Objective Satellite-Based Overshooting Top and Enhanced-V Signature Detection

Kristopher Bedka

Science Systems and Applications, Inc. (SSAI) at the NASA Langley Research Center

1 Meet Kristopher Bedka

• Researcher for 7 years at the Cooperative Institute for Meteorological Satellite Studies (CIMSS), University of Wisconsin

• From August 2009-present, researcher with the NASA Langley Research Center in Hampton, Virginia

Past and Current Research Interests - Objective detection and prediction of newly developing convective storms (i.e. convective initiation)

- Satellite-derived and their height assignment characteristics

- Satellite observed signatures of convectively-induced aviation turbulence

- Objective detection of the overshooting top and enhanced-V signature

- Satellite-derived height, vertical thickness, and cloud microphysical property retrievals and product validation

- Innovative satellite imagery/product visualization techniques 2 Meet Kristopher Bedka

• Researcher for 7 years at the Cooperative Institute for Meteorological Satellite Studies (CIMSS), University of Wisconsin

• From August 2009-present, researcher with the NASA Langley Research Center in Hampton, Virginia

Past and Current Research Interests - Objective detection and prediction of newly developing convective storms (i.e. convective initiation)

- Satellite-derived winds and their height assignment characteristics

- Satellite observed signatures of convectively-induced aviation turbulence

- Objective detection of the overshooting top and enhanced-V signature

- Satellite-derived cloud height, vertical thickness, and cloud microphysical property retrievals and product validation

- Innovative satellite imagery/product visualization techniques 3 Collaborations and Contributors To This Talk

• UW-CIMSS Contributors - Wayne Feltz: CIMSS Satellite Nowcasting and Aviation Applications Program management

- Jason Brunner: Overshooting top and enhanced-V product development

- Richard Dworak: Overshooting top comparisons with , cloud-to-ground , aviation turbulence, radar reflectivity, and CloudSat observations

- Lee Cronce: Software development for GOES-R program

• NASA Langley Contributors - Patrick Minnis: NASA and Radiation team program management

- Patrick Heck and Szedung Sun-Mack: Overshooting top implementation in NASA CERES cloud property retrieval package

• Martin Setvak: Invaluable insight and expertise on satellite observations of convective cloud top features

• Nikolai Dotzek and Peter Groenemeijer: Provided ESWD database and very useful feedback on SEVIRI OT detection study 4 Talk Outline

• Challenges and techniques for objective satellite-based overshooting top (OT) detection

• OT detection using ~11 μm IR Window temperature gradients 1) Methodology, comparisons with WV-IRW BTD 2) Product examples with NOAA/NASA imagery 3) MSG SEVIRI examples 4) Product validation using 1.5 year CloudSat OT database 5) Multi-year OT detection distribution over the US and Europe 6) Quantitative OT relationships with hazardous weather

• Objective enhanced (i.e. cold) V signature detection 1) Methodology 2) GOES/MODIS/AVHRR product examples 3) Product validation and relationships with severe weather

5 Motivation For Development and Users of Objective Satellite Overshooting Top and Enhanced-V Detection Products

• Due to the hazards associated with overshooting top and enhanced-V producing storms, objective detection of these features is a product requirement for the GOES-R Advanced Baseline Imager (ABI)

- Development of the OT/V detection algorithms have been underway for ~2 years

- The GOES-R program requires algorithms to meet temporal latency and product accuracy requirements

- Objective validation of these algorithms is a significant challenge since little “truth” information is readily available

- SEVIRI has comparable spatial resolution to the future GOES-R ABI, making its data a useful testbed for ABI algorithms

6 Motivation For Development and Users of Objective Satellite Overshooting Top and Enhanced-V Detection Products

• Other users of OT and Enhanced-V Detection products include:

- The NOAA Hazardous Weather Testbed via the GOES-R Proving Ground

- National Center for Atmospheric Research (NCAR) Graphical Turbulence Guidance-Nowcast (GTG-N) system

- NASA Clouds and Earth’s Radiant Energy System (CERES) cloud property algorithm

7 What is an Overshooting Top?

Overshooting Top: A domelike protrusion above a cumulonimbus anvil, representing the intrusion of an updraft through its and/or the (from the AMS Glossary)

•Small in size (< 15 km) and often short lived

•Indicates a storm with a very strong updraft, hazardous for aviation operations if a plane were to fly through an overshooting top (OT)

•Interaction of storm updraft with stable tropopause layer generates turbulent gravity waves which can propagate far away from their source region and affect air traffic

•Often indicates the region of a with the heaviest rainfall

•Responsible for injecting water vapor into the stratosphere. This has significant climate impacts because water vapor is an important greenhouse gas.

