1. NUCAPS Usage in Operational Forecasting
AMS Short Course:
GOES-R and JPSS Preview for Users
January 10, 2016
Daniel Nietfeld
Science and Operations Officer (SOO) NOAA-NWS-WFO Omaha
Deputy Team Lead for Product Readiness & Operations (PRO) NOAA/NESDIS/GOES-R

2. Hopeful Takeaways
The Appeal of Hyperspectral Profile Retrievals (NUCAPS Soundings)
Understanding of what these are (and aren’t)
Issues for Forecasters to be aware of
Possible uses
Examples
Software to obtain, display, modify, and compare

3. “Day in the Life” of a NWS Forecaster in the Midwest
Convection is a common forecast problem
Thermodynamic environment is KEY
Accustomed to looking at the 12Z RAOB, with density of ~ 2 per state

4. Upper Air RAOB Network

5. Day in the Life of a Forecaster in a Midwest WFO
Convection is a common forecast problem
Thermodynamic environment is KEY
Accustomed to looking at the 12Z RAOB, with density of ~ 2 per state
During the pre-convective, early afternoon, I modify the 12Z RAOB for current surface conditions, and try to modify it for any changes in the airmass (from upstream)

6. Modified for 19Z changes in
T and Td
CAPE: 460 J/KG
CAPE: J/KG

7. 19Z Profile Supported Development of 4 Violent Tornadoes

8. Day in the Life of a Forecaster in a Midwest WFO
18Z (or 19Z) Special RAOB is a luxury
I don’t have to guess about the airmass changes
I typically still need to tweak the surface conditions due to the sensitivity to dewpoint
Only valid for the point of the RAOB
We occasionally get an Aircraft observation
I look at all of my data with some sense of the margin of error (and I try to learn what that margin of error is).
Observations from instrumentation
NWP

9. Imagine having thousands of 18-19Z soundings!!

10. Along Comes NUCAPS from the Suomi NPP Satellite
NOAA Unique CrIS ATMS Processing System
Operational CrIS+ATMS physical retrieval algorithm
NUCAPS vertical temperature and moisture profiles are available from NPP operationally
Example NUCAPS Skew-T Profile in AWIPS-II NSHARP
Observed Radiosondes
Example NUCAPS Coverage
Parcel LR
Temperature
Dew Point

11. Along Comes NUCAPS from the Suomi NPP Satellite
18-19 UTC!
NOAA Unique CrIS ATMS Processing System
Operational CrIS+ATMS physical retrieval algorithm
NUCAPS vertical temperature and moisture profiles are available from NPP operationally
Example NUCAPS Skew-T Profile in AWIPS-II NSHARP
Observed Radiosondes
Example NUCAPS Coverage
Parcel LR
Temperature
Dew Point

12. 4 consecutive passes over CONUS
This shows 4 consecutive passes. North of 40 N there's overlap from one pass to the next. But there are regions over the southern US -- small regions -- that are unsampled. Of course, these regions differ from day to day as the orbits shift eastward from one day to the next.
4 consecutive passes over CONUS
overlap north of ~40N

13. NPP Polar Orbiting Global Coverage

14. NUCAPS NOAA Unique CrIS ATMS Processing System
CrIS: Cross-track Infrared Sounder (1305 channels)
ATMS: Advanced Technology Microwave Sounder (22 channels)
All instruments on Suomi/NPP
East Coast: 05z/17z
Plains: 07z/19z
West Coast: 11z/23z
Svalbard
Downlink
NSOF
(NDE)
NWS
Gateway
SBN
CONUS Data Flow
WFO

