly configurable nesting strategy
9Operational Model Overview
- How MAQSIP-RT differs from CMAQ Model Aspects
- Sophisticated optimized cloud package (KF-deep,
McHenry-shallow, resolved, sub-grid-layer all
interacting simultaneously) - Sophisticated photolysis
- Online spectral calculation of clear-sky J
- Advanced cloud-optics interactions with J
- Online calculation of fractional land-cover-based
dry deposition velocities - Specialized treatment of shoreline-grid-cell PBL
characteristics - Optimized advection, diffusion, and QSSA chemical
solvers - Fully parallelized using OpenMP
- Latitude/elevation/month lateral BCs for
coarse-most grid - High Temporal Resolution Time-stepped BCs for
all nests - Latitude/month diffusive top boundary condition
for all nests
10Computational Forecast Domains
11Computational Forecast Domains
- MAQSIP-RT grids windows into MM5 domains
- Example 45km East-of-Rockies Grid Showing
Terrain Height
45km MAQSIP East-of-Rockies Grid
12Computational Forecast Domains
- MAQSIP-RT grids windows into MM5 domains
- Example 15km NE Corridor Grid Showing Terrain
Height
15km MAQSIP NE Corridor Grid
13Computational Forecast Domains
- MAQSIP-RT grids windows into MM5 domains
- Example 15km SE US Grid Showing Terrain Height
15km MAQSIP SE US Grid
14Forecast Product Overview
- Forecast Products (Website Delivered)
- MAQSIP-RT Nested Model Runs
- 48-hour 45km-scale forecasts (Eastern 2/3 of US)
- 24-hour 15km-scale forecasts (NE, SE, Texas)
- 14- or 24-hour 5km-scale forecasts (NH,
Birmingham,Houston) - Animated Forecast Maps
- Chemical (MAQSIP-RT) Parameters
- Surface with Wind Vectors
- Ozone
- CO
- NOx, NOy
- Upper Air with Wind Vectors
- Ozone
15Forecast Product Overview
- Forecast Products (Continued)
- Animated Forecast Maps MM5
- Meteorological (60hr fcsts twice daily)
- Surface
- Precip (Convective/Non-convective)
- Isotachs/Vectors
- RH/T/Vectors
- MSLP 2m Td, T
- UPA
- 300/500/850mb heights/vectors/isotachs
- 500mb Abs Vort
- 700mb Height/RH
- Other
- Convective Cloud Tops PBL Height
- SkewTs at various stations
16Forecast Product Overview
- Forecast Products (Continued)
- Static Forecast Maps MAQSIP-RT
- Peak 1-hour Average Ozone colored according to
EPA categorical forecast codes (green/yellow/orang
e/red/purple) - Peak 8-hour Average Ozone colored according to
EPA categorical forecast codes (green/yellow/orang
e/red/purple) - Delivery Schedule
- 12z, 45/15km grids Peak 1-hour/8-hour maps
complete by 115pm EDT animations later - 00z, 45/15/5km grids All products by 430am EDT
17New England Air Quality Study (NEAQS)
Operational MM5 Forecasts
- MM5 results extracted at selected profiler sites
were supplied to Jim Wilczak at NOAA/FSL for
incorporation into experimental ensemble
temperature forecast product(s). MCNCs version
of the MM5 was run at 3 resolutions, depicted
above. The outer domain was typically run for
60-hours at 45km, twice daily. The large interior
(15km) domain was run for 48(36) hrs at 12z(00z),
and the highest resolution (5km) domains were run
for either 14hrs (Birmingham/Houston) or 24hrs
(New England) at 00z. The MCNC MM5 was
initialized with the Eta-EDAS AWIP3D (40km)
dataset and used a 6-hour dynamic initialization
based on gridded Eta forecast data. The 12z run
utilized 06z AVN (GFS) large-scale boundary
conditions, whereas the 00z run used 00z Eta
boundary conditions. No special meteorological
data was incorporated. The MCNC MM5 was
configured with the OSU-LSM using Eta-based
soil-moisture/soil-temperature fields, with
adjustments to account for differences in
soil-texture maps between the Eta and MM5
versions of the LSM. The MRF PBL scheme was used,
along with the KF deep convection scheme, the
RRTM long-wave scheme, the Dudhia short-wave
scheme, and the Reisner-1 microphysics scheme.
Output was provided in real-time through an
ftp-server to FSL in Boulder, CO.