•A key component of the enhanced-V signature 8 Challenge: Detect Detailed Cloud Top Temperature Patterns Within Hazardous Storms Using Current Operational Geostationary IR Imagery

• The overshooting top and enhanced-V signatures are much less distinct in 3-4 km (at nadir) GOES-12/SEVIRI vs. 1 km MODIS IR imagery, making them more difficult to detect with an objective algorithm

MODIS GOES-12

9 9 Overshooting Top Algorithm Design Overview

Characteristics of Overshooting Tops in Satellite Imagery

Visible/Near-Infrared Characteristics 1) “Lumpy” texture relative to smooth anvil cloud and shadowing at high solar zenith angles in ~0.6 μm visible imagery

2) Variability in ice crystal effective radius and ~4 μm channel reflectance, with smaller ice crystals and higher reflectance indicating a severe storm (Rosenfeld et al. 2008).

Day/Night Infrared-only Characteristics 1) Positive upper-level water vapor (WV) - IR window (IRW) channel BT differences (BTD), due to tropospheric water vapor injected into the lower stratosphere by updrafts associated with overshooting convective cloud tops (Ackerman et al. 1996, Schmetz et al. 1997, Setvak et al. 2007)

2) Isolated region of very cold IRW BT relative to the surrounding warmer anvil cloud. Anomalously cold BTs are caused by persistent moist adiabatic ascent, allowing the BT to be much colder than any temperature that would be found in a co-located sounding or NWP model profile 10 OT Detection Using Visible + Near-IR Signatures

Increased texture Overshooting Top in visible channel at high solar zenith angle triggers false detections in Berendes et al. (JGR, 2008) or other visible texture based algorithms

11 11 IRW-Texture Overshooting Top Detection Technique

Description • Overshooting tops appear as small (< 15 km diameter) clusters of cold cloud pixels in ~11 μm IRW channel imagery

• Spatial gradients in IRW BTs (i.e. IRW-texture) can be used in combination with NWP tropopause information and BT thresholds derived from analysis of 450 OT/enhanced-V events (Brunner et al. (WAF, 2007)) to objectively identify OT pixels

Strengths • Offers a more consistent day/night OT detection capability with greater accuracy than the existing operational OT detection method based on the 6 to 7 µm water vapor (WV) minus 10.7 µm infrared window (IRW) channel brightness temperature difference (BTD)

• Small time latency, product can be produced over a US-sized domain ~45 seconds after algorithm code is executed

• Code is flexible and operates on both GEO rapid scan (5-min) and operational (15 to 30-min) scanning patterns in addition to LEO instruments (AVHRR/MODIS)

•Product shows strong relationships with occurrences of cloud-to-ground lightning, aircraft turbulence, maximum in-storm radar reflectivity, and severe weather reports 1212 Physical Description Overshooting Top Algorithm Processing Steps Summary

1) Read 10.7 μm IR Window (IRW) channel BT and NWP Tropopause Temperature

2) Identify pixels with IRW BT ≤ 215 K and NWP tropopause temp

1) Starting from the coldest BT in the list, ensure that no cold pixels are within 15 km of each other. Pixels that satisfy these criteria are called “candidate overshooting pixels”

2) Sample the surrounding anvil at an 8 km radius in 16 directions around the candidate overshooting pixel - 8 km is chosen to sample the region outside an OT, since OTs are generally < 15 km in diameter. We don’t want to sample an OT when computing mean surrounding anvil BT

1313 Physical Description Overshooting Top Algorithm Processing Steps Summary

5) Pixels at least 6.5 K colder than surrounding anvil are considered “overshooting center pixels” - These pixels are used as input within the enhanced-V detection algorithm

6) Within a search box centered on the OT center pixel, find pixels at least 50% colder than the surrounding anvil mean BT - We must identify the entire “top” and not just one pixel within the top

IRW-Texture Candidate Pixels: Anvil BT Sampling White Pinwheels: Candidate OT Significantly Colder Than Anvil MODIS 0.25 km Visible Black Pinwheels: Cloud Top BT Pattern Too Uniform=Not an OT Final OT Detection Pixels

1414 CloudSat Overpass Overshooting Top Physical Description Threshold Determination Minimum OT Temperature

Cumulative Frequency Diagram: 96% of the 450 Overshooting Top Min • The OT minimum IRW BT and IRW BTs were less than 215 K the mean surrounding anvil BT difference were manually computed for 450 enhanced-V producing storm events over the U.S. (Brunner et al. (2007))