15. Which CrIS channels are used for physical retrieval step? 399 of them
24 for surface temperature
87 for atmospheric temperature
62 for water vapor
Channels chosen most sensitive to one gas and least sensitive to other gases
All ATMS channels (22 of them) are used in NUCAPS
NUCAPS soundings are produced using a regression and a physical retrieval. The Regression step uses coefficients trained from cases that used all CrIS/ATMS data and ECMWF model output on select days in the past to find the best statistical relationship between the satellite data and the profiles of Temperature and moisture (“Best” being subject to the constraints of the weather on the test days, of course). That training yielded a set of static coefficients that are used with present-day regressions. Regressions with the data use *all channels*, so they capture more vertical structure than in the physical retrievals that uses a subset of CrIS channels.
When a set of soundings is being produced, regression is done first with all NUCAPS data to get temperature and moisture profiles, these data are used to do cloud-clearing – more on that in a minute.
Then the physical retrievals are done, sequentially. The physical retrieval is done using a subset of
399 of more than 1300 CrIS channels, shown on this chart. In the physical retrieval, the algorithm sequentially solves for temperature, moisture, ozone and other gases. The first guess for this is climatology: Think about what that means if you are in a regime that deviates significantly from normal. There are 24 channels (highlighted in green) from which surface temperature is retrieved, 62 (in red) from which water vapor is retrieved, etc.
The channels selected are most sensitive to one gas while being less sensitive to other gases. That is, they have high (but not perfect) spectral purity.  All 22 ATMS channels are used in NUCAPS. The ATMS retrieval is the solution if either the cloud clearing fails or if the CrIS Physical retrieval fails.
Reference for how atmospheric data can be obtained from various wavelengths:
COMET’s Module on “Advanced Satellite Sounding: The Benefits of Hyperspectral Observation - 2nd Edition”

16. Effective vertical resolution
(Tropical Western Pacific)
(Southern Great Plains)
NUCAPS resolves about 9-10 vertical levels of water vapor, and about 20 layers of temperature, in the atmosphere. The charts show the averaging kernel functions for temperature and it shows how the effect of one CrIS spectrum might be spread out in the vertical.  There is some geographic variability to the kernels.
E.Maddy, IEEE 2008
NUCAPS resolves about 9-11 vertical levels of water vapor in the entire atmosphere
NUCAPS resolves about 20 vertical levels of temperature in the entire atmosphere

17. Summary of products from NUCAPS (and AWIPS-II)
gas
Precision
d.o.f.
Interfering Parameters
Sensitivity
Temperature Profile, T(p), SST, LST
1.5K/km
6-10
Emissivity, H2O, O3, N2O
surface to ~1 mb
Water Profile, H2O(p)
15%
4-6
CH4, HNO3
surface to ~300 mb
Cloud Top Pressure
Cloud fraction
25 mbar,
1.5K, 5% 2 18
CO2, H2O
surface to tropopause
Ozone, O3
10% 1+
H2O, emissivity
Lower stratosphere
Carbon Monoxide, CO
 1
H2O, N2O
Mid-troposphere
Methane, CH4
1.5%
H2O, HNO3, N2O
Carbon Dioxide, CO2
0.5%
H2O, O3, T(p)
Sulfur Dioxide, SO2
 50%
< 1
H2O, HNO3
Volcanic flag
Nitric Acid, HNO3
emissivity
H2O, CH4, N2O
Upper troposphere
Nitrous Oxide, N2O
 5%
H2O, CO
This slide shows what is available from NUCAPS Soundings -- and what is in AWIPS II (in the red box). Think of Degrees of Freedom as something like layer resolution -- that is, there are approximately 6-10 layers of temperature resolved in the profile, and 4-6 layers of moisture. Do not expect to see very thin layers in a NUCAPS sounding.
These degrees of freedom are valid for the Physical Retrieval that moves around/alters the Regression first guess.
Chemical species in NUCAPS that aren't available in AWIPS can be accessed via NOAA's CLASS.
O3 = Ozone, CO = Carbon Monoxide, CH4 = Methame
CO2 = Carbon Dioxide, SO2 = Sulfur Dioxide, HNO3 = Nitric Acid
N2O = Nitrous Oxide
Ammonia = NH3 = 7.5*BT(867.75)/(BT(861.25)+BT(873.5)) mg/m^2
Clarisse, L., C. Clerbaux, F. Dentener, D. Hurtmans and P.F. Coheur
Global ammonia distribution derived from infrared satellite
observations. Nature Geoscience v.2 p
IASI products:
HNO3: -1.4 K/100% at 879, -1.17K/100% at
N2O: -0.44H/5% at 1300, K/5% at
CFCl3(F11): 0.08/10% at 847
CF2Cl(F12): 0.124/10% at 922
SO2: -0.2K/1000% at 1345,1371 cm-1
CCl4: K/10% at 795

18. Atmospheric Soundings from NUCAPS
Operational since 2013
Resolution over the entire atmosphere:
20-layer temperature resolution
9-11 layers water vapor resolution
Best resolution in mid-troposphere
First Guess : NCEP-based climatology and GFS Surface Pressure
Generally, the best resolution is mid-tropospheric. The 1st guess is derived from climatology.
A microwave-only retrieval will be done (for temperature and moisture only – no trace gases) if the physical retrieval using CrIS or if the cloud-clearing fails.