18New England Air Quality Study (NEAQS)
Operational MM5 and SMOKE Forecasts
- In support of the ARL hybridized air quality
forecasting system, MCNC supplied meteorological
and emissions data to Ariel Stein at ARL. A
baseline data-stream was developed in the
May/June timeframe and operational datasets were
made available beginning July 1. These continued
through the end of the NEAQS. The meteorological
data was derived from the operational MM5
forecasts produced at MCNC. The emissions data
resulted from operational runs of the SMOKE
emissions processing/modeling system, driven by
outputs from the MM5 forecasts. (This allows
meteorologically sensitive emissions rates to be
affected by forecast meteorology.) Output was
provided via an ftp-server to ARL in Silver
Spring, MD. Because the same meteorological data
was used to drive the MCNC MAQSIP-RT model in
real-time, a direct comparison of the hybrid
model results and the MAQSIP-RT model results is
possible.
19New England Air Quality Study (NEAQS)
Operational MAQSIP-RT Forecasts
- A continuous archive of forecasts was produced
beginning July 1 and extends through Sept 30,
2002. Forecasts were run twice daily, at 00z and
12z. MCNC provided merged datasets at all three
scales to Stu McKeen at FSL for real-time plots
against NEAQS field data being collected at
surface locations and aboard the NOAA ship Ron
Brown. - Further, forecast maps and animations were made
available over the MCNC Web-site located at
http//emc.mcnc.org/ SECMEP/projects, under the
heading NOAA New England AIRMAP Field Study, for
any NOAA investigator. These maps were used by
Wayne Angevine (Aeronomy Lab), who directed a
profiler site near Durham, NH, and provided
forecasts to the Ron Brown. An example of this
type of plot is shown in Figure 3, compared
against verification as depicted by the gridded
surface-monitor-based observations disseminated
through EPAs AIRNOW Web-site (http//www.epa.gov/
airnow). - MCNC conducted the 00z 45km runs in parallel,
using (or not) surface monitor-based data
assimilation. In the data-assimilation algorithm
currently available at MCNC, hourly data from
more than 1000 surface ozone monitors east of the
Rockies is retrieved shortly after 7PM EDT. These
data are processed and error-checked before being
supplied to a Lagrangian transport model. The
Lagrangian model utilizes the 12z MM5 forecast
meteorological winds and daytime mixing depths to
advect the ratio of observed-to-forecast (from
the 12z run) ozone values each hour from each
monitor, starting at 10AM LDT. Each ratio-event
results in a downstream trajectory envelope which
is allowed to have a vertical extent consistent
with the forecast mixed layer depth. At the end
of the assimilation cycle, the mean ratio along
each trajectory envelope, including all
intersecting envelopes, is calculated, and then
applied to the first-guess ozone field at 00z
from the 12z forecast run. The resulting values
at 00z are used to initialize the forecasts.
20New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 5
- Aug 5 1st day forecast vs. AIRNOW obs
- 15km NE MAQSIP-RT
21New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 10 1st day forecast (left) vs. AIRNOW obs
(right) - 45km MAQSIP-RT
22New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 11 2nd day forecast (left) versus 1st day
forecast (right) vs. AIRNOW obs - 45km MAQSIP-RT
23New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 12 2nd day forecast (left) versus 1st day
forecast (right) vs. AIRNOW obs - 45km MAQSIP-RT
24New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 13 2nd day forecast (left) versus 1st day
forecast (right) vs. AIRNOW obs - 45km MAQSIP-RT
25New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 13 1st day forecast (left) versus vs. AIRNOW
obs - 15km NE MAQSIP-RT
26New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 14 2nd day forecast (left) versus 1st day
forecast (right) vs. AIRNOW obs - 45km MAQSIP-RT
27MAQSIP-RT Forecasts, 2002 New England Air
Quality Study (NEAQS) Preliminary Results Aug
10-18
- Aug 14 1st day forecast (left) vs AIRNOW obs
- 15km NE MAQSIP-RT
28New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 15 2nd day forecast (left) versus 1st day
forecast (right) vs. AIRNOW obs - 45km MAQSIP-RT
29New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 15 1st day forecast (left) vs. AIRNOW obs
- 15km NE MAQSIP-RT
30New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 16 2nd day forecast (left) versus 1st day
forecast (right) vs. AIRNOW obs - 45km MAQSIP-RT
31New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 16 1st day forecast (left) vs. AIRNOW obs
- 15km NE MAQSIP-RT
32New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 17 2nd day forecast (left) versus 1st day
forecast (right) vs. AIRNOW obs - 45km MAQSIP-RT
33New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 17 1st day forecast vs. AIRNOW obs
- 15km NE MAQSIP-RT
34New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 18 2nd day forecast (left) versus 1st day
forecast (right) vs. AIRNOW obs - 45km MAQSIP-RT
35New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 18 1st day forecast (left) AIRNOW obs
- 15km NE MAQSIP-RT
36New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- Aug 19 1st day forecast (left) vs. AIRNOW obs
- 45km MAQSIP-RT
37New England Air Quality Study Real-Time Forecast
Time-Series Example
- Example FSL-generated plot of multiple MCNC
MAQSIP-RT 15km forecasts versus measurement data
collected at the Harvard Forest Surface Site on
August 9-10, 2002 during the New England Air
Quality Study.