• A 215 K Minimum OT BT would detect 96% of these 450 OT (black line) The frequency of OT minimum IRW BTs less than BT events (see dot on black line to values along the bottom x-axis scale for 450 “enhanced-V” described by Brunner et al. (2007). (grey line) The the left) frequency of the OT minimum BT and mean surrounding anvil BT difference less than BT difference values along the top x-axis scale for the same 450 thunderstorm database. Circles along the two lines represent criteria used in the IRW-texture OT detection method. 1515 Overshooting Top Physical Description Threshold Determination OT-Surrounding Anvil BT Difference

• A 6.5 K minimum OT minus Baseline mean surrounding anvil BT OT would detect 84% of all OT Algorithm Criterion is events in Brunner et al. 6.5 K database (see dot on grey line to the left)

• A reduced OT-anvil BT difference criterion (i.e. 4 K) would detect marginal OT (black line) The frequency of OT minimum IRW BTs less than BT values along the bottom x-axis scale for 450 “enhanced-V” events but with too many thunderstorms described by Brunner et al. (2007). (grey line) The false detections frequency of the OT minimum BT and mean surrounding anvil BT difference less than BT difference values along the top x-axis scale for the same 450 thunderstorm database. Circles along the two lines represent criteria used in the IRW-texture OT detection method. 1616 IRW-Texture Overshooting Top Detection Technique

Weaknesses • OTs are not well detected in storms with limited (< 8 km) anvil extent

• Marginal OT events with an overshoot magnitude of < 0.5 km and/or little to no horizontal IRW BT gradient will not be detected by this method

• The highest cloud tops are not always co-located with the coldest IRW temperatures (Setvak et al. (Atmos. Res. 2010)), especially for “older” OTs

• Since largest observed OT is < 15 km in diameter, algorithm settings dictate that OT detections must be > 15 km away from other OT detections. Only one of two OTs will be detected if the two are spaced less than 15 km apart

• Locations within a very cold, non-convective cirrus plume exhibiting significant texture in the IRW BT field can be misclassified as an OT

• Very cold cirrus anvil edges can be misclassified as an OT

Dependencies • Product depends on the use of a NWP tropopause temperature forecast, so the product will not operate (as accurately) if the NWP data is unavailable 17 Comparison Between MODIS IRW-Texture and WV-IRW BTD Output

IRW-Texture Method Detects Suggested MODIS WV-IRW Highly Textured Areas in BTD Threshold (2 K) Over- Visible Imagery Commonly Diagnoses OT Coverage Associated with OTs 18 18 Comparison Between MODIS IRW-Texture and WV-IRW BTD Output

MODIS 0.25 km Visible IRW-Texture OT Detections

Above Anvil Cirrus Plume

FALSE OT MODIS 1 km 10.7 Micron WV-IRW BTD Detections DETECTIONS

Bedka19 et al. (JAMC, 2010) GOES-12 OT Detection Product Animation

Severe Weather Reports2020 During the Animation Period Comparison Between OT Detections From Current GOES and Future GOES-R ABI

Blue: GOES-12 + MODIS-based Proxy ABI Detects

Red: MODIS-based Proxy ABI Detects Only

Comparison of time-matched MODIS with GOES-12 for 50 events shows that GOES-12 minimum IRW BT in the OT region was 12 K warmer than MODIS

As a result, fewer OTs are detected in GOES-12 than what will be detected in future GOES-R ABI or MTG data

21 21 MSG SEVIRI Overshooting Top Detection Examples

22 International Space Station Photograph Of A Thunderstorm With An Overshooting Top Over The Ivory Coast, 5 February 2008

Overshooting Top

3 km MSG SEVIRI 10.8 µm IR Window 1 km MSG SEVIRI Visible Image With 1 km MSG SEVIRI Visible Image Brightness Temperature Image Overshooting Top Detection

Overshooting Top Observed in Overshooting Top Observed in Objective Overshooting Top Space Station Photo Space Station Photo Detection23 European Severe Weather Database Quality Controlled Severe Weather Reports: 25 May 2009

QC1: report confirmed by reliable sources

QC2: report fully verified

24 SEVIRI OT Detections 25 May 2009

1000 UTC 1100 UTC

1200 UTC 1300 UTC

1400 UTC 1500 UTC

1600 UTC 1700 UTC

25 SEVIRI IRW-Texture OT Detection Example Low Countries 25 May 2009

OTs are detected throughout these storms, but some are missed for storm “A” which exhibits a cold ring signature

Would WV-IRW BTD have detected the storm “A” OT (see black arrow)?

26 IRW-Texture vs. WV-IRW BTD 25 May 2009 A

WV-IRW BTD > 0 K identifies the entire deep convective anvil cloud

WV-IRW BTD values are insignificant A for the OT signature in storm A (see white circle)

This illustrates the challenges of objective OT detection using IR-based techniques since the IRW BT WV-IRW and WV-IRW BTD are often well BTD correlated A

27 SEVIRI IRW-Texture OT Detection Example Italy/Slovenia 25 May 2009

28 Objective OT Validation

29 Qualitative Overshooting Top Detection Validation Using CloudSat

CloudSat Radar Reflectivity Cross Section True OT Location

GDAS Tropopause Height • The CloudSat-observed overshooting top has a diameter of 11 km and has a peak height 1.5 km above the GDAS tropopause

• The ABI detection algorithm Co-Located MODIS Data: 20080509 at 2315 UTC accurately identifies this feature (lower- left panel) in addition to several other probable overshooting tops not directly observed by CloudSat

True OT Location True OT True OT Detected By ABI Location Location Algorithm

CloudSat 30 MODIS 250 m Visible Overpass Proxy ABI 2 km IR Window BT 30 With OT Detections Qualitative Overshooting Top Detection Validation Using CloudSat

IRW-Texture

WV-IRW BTD

Data destriped by L. Gumley 31 (UW-SSEC) but stripes remain Overshooting Top Detection Comparison With CloudSat OT False Detect Without NWP OT Detected Without NWP

Use of NWP Tropopause reduces overall number of OT detects at the expense of missed detection of marginal events

IRW-Texture (With/Without NWP Tropopause Check)

32 CloudSat-based Overshooting Top Validation Approach

• 1.5 years (> 3000 granules) of CloudSat Cloud Profiling Radar (CPR) data was manually investigated to find direct OT overpasses – 1:30 PM overpass time precedes maximum convective storm activity

• CloudSat selected over CALIPSO because CALIPSO cloud top height product is limited to the GEOS-5 tropopause and CloudSat allows us to better see the vertical structure of the deep convection

• The highest cloud top must be at least 0.5 km above the surrounding anvil cloud for 2 adjacent CloudSat profiles to ensure the presence of a coherent OT signature.

• This implies that the smallest OT in the final OT database can be ~3.4 km in diameter in the along track direction (CloudSat spatial resolution=1.7 km).

• For daytime events, Aqua MODIS 250 meter spatial resolution visible imagery along the CloudSat overpass is also analyzed to further verify that CloudSat observed an OT.

• OT events that meet these criteria are assigned to the final OT database. 114 total 33 33 OT events were found CloudSat-based Overshooting Top Validation Approach

• CloudSat observed OTs are treated as both single coherent entities (i.e. “top regions”) and as the individual pixels that compose the top

• This distinction is important because a detection method may not identify every pixel within an OT along the CloudSat track but could identify some portion of the top region. This should be rewarded within the validation framework.

• The exact location and spatial bounds of an OT is subjectively determined through close examination of the CloudSat 2B-GEOPROF profile

• The following metrics are used to evaluate the accuracy of IRW-texture and WV-IRW BTD OT detections: 1) OT pixel false alarm rate (FAR) and 2) probability of detection (POD) of all possible top regions

• OT Pixel FAR=Number of incorrect OT pixel detections along CloudSat overpass Total number of OT pixel detections along CloudSat overpass

OT Top Region POD=Number of accurately detected top regions along CloudSat overpass Total number of top regions along CloudSat overpass 34 3434 Comparisons of OT Detection Performance Relative to CloudSat

Locations of 114 CloudSat Observed Overshooting Tops From April 2008 to September 2009 • OTs are detected more often over land than water

• Errant NWP tropopause temperature analysis over water could contribute to this

• Heymsfield et al. (2010) shows that the maximum upward vertical motion in land-based convection is of greater magnitude and at a higher vertical level than convection over ocean

• OTs over land would likely be of larger magnitude and would therefore appear more prominently in IRW imagery, leading to a higher CloudSat Observed OTs Detected By ABI Algorithm detection rate using the IRW-texture CloudSat Observed OTs Missed By ABI Algorithm method 35 Quantitative Validation Statistics: MODIS-based Proxy ABI

Percentage of Detected Pixels Outside • WV-IRW BTD averaged over a 3x3 box CloudSat OT Locations to minimize impact of residual striping on (FAR) results (Chung et al. 2008)

• IRW-texture meets GOES-R algorithm specification (< 25% FAR)

• Removal of NWP tropopause check increases POD and the expense of increased FAR. - Are these truly “errors” or just marginal OT events?

• As illustrated in previous graphics, WV- IRW BTD shows FAR that is comparable to POD

• WV-IRW BTD > 0 K more representative of a deep convective cloud mask than OT detection mask

36 Quantitative Validation Statistics: MODIS-based Proxy ABI

Percentage of Detected Pixels Outside CloudSat OT Locations (FAR)

• Global CloudSat OT database allows us to compare current vs. “future” algorithm performance. MET-7 and MTSAT cannot yet be processed in our framework

• The coarser resolution of current GEO reduces POD, but FAR is comparable to proxy ABI

37 Applications of OT Detection Output

38 Using Objective Overshooting Top Detections To Improve CERES Convective Cloud Top Height CERES: Clouds and Earth’s Radiant Energy System, A NASA/NOAA Climate Monitoring Instrument

Issue: In convective cloud updraft regions, OTs reach heights far above the tropopause, but the tops are often assigned to the NWP-defined tropopause height. Observed cloud top temperatures are significantly colder than any point in the NWP profile, so no realistic height can be assigned.

Solution: Use an objective overshooting cloud top detection algorithm to improve cloud top height assignment in these regions. Combine 8 K/km with difference between cloud top and MOA tropopause temperature to adjust the cloud top height upward in overshooting top regions

0.25 km Aqua MODIS Visible With 1 km Aqua MODIS Band 31 BT Overshooting Top Detections Colored By Congo: 3/19/2009 at 1225 UTC 0.25 km Aqua MODIS Visible Upward Height Adjustment Magnitude

3.0 2.5

2.0

1.5

1.0

0.5 0 39 Using Objective Overshooting Top Detections To Improve CERES Convective Cloud Top Height CERES: Clouds and Earth’s Radiant Energy System, A NASA/NOAA Climate Monitoring Instrument

CloudSat Overpass Shown Below • Height adjustment in overshooting top (OT) regions has been integrated into CERES Version 3 cloud software

• The example below shows that upward height adjustment better matches CloudSat heights Colored Circles: Overshooting Top Detections

* Original CERES Height **** **** * Improved CERES Height CloudSat Height

CloudSat Observed NWP Tropopause Height Overshooting Tops

40 Long-Term Overshooting Top Databases and OT Relationships With Hazardous Weather

• The IRW-texture algorithm has been used to produce a 5+ year database of GOES-12 and SEVIRI OT detections over the U.S. and Europe/North Africa

- These databases illustrate where and when strong storms are most frequent across these regions

• The GOES and SEVIRI OT databases can also be used to show how detected OTs relate to hazardous weather. OTs are compared with:

1) U.S. Storm Prediction Center and European Severe Weather Database (ESWD) severe storm reports

2) Objective 1-minute in-situ turbulence observations from United Airlines Boeing 737/757 aircraft

3) Cloud-to-ground lightning strikes from the U.S. National Lightning Detection Network (NLDN)

4) Radar reflectivity observations from the U.S. NEXRAD network 41 Rapid + Operational Scans

Day + Night Color Daytime: 9 AM – 9 PM Ranges Do Not Match Nighttime; 9 PM – 9 AM

Operational Scans Only 42 SEVIRI Overshooting Top and Land Surface Elevation

43 Diurnal Variation in OT Activity Over the GOES-12 and SEVIRI Domains

• The time of peak OT activity is approximately the same between Europe and the U.S.

• OTs are more frequent over the GOES-12 domain at night due to:

1)The presence of the Caribbean and Gulf Stream Ocean current, which exhibits a weaker daytime OT bias than most land regions

2)The Great Plains region which shows a clear night-time OT bias

44 Qualitative Comparisons of OT Detections With Severe Weather Reports

• OTs are often detected where severe weather is reported, though it is clear that an OT detection does not guarantee a severe storm

• Linear patterns in the OT field show that individual severe storm cells can be identified over long time periods 45 Qualitative Comparisons of OT Detections With Severe Weather Reports

• OTs are often detected where severe weather is reported, though it is clear that an OT detection does not guarantee a severe storm

• Linear patterns in the OT field show that individual severe storm cells can be identified over long time periods 46 Quantitative Comparisons of OT Detections With Severe Weather Reports

• Statistical comparisons of OT detections with severe weather events can be challenging and may produce misrepresentative results because:

1) Not all severe weather events are reported, especially over Europe (e,g. events during the night and/or in unpopulated areas, see Dotzek et al. (Atmos. Res.,2009))

2) The reported time and location of events may not be perfectly accurate which can affect comparisons since a fixed time/distance match criteria are often used

•The OT-severe weather event comparisons were done for this talk in two ways:

1) Determine how often OTs were detected within a given time/distance from a confirmed severe event (GOES-12 & SEVIRI) - But, does a “miss” (i.e. no OT near severe event) mean that the algorithm did not detect the OT or does this mean that an OT is not required for a storm to be severe?

2) Determine how often severe weather is found within a given time/distance from a detected OT (GOES-12 only) 47 Comparison of GOES-12 OT Detections With Severe Weather Reports: 2004-2009 Warm Seasons

Severe Weather to OT Comparison

# of OTs Near Severe Weather Number of Severe Severe Weather Match Percentage Type Weather Events Events INTERPRETATION 2633 4684 56.2% OTs are detected within +/- 30 mins Severe 30804 52743 58.4% and 60 km of severe storms for

Large 28787 56114 51.3% 54.8% of severe events

All Types 62224 113541 54.8%

OT to Severe Weather Comparison Severe Weather # of OTs With # of OTs Without Match Percentage Type Severe Weather Severe Weather

Tornado 9107 426269 2.1% There is a 25.1% probability that a Severe Wind 72535 362841 16.7% storm with an OT detection will be Large Hail 61330 374046 14.1% severe

All Types 109042 326334 25.1% 48 Relationship Between GOES-12 Minimum IRW BT and Severe Weather: 2004-2008 Warm Seasons

• Severe weather risk increases with decreasing minimum OT IRW BT

• Relationship is most pronounced for severe wind storms

49 Comparison of GOES-12 OT Detections With Severe Weather Reports: 2004-2009 Cold Seasons

Severe Weather to OT Comparison

# of OTs Near INTERPRETATION Severe Weather Number of Severe Severe Weather Match Percentage Type Weather Events Events OTs are detected within +/- 30 mins and 60 km of severe storms for 31.6% Tornado 625 1436 43.5% of severe events. Warm Severe Wind 1729 6673 25.9% season=54.8%

Large Hail 2538 7353 34.5% OTs may be less pronounced in cold

All Types 4892 15462 31.6% season severe storms

OT to Severe Weather Comparison Severe Weather # of OTs With # of OTs Without Match Percentage There is a 12.1% probability that a Type Severe Weather Severe Weather storm with an OT detection will be Tornado 1478 52667 2.7% severe. Warm season=25.1%

Severe Wind 3400 50745 6.3% OT-severe weather relationship is not Large Hail 3682 50463 6.8% as well defined during cold season

All Types 6522 47632 12.1% 50 Comparison of Severe Weather Reports With SEVIRI OT Detections: 2004-2009 Warm Seasons

• Confirmed (QC1 and QC2) ESWD events are compared with SEVIRI OT detections for the 2004-09 warm seasons. Events with a temporal confidence of +/- 1 hour are used. Distance match criterion based on temporal confidence. See Bedka (Atmos. Res., 2010) for methodology

• Severe wind and large hail stats are very consistent with those over the U.S., but the tornado stats are significantly different (56% vs. 14%). Further research should 51 be done to understand this difference Relationship Between OTs and Maxima

• OTs were detected for each of the storm cells labeled A-D

• Localized radar reflectivity maxima are present near the OT region 52 • Comparison of 2008 warm season OTs with U.S. national composite radar reflectivity mosaic

• Maximum reflectivity within 9 km2 region surrounding OT is recorded

• Coldest OTs are associated with heaviest rainfall

• Combination of false alarm, radar network gaps/outage, and boundary layer rainfall evaporation in dry boundary layers contributes to the 30% of 212- 215 K OTs with reflectivity < 30 dBZ

• 5-min base reflectivity manually tracked +/- 30 mins of OT detection for 200 OT-producing storms

• Rainfall rapidly intensifies within the storm before OT and peaks at the time of the OT detection

• Theory: Strong updraft produces compensating downdraft, forcing heavy to the surface?

53 Overshooting Top Impact on Aviation Operations

Magenta Box: OT Detection Red: Severe Turbulence Green: Moderate Turbulence and Blue: Light Turbulence

54 Overshooting Top Impact on Aviation Operations

Flight Density of United Airlines Boeing 737 6-year Climatology of GOES-12 and 757 Aircraft That Collect Objective EDR Overshooting Top Detections Turbulence Measurements

Southwest and Delta Airlines have recently begun collecting EDR turbulence data, providing better 55 geographic coverage Quantitative Relationship Between OT Detections and In-situ Turbulence Observations

Comparison Between GOES-12 OTs and United Airlines Objective EDR Turbulence Observations During 2005-2008 Warm Seasons

Increased turbulence for non-OT cold pixels at larger radii likely associated with other distant OTs

56 From Bedka et al. (JAMC,2010) Quantitative Relationship Between OT Detections and Cloud-To-Ground (CG) Lightning

OT detection output compared to NLDN CG Stronger storm updrafts produce larger strikes in 2008. Comparison to “total lightning” is overshooting magnitude and increase a future work item charge separation in the storm core, yielding concentrated CG lightning in the CG strikes are much more frequent near OTs than OT region cold pixels in a spatially IRW uniform BT field 57 Summary of Objective Satellite-Based Overshooting Top Detection Research

• Development of an accurate objective satellite OT detection algorithm is challenging - Validation results show that use of spatial IRW BT gradients and BT thresholding provides the best balance between POD and FAR of currently available methods - Algorithm validation will continue to expand over time, U.S. NEXRAD “echo top” mosaic being used for upcoming validation studies

- The work of Setvak et al. illustrates that IRW-only methods may not work effectively in all situations, especially as OTs become “older” and mix with the ambient environment

• Long-term geographic OT detection distributions show realistic results and illustrate similarities and differences between convective storm intensity/frequency over the U.S. and Europe

• OT detection output provides value added information for forecasting locations of severe weather, CG lightning, aviation turbulence, and radar reflectivity maxima, especially in data void regions - The IRW-texture algorithm is being developed for future operations in the GOES-R era, but offers benefits to today’s forecasting problems with current LEO/GEO instruments

- I welcome collaborations with motivated researchers/forecasters who will commit to evaluating detection output and providing feedback. 58 Thank You For The Interest and Your Time!

Any Questions??? (Enhanced-V Detection Slides To Follow If Time Available)

59 Objective Enhanced-V Signature Detection

60 What is an Enhanced-V?

The appearance of an enhanced-V in infrared imagery resembles a V- or boomerang-shaped area of cold IR window channel brightness temperatures (BT), and is bordered by an area of warm BT downwind. (From Meteorology: Understanding the Atmosphere (Ackerman and Knox, 2001) )

• The enhanced-V is often seen in infrared satellite imagery before the onset of severe weather (damaging winds, hail, and/or tornadoes) and is an important indicator of a severe thunderstorm.

- McCann (1983) found that the median lead time (time from the initial identification of the feature on satellite imagery to the first reports of severe weather on the ground) is about 30 minutes.

- Brunner et al. (2007) found that ~92% of enhanced-V storms with a specified set of temperature parameters produced severe weather during the 2003 and 2004 warm seasons

61 Components of the Enhanced-V and Relationship with Severe Weather

AVHRR image courtesy of Martin Setvak (CHMI) 62 V-Signature vs. Anvil Thermal Couplet Detection

• Though the enhanced-V signature can appear quite differently in many cases, the anvil thermal couplet (ATC) is (almost) always present, making accurate detection of the ATC and thus the V-signature far more attainable

MODIS & AVHRR IR Window Degraded to 2 km ABI Resolution All Images Cover the Same Horizontal Distance and Use Same Color Enhancement, Illustrating Variations in the V-signature Across Events

63 63 Enhanced-V Anvil Thermal Couplet Detection Processing Schematic

INPUT: Overshooting Top Detection Mask

Search region in the eastern semicircle of the OT for potential couplets

Ensure that potential couplet is not along the anvil cloud edge

Ensure that potential couplet is a focused area of warm BT surrounded by colder anvil

Ensure presence of a relatively large anvil cloud with a singular warm area

Output Enhanced-V Thermal Couplet 64 Mask 64 Enhanced-V Detection Algorithm Summary

• Input Datasets: Proxy ABI IR window data are generated from 1) 1 km AVHRR/MODIS, 2) 3 km MSG SEVIRI, and 3) 2 km WRF NWP output from UW- CIMSS. An NWP tropopause temperature analysis is used to ensure that a pixel is indeed “overshooting” the tropopause

• Overshooting top pixels are first identified

• Overshooting top detections are then passed to the enhanced-V anvil thermal couplet detection algorithm

• Thermal couplets are located within the “arms” of the enhanced-V signature, downstream (i.e. east) of the overshooting top location

• A series of tests are performed to ensure that the thermal couplet is a focused area of warm temperatures within a relatively large continuous anvil cloud, consistent with observations of enhanced-V producing storms

• Several additional parameters are output that can be used as quality control

65 65 66 67 Overshooting Top and Thermal Couplet Detection Product Output

2 km Proxy ABI From 1 km MODIS 10.7 μm Imagery 4/7/2006 at 1825 UTC

* Overshooting top location indicated by blue symbol (only OTs with thermal couplets are shown) * Enhanced-V ATC detections indicated by green symbol 68 * 4 of 5 possible enhanced-V ATCs are detected accurately68 in this MODIS case Overshooting Top and Thermal Couplet Product Output

5/10/2004 2317 UTC 5/10/2004 at 2317 BlueUTC Box: 2 km OT Pr Detectionsoxy ABI From 1 km AVHRR Green Box: Enhanced-V ATC Detections White Lines: Enhanced-V Signatures Severe Storm Reports +/- 30 Mins 1 2 From Image Time Numbers Correspond To The Storms Labeled In the Left Panels

1 4 3 3 4 2

69 69 View Slide Show to See Figures Underneath Overshooting Top and Thermal Couplet Product Output

5/10/2004 2317 UTC Blue Box: OT Detections Green Box: Enhanced-V ATC Detections White Lines: Enhanced-V Signatures Severe Storm Reports +/- 30 Mins 1 2 From Image Time Numbers Correspond To The Storms Labeled In the Left Panels

1 4 3 3 4 2

70 70 Performance Estimates Enhanced “V”/Anvil Thermal Couplet Detection

* No objective “truth” dataset currently exists for defining the presence of enhanced-V’s

* Brunner et al. (Wea. Forecasting, 2007) have manually identified 450 enhanced “V” cases within MODIS and AVHRR data during Summer 2003 and 2004. Additional V cases from recent years have also been identified.

* 1 km MODIS and AVHRR longwave IR (11 micron) imagery have been remapped to ABI navigation and spatial resolution to simulate future ABI imagery

* The enhanced “V” anvil thermal couplet detection algorithm is run for about half of these manually identified V cases and POD/FAR statistics are computed over the entire MODIS/AVHRR overpass

71 Anvil Thermal Couplet Detection Performance Relative to Brunner et al. V-Signature Database

False Alarm Ratio Probability Of Detection False Alarm Ratio Accuracy Requirement

62% 24% (126 accurately detected (40 false detected / (40 false 25% / 203 V-signature events) detected + 126 accurately detected))

• 76% of all detected enhanced V anvil thermal couplets were associated with severe weather.

• 66% of non-detected enhanced V anvil thermal couplets were associated with severe weather, indicating that the ABI algorithm focuses on high impact events

72 72 Enhanced-V Anvil Thermal Couplet Detection Validation Stats: GOES-12 vs. ABI

Input Imagery % of Accurately Detected to Enhanced-V False Probability Enhanced-V ATCs with ATC Detection Alarm of Severe Weather within +/- Ratio Detection 30 minutes and 60 km of 125 Events the OT location

Proxy ABI from 21.5% 67.2% 77.4% MODIS/AVHRR

GOES-12 5.9% 12.8% 100.0%

• Compare enhanced-V detection output for time-matched GOES and AVHRR/MODIS imagery. Only 125 of the 203 events are included here

• Our ability to detect enhanced-Vs in current GOES-12/13 imagery is 73 diminished due to coarser image spatial resolution Enhanced-V Detection Summary

• Algorithm focuses on detection of anvil thermal couplets (ATC), not the enhanced-V signature itself

• ATC detection method is based on OT detection output

- No OT detection = No ATC detection

- ATCs can be in the eastern semicircle of an OT. This is not a significant issue over the U.S., but the code may need some adjustment to operate effectively over Europe

- The algorithm will detect cold ring shaped storms in addition to U/V storms, provided there is a significant cold/warm BT couplet between the OT and ring center

• Algorithm validation based on database of 450 known AVHRR/MODIS-observed enhanced-V events (Brunner et al. 2007)

- Algorithm meets 25% FAR GOES-R performance specification

- Severe weather occurrence higher for detected ATCs vs. those that are undetected 74 Thank You For The Interest and Your Time!

Any Questions???

75 Backup Slides

76 Anvil Height Anvil 11 um BT OT Peak Height OT Min 11 um BT

MODIS+CloudSat OT Characteristics

OT-Anvil Height Difference Lapse Rate= ------OT-Anvil BT Diff

77