19. How are Sounding Produced
Regression uses static coefficients derived from ECMWF output for selected focus days (used to create a first guess/cloud-clearing field that uses climatology).
Regression (for cloud-clearing) uses all CrIS and ATMS channels
Retrieval minimizes Observation-Calculation value
If the environment really deviates from normal, the retrieval has a lot of work to do to converge to a solution.
If this doesn’t converge to a solution, sounding is still produced, but flagged
Retrieval uses a subset of channels
Shape-preserving; adjusts regression to the solution
Final Sounding is computed on 100 levels that are standard for a Radiative Transfer Model
More levels than warranted by Sounder resolution
You see these 100 levels in AWIPS 2
A physical retrieval is used to produce the sounding, minimizing the difference between observations and calculations. If the environment really deviates from normal, the retrieval has to do a lot of work to converge to a solution. If that convergence to a solution doesn't happen, the sounding is still produced, but flagged.
What is OBS-CALC? The OBS Vector is a cloud-cleared radiance spectrum (using the subset of CrIS radiances). The CALC vector is from a radiative transfer model that uses the regression solution – trained with the ECMWF – and climatology.
The final output is mapped to 100 levels that are standard for a radiative transfer model. You'll notice the 100 levels in AWIPS are always the same pressure surfaces. It's really more levels than are resolved.

20. How do NUCAPS soundings appear?
NUCAPS Sounding Points overlain over a Visible Suomi NPP Image
This shows the NUCAPS Sounding points plotted on top a VIIRS .64 micron image. Note that NUCAPS sounding data do not extend to the edge of the swath.
Plotting soundings over data like this allows you to anticipate where soundings might have difficulties because of clouds.

21. NUCAPS Sounding Points overlain over a Visible Suomi NPP Image
4 Points to examine:
NC in clear air
We'll look at 4 different soundings. one in NC in clear air..

22. Clear Air, North Carolina
The sounding in clear air over North Carolina is smoothly varying as is typical for atmospheric soundings, with a well-mixed boundary layer
Clear Air, North Carolina

23. NUCAPS Sounding Points overlain over a Visible Suomi NPP Image
4 Points to examine:
NC in clear air
NC near cloud edge
KY is deep clouds
One in deep clouds over Kentucky

24. The sounding in deep clouds probably failed to converge, and an error flag is available. AWIPS software is being modified so that you know the sounding is bad before you look at it. Soundings such as these that look non-physical are most common in regions of thick cloudiness, so use NUCAPS soundings with caution in such regions. That's a good reason to plot the sounding locations - the green dots -- over satellite information.
Upper level information for this sounding -- such as the tropopause height -- may still be useful.
Deep Clouds, Kentucky

25. Forecast Problem: 29 May 2014 Afternoon/Evening Convection ?
GFS 9 hour fcst
From 12UTC 29 May
Valid 21 UTC 29 May
1915 UTC
Visible
This is a more forecast-specific use of the soundings. Will that broken cumulus field on the border of Iowa/Nebraska develop? The GFS is predicting QPF later in the day.
Will the Cumulus Field develop further? (As suggested by the GFS)

26. NUCAPS Soundings give thermodynamic information at a convenient time!
Suomi NPP overflies the central US around 18 or 19z; this provides timely information on the state of the atmosphere in early afternoon. Here we have three different scenes, zooming in from North America to the Plains to WFO Omaha.
Let's look at the sounding in eastern Nebraska, circled in red.
1842 UTC NPP pass

27. NUCAPS Sounding 1842 UTC 29 May 2014 Would you expect convective development?
T = 79F (METAR = 80F)
Td = 60F (METAR = 63F)
CAPE ~ 50 J/KG
This sounding has a surface temperature and dewpoint close to observations (so we needn't edit the sounding to bring it into accord with observations). CAPE is small. Would you expect convective development?

28. Late afternoon visible imagery shows that the cumulus field has dissipated.
2354 UTC
Visible

29. June 3, 2014 High Risk Severe Weather Event in Omaha
A couple days later, Omaha experienced severe weather. This image shows the storm reports, damage, and two NUCAPS sounding sites from 1849 UTC on 3 June. Note also the gradient in dewpoint in the plotted METARs.
The next page shows the NUCAPS sounding for the site just south of the Omaha WFO -- the one in the Green Box.