38New England Air Quality Study (NEAQS) Preliminary
Results MAQSIP-RT Forecasts Aug 10-18
- MAQSIP-RT shows good skill for many days
- Forecast patterns tend to improve with more
recent forecasts - Forecast time-series at NEAQS ground sites are
encouraging. - MM5 seems to overdo the sea-breeze on several
occasions
39MAQSIP-RT Forecasts, 2002 Aug 2-10 SE US
- 15km SE Day 2 Fcst vs Obs Aug 2
40MAQSIP-RT Forecasts, 2002 Aug 2-10 SE US
- 15km SE Day 2 Fcst vs Obs Aug 3
41MAQSIP-RT Forecasts, 2002 Aug 2-10 SE US
- 15km SE Day 2 Fcst vs Obs Aug 4
42MAQSIP-RT Forecasts, 2002 Aug 2-10 SE US
- 15km SE Day 2 Fcst vs Obs Aug 5
43MAQSIP-RT Forecasts, 2002 Aug 2-10 SE US
- 15km SE Day 2 Fcst vs Obs Aug 6
44MAQSIP-RT Forecasts, 2002 Aug 2-10 SE US
- 15km SE Day 2 Fcst vs Obs Aug 7
45MAQSIP-RT Forecasts, 2002 Aug 2-10 SE US
- 15km SE Day 2 Fcst vs Obs Aug 8
46MAQSIP-RT Forecasts, 2002 Aug 2-10 SE US
- 15km SE Day 2 Fcst vs Obs Aug 9
47MAQSIP-RT Forecasts, 2002 Aug 2-10 SE US
- 15km SE Day 2 Fcst vs Obs Aug 10
48RTOFS 2002 Early September Case Study
RTOFS Max 8-Hour Forecast (00Z Run)
Max 8-Hour Observations
RTOFS to Obs Comparison Max 8-Hour
Slide 11
49MAQSIP-RT SE 15km Grid, 2002 Preliminary
Conclusions
- MAQSIP-RT shows good regional skill, but is
biased somewhat high on some days near and
downwind of both Atlanta and Birmingham and
biased somewhat low on some days in NC and TN.
These urban-corridor biases tend to occur
simultaneously, suggesting either meteorological
or emissions inventory issues. - Urban and downwind urban corridor emissions
inventories, especially Ga, Alabama, NC and TN
may need improvement. - In general
- Subtle effects (presence/absence of shallow
convection, weak synoptic transitions, persistent
moisture pools at mid-layers) are significant if
not dominant in the SE for any given summer - SE US forecasting may be more challenging due to
largely mesoscale nature of the weather in
summertime
50MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)Parts of July and September
51MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)Parts of July and September
52MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)Parts of July and September
53MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)Parts of July and September
54MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)Parts of July and September
- July 7 SW 15km Grid Day 1 Fcst
55MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)Parts of July and September
- July 8 SW 15km Grid Day 1 Fcst
56MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)Parts of July and September
- July 9 SW 15km Grid Day 1 Fcst
57MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)Parts of July and September
- July 10 SW 15km Grid Day 2 Fcst
58MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)
59MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)
60MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)
61MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)
62MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)
63MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)
64MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)
65MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)
66MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)
67MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)
68MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)
69MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)
70MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)Parts of July and September
71MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)Parts of July and September
72MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)Parts of July and September
73MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)Parts of July and September
74MAQSIP-RT Forecasts, 2002 Texas, (Arkansas,
Oklahoma)Parts of July and September
75MAQSIP-RT SW 15km (Texas/Oklahoma/Arkansas) Grid,
2002 Preliminary Conclusions
- MAQSIP-RT shows very good regional skill in the
SW plains - Forecast for Dallas July 7-10 event outstanding
- Forecast for Houston Sept 11-14 event quite good
- Results for Arkansas and Oklahoma encouraging
76Operational MM5 Statistical Evaluation Highlights
- Evaluation by MOS Region
- Evaluation over whole Domain
77Operational MM5 Statistical Evaluation
Highlights Max Temperature
Eta/MM5 statistical plots for maximum
temperatures. The left plots represent Eta/MM5
(45-km) coincident sites, while the right plots
represent Eta/MM5 (15-km) coincident sites. The
center legend applies to both the left and right
plots.
78Operational MM5 Statistical Evaluation
Highlights Min Temperature
Eta/MM5 statistical plots for minimum
temperatures. The left plots represent Eta/MM5
(45-km) coincident sites, while the right plots
represent Eta/MM5 (15-km) coincident sites. The
center legend applies to both the left and right
plots.
79Operational MM5 Statistical Evaluation
Highlights QPF
Eta/MM5 statistical plots for six-hourly
precipitation. The left plots represent Eta/MM5
(45-km) coincident sites, while the right plots
represent Eta/MM5 (15-km) coincident sites. The
center legend applies to both the left and right
plots. In the legend, "ets" represents equitable
threat scores, and HK represents the Hanssen and
Kuipers ("true skill') score.
80Operational MM5 Statistical Evaluation
Highlights QPF
Number of sites in which the amount of observed
precipitation was more closely predicted by Eta
or MM5 for the first eight months of 2002. Note
that we polled a limited number of Eta forecasts
for January and February, most of which were
located in the region surrounding the Carolinas.
The bolded numbers indicate a closer match to
observations.
81Operational MM5 Statistical Evaluation
Highlights QPF
Average accumulated precipitation totals for each
analyzed month in 2002. The calculation was made
using all valid 6-hourly values for which MM5,
Eta, and observations were available. Note that
we polled a limited number of Eta forecasts for
January and February, thus the extremely low
averages. The bolded numbers indicate a closer
match to observations.
82MM5 Statistical Evaluation Highlights
- For the key forecast variables (i.e. max/min
temperature, precipitation), the EMCMM5 either
outperforms or performs comparably to the
NCEP-Eta model. Daily maximum temperature
prediction skill is essentially equivalent,
evaluated for both the 45km and 15-km EMC MM5
models and their domains. The EMC-MM5 is better
able to capture nocturnal temperature minima,
thus providing better minimum temperature
guidance to an operational forecaster.
83MM5 Statistical Evaluation Highlights
- Though the Eta categorical six-hourly (rain
versus no-rain) skill scores are marginally
better than the EMC-MM5, these scores do not
reflect the quantitative bias present in the Eta
model. The NCEP-Eta model is significantly worse
at predicting accumulated precipitation at the
12km scale than is the EMC-MM5 at 45km, being
biased high in amounts by about 29, as compared
to 21 for the EMC-MM5. The NCEP-Eta model also
over-precipitates when compared to the 15-km EMC
MM5, as reflected in a better overall bias score
for the EMC 15-km MM5.
84MAQSIP-RT User Testimonies
- George Bridgers (NC-DENR) Im just now getting
all the data into the office from around the
state yesterday, but the model has done an
outstanding job the past several daysreally all
year. July 31, 2002 - Bill Ryan (PSU) For future reference, the NAQP
performed excellently this week. It killed my
forecast 4 of 5 days. July 19, 2002 - Bryan Lambeth (TCEQ) Overall, the model
performance looks qualitatively good this year.
July 29, 2002 - Emily Harris (ADEQ) This is a wonderful tool
for our at-risk populations to assist in planning
their outdoor activities during ozone season.
Oct 7, 2002
85Overall Conclusions
- The MCNC operational MM5 is as good or better
than the NCEP Eta model in daily max/min
temperature forecasts for the entire year 2002
(to present). - The MCNC operational MM5 produces more reliable
amounts of quantitative precipitation than does
the Eta model for the entire year 2002 (to
present). - The MCNC operational MAQSIP-RT has become a
reliable numerical ozone forecast tool over much
of the nation east of the Rockies during summer
2002. - Subscriptions are available to any state or
agency involved in ozone forecasting.
86Contacts and Web Address
- Web address and Contact Information
- http//emc.mcnc.org/projects/SECMEP/index.html
- Multiscale MAQSIP-RT Forecasting Products
- Email mchenry_at_emc.mcnc.org
- Voice (919) 248-9237