30. June 3, 2014 High Risk Severe Weather Event in Omaha
NUCAPS sounding locations
Overlaid with VIIRS 0.64
1905Z June 3, 2014
Location of OAX in yellow
Northern dot is within a few
KM of KOAX, but under
cloud cover
Southern dot is in a nearly
Cloud-free location, and is
warmer and more humid
OAX
Zooming in again, we see the sounding from the cloudy region -- boxed in green, and also a second sounding to the south where skies are clearer. (And dewpoints are warmer there too)

31. June 3, 2014 High Risk Severe Weather Event in Omaha
NUCAPS ~10 KM south of OAX
1849Z June 3, 2014 – cloudy
scene, but still gives info!
Modified for surface METAR
Ob of T=83, Td=63
SB CAPE = 1851
The NUCAPS sounding from the site in the green box on the previous page -- a cloudy region, but the sounding still gives information. This sounding has been modified so the surface METARs match the sounding values. Surface-based CAPE is 1851.

32. June 3, 2014 High Risk Severe Weather Event in Omaha
NUCAPS sounding locations
overlain on VIIRS 0.64
1905Z June 3, 2014
The boxed site is the one we just looked at. Note that a clear Sounding is just south -- and recall that there is a strong dewpoint gradient in the region.

33. June 3, 2014 High Risk Severe Weather Event in Omaha
NUCAPS sounding
40 km south of OAX
1849Z June 3, 2014
Unmodified
SB CAPE = 686
Here is the unmodified NUCAPS sounding at that southern site. Surface-based CAPE is around but the surface temperature/dewpoint in the sounding do not match the METARs. This is a good sounding to edit.

34. June 3, 2014 High Risk Severe Weather Event in Omaha
NUCAPS sounding
40 km south of OAX
1849Z June 3, 2014
Modified for surface METAR
Ob of T=85, Td=68
SB CAPE = 3095
Once the sounding is edited so that its lowest temperature/dewpoint match the METAR, surface based CAPE exceeds 3000! This kind of information should influence your interpretation of present conditions.

35. Issue #1: Smoothing Vertical resolution is a bit course
~20 temperature layers
~10 moisture layers
Significant smoothing
Identification of warm capping layers?
Identification of dry layers aloft (downburst potential)?

37. Issue #2: Surface/BL Modification
Modification is necessary 99+% of the time due to errors in surface T and Td
Techniques, such as SPC’s SFCOA, have been used to objectively modifying the low levels of a sounding (RAP) using METARs
Automation of Sounding Modification at the Surface and in the BL – ongoing project:
“Improving NUCAPS Soundings for CONUS Severe Weather Applications via Data Fusion” - Dan Lindsey PI

38. Issue #3: Clouds/Rain Errors
Extra caution/scrutiny is needed
Excited about
the recent
improvements!

39. Quality Control - Problems
Life isn’t always perfect in the world of satellite profile retrievals…
Microwave or IR retrieval failures/errors
Regression failures/errors

40. New color scheme based on QC Flag Status (in development on AWIPS2)

41. Why not use the NWP sounding?
Sometimes do, but subject to NWP issues/errors
Soundings within model convection
Convective Parameterization Schemes result in unrealistic profiles

42. Real vs. Modeled Observed GOES Visible HRRR 2-hr forecast 2 hr fcst
from 17Z run
1900 UTC June 16, 2015
Atmosphere with clear, blue sky
1900 UTC June 16, 2015
Atmosphere with deep convection

43. Cold Air Aloft Application – Impacts for Aviation
Dual-Regression Retrievals for March 200hPa

44. Formal Evaluation of NUCAPS in NOAA’s Hazardous Weather Testbed 2015 Spring Experiment
Norman, Oklahoma (OU/NOAA Facility)
5 NWS forecasters, 1 broadcaster per week (30 total; and PI’s)
5 weeks in May and June, 2015
“Can NUCAPS data provide unique value to the severe weather nowcast and warning process?”

45. NUCAPS in 2015 HWT Timing of profiles Most common uses in HWT
East: ~
Central: ~ Z
West: Z
Most common uses in HWT
Analysis of pre-convective environment
Asses instability, boundaries, etc
Analysis of near-storm environment
Comparisons with NWP, RAOBS
Plus ~75 minute latency to AWIPS-II
Sfc/near-surface modifications to profiles necessary in most cases
Clear-sky selections recommended
In general, forecasters felt that, when modified, the profiles provide an adequate and useful representation of the current state of the atmosphere …
… leading them to see the value in having this information to fill the spatiotemporal gaps that exist in observed sounding information.

46. Blog Post: “NUCAPS Sample” May 12 - Pocatello, ID
“The instability seems a little high, but it could be localized.  Will see how the thunderstorms in the area develop over the next few hours….. “
“This thunderstorm moved over the sampled area about two hours later.  It peaked at about dBZ Composite Refl ”
“With our office between ROAB sites, having the NUCAPS soundings will be a good way for us to get a handle on the conditions in our area.”

47. Blog Post: “NUCAPS Sample” May 13 - Midland, TX
Modifed NUCAPS Soundings
Storms moving east into this environment continued to develop and strengthen
Back west, no new development in environment characterized by weaker instability and less moisture
“I liked it and thought it was useful to have today. My office is in between RAOB sites, so it would be nice to have this additional environmental information.”

48. Blog Post: “West Texas Soundings” May 19 – Midland, TX
00Z MAF RAOB
12Z MAF RAOB
19Z NUCAPS near MAF
“The drying of the air at mb since 12Z is reflected by intermediate NUCAPS soundings.”
“The NUCAPS soundings are a good way to see changes in the airmass since the RAOB soundings have been taken.”

49. Blog Post: “NUCAPS compared to Observed IAD sounding” June 8 - Sterling, VA
“18 UTC NUCAPS sounding near IAD, modified for IAD surface data showing 2200 J/kg when compared to 900 J/kg in the observed IAD 18 UTC sounding. The observed sounding also shows an elevated mixed layer and capping near 825 mb which is not seen in the NUCAPS sounding.
Observed 18z RAOB
“I'm still a little suspect of the NUCAPS data as it doesn't show the fine scale detail that is so valuable in a standard RAOB.”
NUCAPS
“Fusing of all the sources is really the way to go, they should all be blended together, instead of having to use them all (NWP, NUCAPS, etc)”

50. NUCAPS Feedback
All participants answered that they understand the differences between space-based soundings and RAOBs
Only 1 NWS participant already uses NUCAPS at home office (Alaska)
20/23 say they will 5%
28%
74%
28%
Daily Survey Q
26%
Daily Survey Q
30% 8%

51. NUCAPS Feedback
General shape and stability/moisture parameter values seemed realistic
Comparable to observed soundings
Important features and details such as capping inversions not depicted well (or at all) in the soundings
Stable layer sometimes apparent (bump); how to interpret this was unknown
Surface/ML modification often necessary, too cumbersome
“Automating the modifications would be great, including the low-level mixing”
QC Flags a must
“QC flags would give me more confidence in the soundings, as it is difficult to judge with just the cloud data.”
Various AWIPS-II requests
Training requests
More severe app examples
Verification statistics

52. Future of NUCAPS in the HWT
2016 Spring Experiment
Code upgrade
Additional satellites
QC flags
Additional visualization options
2017 Spring Experiment
Evaluate automated sfc modification NUCAPS project

53. Summary
Produced with every Polar Orbiter pass, time latency is about minutes
Afternoon Soundings:
16z on East Coast
18z in mid-continent
20-21z along West Coast
Morning Soundings
07z on East Coast
09z in mid-Continent
11z along West Coast
~10 moisture layers and 20 temperature layers in the vertical
Familiarize yourself with what information the soundings can yield
Learn how to modify the soundings if the lowest layers don’t agree with nearby METARs
A mix of infrared and microwave information. Regression can fail, but a sounding will still be produced. Ongoing work to add status flag to AWIPS 2 display.
A summary of this training is on this slide. This concludes the training on NUCAPS sounding data in AWIPS 2.

54. Summary Best uses in clear air or scattered – broken clouds
Convection
Cold air aloft
Less useful in precipitation and thick, uniform cloud cover
Not useful for determining precipitation type in winter weather
Be aware of QC errors and unrealistic profiles

55. How can I utilize these “soundings” ??
Improvements in this area are coming… but try these:
SHARPpy:
RAOB:
NPROVS:
(NPROVS best for archives and comparisons, not real time)
free
commercial – not free
free

56. Obtaining the Data Data available via NOAAPORT CLASS (data site):
Obtaining NUCAPS data from CLASS:

57. Web Links to Learn More
NUCAPS full technical description:
Hazardous Weather Testbed NUCAPS Blog:
NPROVS comparison software:

58. Web Links to Learn More
COMET’s Module on “Advanced Satellite Sounding: The Benefits of Hyperspectral Observation - 2nd Edition”
NUCAPS Training Developed for NWS: