Operational Satellite Derived Tropical Rainfall Potential (TRaP)

Produced by the COMET^® Program

• 1. Introduction
  • 1.1 Module Introduction
  • 1.2 Module Outline
  • 1.3 Module Objectives
  • 1.4 Rainfall Forecasting for Landfalling Tropical Cyclones
  • 1.5 Polar-orbiting Satellite Systems

• 2. TRaP Technique, Heritage, and History
  • 2.1 Introduction
  • 2.2 Leveraging Technology Improvements
  • 2.3 Infusion of Microwave Observations
  • 2.4 The TRaP Formula
  • 2.5 Manual TRaP Technique
  • 2.6 TRaP Technique Improvements 1
  • 2.7 TRaP Technique Improvements 2
  • 2.8 Three Operational Microwave Polar-orbiting Satellite Systems
  • 2.9 Hurricane Michelle Example
  • 2.10 TRaP Assumptions and Limitations

• 3. TRaP Products and Availability
  • 3.1 Introduction
  • 3.2 Access to TRaP Products
  • 3.3 TRaP Product Examples from Hurricane Lily

• 4. Examples from 2003
  • 4.1 Tropical Storm Bill 1 (July 2003)
  • 4.2 Tropical Storm Bill 2
  • 4.3 Tropical Storm Bill 3
  • 4.4 Tropical Storm Bill Summary
  • 4.5 Hurricane Claudette 1 (July 2003)
  • 4.6 Hurricane Claudette 2
  • 4.7 Hurricane Claudette 3
  • 4.8 Hurricane Claudette Summary

• 5. Examples from 2002 - Part 1
  • 5.1 Hurricane Bertha 1 (August 2002)
  • 5.2 Hurricane Bertha 2
  • 5.3 Hurricane Bertha 3
  • 5.4 Hurricane Bertha Summary
  • 5.5 Tropical Storm Fay 1 (September 2002)
  • 5.6 Tropical Storm Fay 2
  • 5.7 Tropical Storm Fay 3
  • 5.8 Tropical Storm Fay Summary
  • 5.9 Tropical Storm Hanna 1 (September 2002)
  • 5.10 Tropical Storm Hanna 2
  • 5.11 Tropical Storm Hanna Summary

• 6. Examples from 2002 - Part 2
  • 6.1 Tropical Storm Edouard 1 (September 2002)
  • 6.2 Tropical Storm Edouard 2
  • 6.3 Tropical Storm Edouard 3
  • 6.4 Tropical Storm Edouard Summary
  • 6.5 Hurricane Isidore 1 (September 2002)
  • 6.6 Hurricane Isidore 2

• 7. Additional Examples
  • 7.1 Hurricane Lili 1 (October 2002)
  • 7.2 Hurricane Lili 2
  • 7.3 Tropical Storm Helene 1 (September 2000)
  • 7.4 Tropical Storm Helene 2
  • 7.5 Tropical Storm Helene 3
  • 7.6 International Support - Caribbean Examples from 2002
  • 7.7 Summary

• 8. Future Initiatives and Improvements
  • 8.1 Introduction
  • 8.2 An Example of Digital Output to Users
  • 8.3 Tropical Rainfall Nowcaster (TraN)

• 9. Summary
  • 9.1 TRaP Summary
  • 9.2 Contacts, Resources, and Internet Links
  • 9.3 References

1.0 Introduction

Welcome to the Operational Tropical Rainfall Potential (TRaP) module, developed by Sheldon Kusselson at the NESDIS Synoptic Analysis Branch (SAB). Many different organizations and individuals support the design and development of the TRaP product, so this really is a collaborative effort. Various versions of the tropical rainfall potential product have been available for the last 15 years. This module will cover the creation of the original version of TRaP and its evolution into the product available today.

The current generation of NOAA and DMSP microwave sensors provides quantitative information on global precipitation to enhance forecasts of potentially hazardous rainfall through tools like the TRaP technique Both NOAA and DMSP have contributed to the design of the next generation of microwave sensors that will fly on the NPP and NPOESS missions over the next two decades. The NPOESS satellite system will have vastly improved data latency, with 95 % of the data available to users within 28 minutes or less. CMIS and ATMS will provide all of the data of the previous generation of microwave instruments with improved resolution and greater multispectral capability. CMIS will also provide passively sensed wind speed and direction, a capability unavailable with previous generation of passive microwave radiometers.

1.1 Module Introduction

Flooding from hurricanes and tropical storms has resulted in considerable loss of property and more deaths in the United States than winds and storm surges combined. Dr. Edward Rappaport of the Tropical Prediction Center (TPC) found that in the United States, during the period between 1970 and 1999, fresh water floods accounted for more than half of the 600 deaths directly associated with tropical cyclones.

1.2 Module Outline

1. Module Introduction
2. TRaP Technique, Heritage, and History
3. TRaP Products and Availability
4. Examples from 2003
5. Examples from 2002 - Part 1
6. Examples from 2002 - Part 2
7. Additional Examples
8. Future Initiatives and Improvements
9. Summary

Here is the outline for the module. After a brief introduction and look at TRaP heritage and history, the module focuses on more recent microwave TRaP products that provide guidance for potential tropical cyclone rain accumulation. Also included is information on TRaP availability on the Internet and current users. The TRaP Examples sections compare TRaP forecasts with observations and model forecasts for several storms from recent hurricane seasons. The examples provide a context for exploration of the strengths and limitations of the TRaP product. Next, future TRaP innovations are discussed. The module concludes with a short summary.

1.3 Module Objectives

• State the basis of the TRaP technique, its formulation, and inputs
• State the assumptions and the limitations of the technique
• Find and access TRaP products on the Internet

What does NPOESS bring to TRaP?

NPOESS will offer:

• Improved spatial coverage
• Improved sensors (polarization, spatial resolution)
• Reduced data latency

This module demonstrates how microwave rain rates can provide guidance on potential tropical cyclone rainfall accumulations through application of the aerial TRaP product.

After completing the module, users will be able to:

• State the basis of the TRaP technique, its formulation, and inputs
• State the assumptions and the limitations of the technique
• Find and access TRaP products on the Internet

With this knowledge, forecasters can modify what they see in the product, so that they can improve the TRaP product for a particular area of interest.

What does NPOESS bring to TRaP?

NPOESS will offer:

• Improved spatial coverage
• Improved sensors (polarization, spatial resolution)
• Reduced data latency

1.4 Rainfall Forecasting for Landfalling Tropical Cyclone

• Forecasting rainfall from landfalling tropical cyclones is a difficult task
• Few rainfall observations are available
with offshore storms

• Initializing numerical weather prediction models with enough detailed storm structure for accurate rainfall forecasts is difficult
• Radar observations available only when storm in close proximity to coast (within ~ 460 km/248 nm)
• Large gaps in radar coverage

Deriving total rainfall estimates for land-falling hurricanes is an inherently difficult task. Looking at observational data is often not enough because few rainfall observations are available offshore with a storm approaching land. Computer model output alone is also not enough because factors such as data assimilation and initialization can be problematic. Looking at radar data is insufficient because shore-based Doppler radar has a limited 250-mile range, and there are also large gaps in the radar’s coverage of any one storm. So, where else can forecasters look to find out what tropical cyclone rainfall is like while it is still offshore?

1.5 Polar-orbiting Satellite Systems

There are polar-orbiting satellite-based microwave radiometers that can measure instantaneous rain rates through clouds. The radiometers on the current generation of geostationary satellites are limited to visible and infrared detection, so they see cloud tops and don't really sense anything below cloud top height in most cases. Microwave radiometers, however, can penetrate through the entire cloud area of tropical cyclones and provide supplemental observational data to the forecaster. Information on cloud water, cloud ice, atmospheric moisture, and temperature are examples of this capability.

2.0 TRaP Technique,
Heritage, and History

2.1 Introduction

The heritage of TRaP actually goes back to the 1980s when it was a GOES (geostationary satellite) technique developed by Spayd and Scofield. The technique was subjective because, in essence, cloud signatures were assigned subjective rain rates. The TRaP formula then used those subjective rain rates in a long and laborious process to derive a single tropical rainfall potential number, the end product. But, these rain-rate values depended on cloud-top trends; hence, the large range in rain rates for particular areas of the tropical cyclone.

2.2 Leveraging Technology Improvements

NOAA Administrator:

• Emphasized the need to take advantages of new technology
• Described benefits of better observation systems/data collection

Panel Members on hurricane forecasting:

• Forecasters will depend on new systems and improve use of technology
• Observation systems have improved both space- and ocean-based data availability, timeliness, and quality

While looking at how to improve guidance for forecasters in the 1990s, the NOAA administrator felt that some of the new technologies coming on line were not being used to their full potential. He felt that new technologies need to employ new systems that use space- and ocean-based remote sensing products such as microwave observations.

2.3 Infusion of Microwave Observations

New satellite instruments and microwave sensing technology led to a transition from the subjective GOES technique for tropical rainfall potential to a more objective polar-orbiting satellite technique in the 1990s. This transition allowed big improvements in deriving rain rate estimates.

2.4 The TRaP Formula

TRaP = (R_avg * D)/(V - 1)

The current TRaP formula is a simplified form of the rainfall potential formula used with the GOES technique and relies on microwave rain rates from polar-orbiting satellites as one of the inputs.

The microwave TRaP technique is used today to produce the tropical rainfall potential. Beginning in 1992, operational meteorologists from NOAA/NESDIS’ Satellite Analysis Branch (SAB) started looking at the rain-rate product produced by the polar-orbiting, Defense Meteorological Satellite Program’s (DMSP) Special Sensor Microwave Imager (SSM/I). The goal was to produce a rainfall potential for tropical disturbances expected to make landfall in the next 24 to 36 hours.

The example shown here is of the DMSP SSM/I microwave rain rate product for Hurricane Georges on the morning of 27 September 1998, when the storm was offshore and 12 to 24 hours from landfall.

2.5 Manual TRaP Technique

The microwave TRaP technique was once done manually by satellite analysts. It was time consuming to draw lines across the digital rain rate values and only produce one or two tropical rainfall potential numbers.

2.6 TRaP Technique Improvements 1

In 2000, more of the TRaP algorithm was automated and calcualted by the computer. It integrated all of the storm’s microwave rain rates to produce a graphical product based on the current movement of the storm, compared with just one rainfall potential number.

2.7 TRaP Technique Improvements 2

In 2001, major improvements included going to a higher resolution rain-rate product from the NOAA-15 Advanced Microwave Sounding Unit (AMSU) and using the Automated Tropical Cyclone Forecast System (ATCF) for the future movement and speed of the storm. Adding the NOAA-16 polar orbiter doubled the coverage for rain rates over a particular storm, providing the ability to do TRaPs every 6-12 hours. Another important addition was the use of the rain-rate product from the NASA Tropical Rainfall Monitoring Mission (TRMM) Microwave Imager (TMI). The first TRMM mission was launched in 1997 and continues to provide data into 2004.

2.8 Three Operational Microwave Polar-orbiting Satellite Systems

Today, tropical rainfall potential products are produced from the three satellite systems with microwave sensors. This example is from Hurricane Michelle in November 2001. Each satellite has different characteristics, and SAB makes use of the microwave rain rate data from all three. TRMM, unlike NOAA AMSU and DMSP SSM/I, is not a polar-orbiting satellite. It covers the tropics between 38 degrees north and 38 degrees south latitude.

When any of these satellites passes over a storm, a tropical rainfall potential can be generated automatically. As one can see, the TRMM TMI has a higher-resolution five-kilometer footprint compared to NOAA AMSU and DMSP SSM/I. However, one can get many more passes with the NOAA and DMSP satellites than with the TRMM, which generally sees only one pass over a given location in a 24-hour period. Each system has its own characteristics and all available satellite data are used to produce the TRaP product. This is also a good way to get people looking at tropical cyclone rainfall using platforms other than geostationary satellites.

2.9 Hurricane Michelle Example

For Hurricane Michelle between 0643 and 1210 UTC on 3 November 2001, three different tropical rainfall potentials are produced while the storm is between the Yucatan Peninsula and Cuba. There are slightly different results for each satellite platform, because each of the satellite’s sensor and rain rate algorithms is different. In this case, the AMSU TRaP is higher than the TRMM.

2.10 TRaP Assumptions and Limitations

Assumptions:

• Satellite-derived rain rates are correct
• No change in the size, shape, and magnitude of the area of the rain rates
• Rain area moves in the forecast direction and speed of the storm

Limitations:

• No outside influences on the storm are permitted
• Presence of fronts/upslope conditions indicates need to increase TRaP value
• Presence of shear conditions indicates need to decrease TRaP values
• Presence of dry air entrainment indicates need to decrease TRaP values

There is no one technique or model forecast that is perfect, and TRaP won't give you a perfect forecast. However, it does provide guidance information for the forecaster to use to improve his or her rainfall forecast. So forecasters should note that there are essential assumptions and limitations to this product that they must know and understand.

Some of these assumptions are:

• The satellite-derived rain rates are correct. They are valid for the particular satellite’s snapshot pass over the storm, and the rain rates are only as good as the algorithms that go into producing the TRaP product.
• There is no change in the size, shape, or magnitude of the area of the rain rates. So, once you have that snapshot, the algorithm just assumes that those rain rates don't change at all for the next 24 hours.
• The rain area moves in the forecast direction and speed of the storm.

Some of the TRaP limitations are:

The TRaP technique cannot incorporate outside influences on the storm. In practice, however, there are many possible outside influences on a storm, so the following guidance should be used.

If there is a front and/or upslope conditions, a forecaster should know to increase the TRaP values. On the other hand, if dry air or shear are getting involved in the storm or will get involved over the next 24 hours as the storm approaches or makes landfall, then a forecaster should know to decrease the TRaP values. Examples where the TRaP algorithm does or does not do well will be presented later in the module. Most of the case studies indicate that the TRaP technique is accurate as long as these assumptions are reasonably well satisfied and their limits are taken into consideration by forecasters.

3.0 TRaP Products and Availability

3.1 Introduction

• TRaP Products on the Web
• TRaP Users
• Sample TRaP Products

Before looking at some TRaP examples, let's quickly look at where you can view the TRaP product. Right now, today, you will see the tropical rainfall potential product for named or numbered storms SAB has been covering in the last few weeks. This product is automatically produced by a workstation in Camp Springs, Maryland and is usually posted to the Website when the storm has wind speeds of 35 knots or greater and is within 24 to 36 hours of landfall. Tropical rainfall potentials are not posted to the homepage until all of the previous assumptions and criteria are met and they have been evaluated by an experienced analyst.

3.2 Access to TRaP Products

In the United States, most TRaP users are from the National Center for Environmental Prediction (NCEP), the Hydrological Prediction Center (HPC), and the Tropical Prediction Center (TPC). The National Weather Service’s Central Pacific Hurricane Center (CPHC) also uses the TRaP product when there are storms approaching the Hawaiian or other Pacific Islands. They are particularly interested in what kind of rainfall to expect as the storm approaches. Other places using the product are the National Weather Service Forecast Offices and National Weather Service River Forecast Centers, where they can at least view the product on the TRaP homepage, but cannot yet physically display digital data on a workstation at their offices. This functionality will be added in future iterations of the TRaP product.

SAB produces TRaP products for all of the Eastern Hemisphere whenever the storms are within 24 hours of landfall. Many countries have expressed a real desire to have this product, not just on the homepage, but in a digital format so that they can compare it with model guidance and observational guidance. Australia, Canada, India, Belize, Vietnam, and others are all viewing the TRaP product on the Internet.

TRaP Product Access:

Posting

•  TRaP product posted to homepage
   http://www.ssd.noaa.gov/PS/TROP/trap-img.html
  
•  Customer e-mail notification
  
•  FTP access to digital products

Generated Message:

Subject: New TRaP Created
Date: Fri, 5 Sep 2003 17:55:33 GMT

Dear TRaP Customer,

A new TRaP has been generated in your area of interest. This TRaP was generated for FABIAN using 1438Z AMSU rain rate data and the 1444Z KNHC bulletin, and will be avaliable for 48 hours.

To access it please go to
ftp://gp16.wwb.noaa.gov/pub/TRAP and find:

2003FABIAN.WTNT45.KNHC.051444.AMSU.051438.24.Z
2003FABIAN.WTNT45.KNHC.051444.AMSU.051438.24.txt.Z
2003FABIAN.WTNT45.KNHC.051444.AMSU.051438.24.GIF

in directory "Atlantic"

For information on how to decipher the text file please see the document located in directory "README". If you have any questions or wish to have your name removed from our notification system, please contact Michael Turk.

TRaP Product Access:

E-mail Notification

Sheldon Kusselson or Michael Turk.

• Overview of the TRaP product
• Link to TRaP examples
• Verification information
• Product formats

3.3 TRaP Product Examples from Hurricane Lily

Here is an example of the various TRaP products that are now automatically being produced in Camp Springs, Maryland. For each TRaP product that is produced, four individual six-hour TRaPs out to 24 hours and one 24-hour composite can be produced. Remember that the automatic TRaPs are generated any time a storm is given a name or a number. This is one example from Hurricane Lily. Six-hour TRaPs like the set shown here do not currently animate on the homepage, but will sometime in the near future. Also planned for the Website are separate pages that provide TRaP output organized by region. Currently, one 24-hour TRaP graphic is available on the homepage.

4.0 Examples from 2003

4.1 Tropical Storm Bill 1 (July 2003)

Tropical Storm Bill was a system that threatened the U.S. Gulf Coast in July, 2003. Here, the NOAA-16 AMSU provided a pass about 24 hours before the storm made landfall.

4.2 Tropical Storm Bill 2

This slide compares the TRaP output with observed rainfall for tropical storm Bill. The observed rainfall product, also known as the Stage IV Quantitative Precipitation Estimate, is generated by NCEP and represents a quality-controlled 24-hour composite of rainfall estimates derived from multi-sensor observations including radar, satellite, and rain gauge.

The TRaP product did a good job of predicting the total rainfall amounts, but the location of the TRaP rainfall is different from the observed rainfall.

4.3 Tropical Storm Bill

This slide compares 24-hour forecast rainfall totals from both Eta and GFS models with the TRaP and observed totals. Both of the model runs underestimated the total rainfall by a factor of two.

Looking at the rainfall observations, the maximum amount was on the southern Mississippi/Alabama coast. The TRaP did rather well because the storm behaved well, and it was a pretty good forecast based on that snapshot image. This case nicely illustrates where satellite data provided good guidance to the forecaster.

4.4 Tropical Storm Bill Summary

• Little change in magnitude and area of rain rates
• Actual track of storm was faster and to the east
• Two parts of the storm merged to produce the final rainfall amounts

In analyzing the TRaP product in light of the observed rainfall, it seems that the product did well on the magnitude and area of the rainfall, but miscalculated the landfall location. From other analyses, it was found that the actual track of the storm was faster and farther east than the HPC track forecast. The western half of the storm with a ten-inch maximum precipitation area merged with the other part of the storm and created a final rainfall total of nine inches.

4.5 Hurricane Claudette 1 (July 2003)

In 2003, Hurricane Claudette was bearing down on the Texas coast when the NASA TRMM satellite provided instantaneous rain rates from its recent pass. As can be seen in the TRaP image, three areas of heavy rain were forecast by the TRaP product.

4.6 Hurricane Claudette 2

This slide compares the observed 24-hour rainfall on the left and the 24-hour TRaP product on the right. In this case, the TRaP product produced rainfall amounts consistent with the Stage IV-observed rainfall product.

4.7 Hurricane Claudette 3

The GFS model did significantly better than the Eta model on this storm, however, it still underestimated the total amounts by more than an inch.

4.8 Hurricane Claudette Summary

• Little change in both area and magnitude of rain rates
• Good 24-hour forecast of storm's movement
• Rain areas moved in forecast direction and speed of storm

The TRaP product for hurricane Claudette performed quite well on the magnitude and overall coverage of rainfall.

Claudette took on a more westerly track and accelerated somewhat after making landfall, but maintained its speed and direction as it moved across Texas and towards northern Mexico.

Notice the inland spreading of the heavier 7- to 8-inch totals seen in the 24-hour TRaP product on the left. This is the result of extrapolation of more intense convective elements that the TRMM rainfall product was able to resolve just prior to the storm making landfall. Recall that the resolution of the TRMM TMI is three times greater when compared with the AMSU or SSM/I. AMSU and SSM/I would have considerably smoothed and underestimated these smaller convective areas, so that the net effect on the 24-hour TRaP product would also have been a reduction in the overall rainfall totals.

5.0 Examples from
2002 - Part 1

5.1 Hurricane Bertha 1 (August 2002)

2002 was a busy season for the production of TRaPs, especially over the Gulf of Mexico. Hurricane Bertha is one example. Whenever there is a complete pass over a storm from a NOAA, TRMM, or DMSP satellite, a rain-rate image is extracted. In this case, a NOAA AMSU pass was used as the storm was approaching southern Alabama and southeast Mississippi. At the same time, the latest 24-hour tropical storm track forecast is obtained from the TPC. To create the storm track, the best human input is integrated with numerical model guidance, to forecast the direction and speed of the storm within the 24-hour rainfall potential outlook period. This track forecast is then used to create the TRaP product. In the best storm forecasts, there is a lot of collaboration with the HPC, the National Weather Service Office in Miami, and lots of other people before TPC puts out that best track forecast. The computer generates a 24-hour rainfall potential from that snapshot rain-rate image and the latest hurricane track forecast coming out of TPC.

5.2 Hurricane Bertha 2

From the rainfall observations, you can see that the maximum amount was on the southern Mississippi/Alabama coast. The TRaP did rather well because the storm behaved well, and it was a pretty good forecast based on that snapshot image. SAB looks at satellite data because satellite input may provide good guidance to the forecaster.

5.3 Hurricane Bertha 3

The numerical models can sometimes underestimate forecasts of tropical storm rainfall. This is one of the reasons why SAB believes that using a product like TRaP will provide useful guidance to forecasters and also provide useful input to river forecast models. In the lower right panel, we can see that the GFS precipitation forecast actually verified quite well compared to the Eta, which had a little more trouble, particularly for some of the heavier, more localized amounts.

5.4 Hurricane Bertha Summary

The TRaP product worked well in this case because there was little change in the magnitude and area of rain rates as seen in that snapshot image. The whole rain area of the storm moved in the direction and speed predicted, while outside influences were minimal. These factors combined to give the forecaster confidence in the TRaP output for the storm as it headed toward southern Alabama and southeast Mississippi.

5.5 Tropical Storm Fay 1 (September 2002)

Tropical Storm Fay was a bit different from Bertha. The TRaP product started out with an SSM/I pass of the storm. A large area of rain, with very high rain rates of more than 1¼ inch per hour, was observed. The SSM/I, however, has a 15-kilometer resolution. So a few more localized intense rain-rate areas probably existed within that 15-kilometer box. Taking the storm track forecast from TPC, a tropical rainfall potential product was derived and displayed on the SAB TRaP homepage.

5.6 Tropical Storm Fay 2

This slide shows a comparison between the TRaP product and the observed rainfall for Tropical Storm Fay. Since the TRaP product is somewhat of an extrapolation from the instantaneous rain rate, it often shows larger areas of rainfall potential with larger maximum amounts.

The maximum amounts worked pretty well, but the aerial extent smaller, which implies that the storm probably weakened as it moved inland, with the intensity near the center staying about the same. This kept the maximum amount fairly accurate, but that maximum amount was focused closer to the center of the storm as it moved inland.

5.7 Tropical Storm Fay 3

How did the model guidance do? If you had looked at model guidance from the AVN and the Eta-32 runs that evening, you probably would have tossed it into the trash can. The models didn't really give you very much in the way of guidance. There is a lot of satellite data that does go into the models, but there may have been a problem with data assimilation. The TRaP model isn't in competition with the models. It is another guidance product for the forecaster to look at, especially if there is little faith in the model forecast for a storm as it moves inland.

5.8 Tropical Storm Fay Summary

To summarize, for Tropical Storm Fay there were some changes in both the area and magnitude of the rain rates. That is probably why the area of rainfall potential was smaller than the TRaP product showed. The rain area did move in the forecast direction and lined up well with the observed track. The storm experienced a minimal amount of outside influence, such as the movement into a very moist environment with no shear or dry air intrusion, so the TRaP forecast remained relatively accurate.

5.9 Tropical Storm Hanna 1 (September 2002)

For Tropical Storm Hanna, the rain-rate image shown here came from the NASA TRMM satellite. The TRaP product shows the 24-hour rainfall potential based on the instantaneous rain rate and the storm track from TPC. How did the TRaP product perform for this case?

5.10 Tropical Storm Hanna 2

Again, for reasons stated earlier and for other reasons that are unknown at this time, the models tended to have a hard time generating sufficiently heavy precipitation with these storms. The TRaP product can supplement model forecasts, even though this TRaP underestimated the total rainfall amounts somewhat. There could have been more outside influences as the storm penetrated further inland. Another point to keep in mind is that the performance of an extrapolation technique like TRaP tends to degrade toward the end of the 24-hour period. And that degradation is even more evident in cases where outside influences on the storm are more pronounced.

5.11 Tropical Storm Hanna Summary

Hanna was an example where both the model and TRaP rainfall amounts were lower than the observed totals. The TRaP forecast total, however, was much closer to the final amount. The storm did follow the forecast track, but there was some change in the area and magnitude of the heaviest rainfall totals.

6.0 Examples from 2002 -
Part 2

6.1 Tropical Storm Edouard 1
(September 2002)

Just to illustrate that not all TRaPs perform that well, here are some examples in which TRaPs did not do very well. Tropical Storm Edouard is a classic. This storm had intense rain rates, a well-defined storm structure, and was generally forecast well by the TPC. Indications were of a slow moving storm with very high rain rates. The TRaP output produced 16 inches of rain. However, did 16 inches of rain fall on the Florida coast?

6.2 Tropical Storm Edouard 2

The 24-hour observed rainfall totals shown here indicate that Edouard only produced a 4.5 inch maximum over portions of central Florida.

Why didn’t the storm produce 16 inches of rain as the indicated by the TRaP product?

6.3 Tropical Storm Edouard 3

There are three reasons why the TRaP product underestimated the rainfall with Edouard. First, the system was sheared. Second, there was dry air entrained into the storm. Third, just as it was making landfall, the storm suddenly sped up from that snapshot image at 2056 UTC on 4 September. The somewhat slower TPC track forecast that provided input for the TRaP resulted in an overestimate of rainfall amounts. Those were the three big factors responsible for decreasing the rain rate during that 24-hour period of the TRaP product. So it is not surprising that what eventually fell was one-fourth of what the TRaP product forecast.

6.4 Tropical Storm Edouard Summary

To summarize for Tropical Storm Edouard, there was significant shear impacting the system that weakened the storm in this sequence of images. The storm pretty much fell apart as it crossed into central Florida. Three factors worked together to lessen the actual rainfall totals: environmental shear, dry air entrainment, and increased storm motion.

6.5 Hurricane Isidore 1 (September 2002)

Hurricane Isidore is another interesting example. This storm had a rebirth of sorts after leaving Mexico and falling apart in the central Gulf of Mexico. As a result, the TRaP product was highly disorganized. Twelve hours later over the Central Gulf, the storm became better organized, as did the TRaP product. By the time Isidore approached the Central Gulf Coast of the U.S. mainland on the morning of 26 September 2002, the storm had become much better organized. What do you think the TRaP product showed?

6.6 Hurricane Isidore 2

The TRaP product showed a little bit better organization. Usually the TRaP product is not produced when the storm is very close to the coast, mainly because microwave rain rates are more accurate over water than over land. Also, forecasters have their radar to look at once most of the storm rainfall is over land. In this case, there were some pretty good rain rates close to the coast. The storm’s center was just offshore and the TRaP product actually worked out fairly well. It is known that a weakening storm necessitates a decrease in rain amounts. That would normally be the case. But here, Isidore encountered a frontal system that helped increase the TRaP amount and offset the effects of the storm weakening over land.

7.0 Additional Examples

7.1 Hurricane Lili 1 (October 2002)

This slide shows some TRaPs from Hurricane Lili and the associated verification data. A polar-orbiting pass has to show better than 75% of the rainfall in the storm to be used as part of a TRaP product. In this case, in the period before landfall, there were at least three good passes over the storm to produce 24 hour TRaP guidance products. Focus on the observed rainfall images in the lower portion of the slide. As often happens, especially when a storm is at least 24 hours from landfall, a lot of the convection starts developing after the snapshot image. The result is that this newer convection does not get incorporated into the TRaP product. So, in many cases, a lot more rainfall is generated well out ahead of the TRaP product guidance, especially when the storm is 18 to 24 hours from landfall. As the storm gets closer to landfall, rainfall estimates seem to get a little bit better, as was seen with this example.

7.2 Hurricane Lili 2

These images compare computer guidance with the TRaP output (upper left), and the Stage 4 rainfall totals (upper right). Notice that both AVN and Eta models underestimated the forecast rainfall amounts.

By exploiting the various satellites available to operations, the TRaP product for this case clearly demonstrates the ability of satellite-based guidance to help improve forecasts for critical weather situations like the tropical cyclone expected to make landfall.

7.3 Tropical Storm Helene 1 (September 2000)

Tropical Storm Helene was an operational case from September 2000. TRaP guidance will be compared with model data and observations. This slide shows the TRaP forecast for Helene. TRaP output was generating a maximum amount of 11 inches over the Gulf of Mexico, whereas none of the models was forecasting more than about four inches. This was guidance information provided to the NCEP HPC Branch, which is co-located with the SAB within the National Precipitation Prediction Unit.

7.4 Tropical Storm Helene 2

HPC forecasters used the TRaP rainfall guidance in their discussion material as the storm approached landfall. None of the computer model forecasts were indicating amounts comparable to the TRaP product as the storm approached the Florida coast.

7.5 Tropical Storm Helene 3

This slide compares the 24-hour TRaP output with observed rainfall totals for the same period. Even if the TRaP guidance was too slow, it would have given the forecaster an indication that rain amounts in excess of those predicted by the models, were possible along the coast and inland. This case illustrates that the TRaP product may give a heads up to the forecaster of the potential for heavy rainfall as a tropical cyclone nears the coast and moves inland.

As mentioned earlier, the convection ahead of the tropical cyclone is often completely missed by that snapshot polar-orbiting satellite image. This means that when the TRaP is generated 24 hours from landfall, it will often appear to be a little slow in bringing that heavier amount to the coast. TRaP will otherwise do a pretty good job of giving the forecaster an idea of how much rain to expect as the storm itself comes inland.

7.6 International Support - Caribbean Examples from 2002

As mentioned previously, TRaPs are not produced for the U.S alone. SAB is building international support. Here are some examples from the Caribbean: In 2001, Iris went into Guatemala. In the same year, Michelle went across Cuba and strayed just south of Florida. SAB is trying to build support for TraP not just in the United States and the Western Hemisphere, but also globally.

7.7 Summary

There is a lot more work to be done. By no means has TRaP development reached an end. Currently TRaP is not fully automated. An operational satellite meteorologist will look at TRaP output when a storm is within 24 to 36 hours of landfall to make sure that all of the rain in the storm is covered, and that the TRaP product looks reasonable before it is made available on SAB’s homepage. A number of initiatives and improvements have been identified for TRaP and are reviewed on the following pages.

8.0 Future Initiatives and Improvements

8.1 Introduction

• SAB partnerships with NWS River Forecast Centers
• Validation efforts
• Six-Hourly TRaPs organized by worldwide storm basin
• Continued evaluation of the Hydro-Estimator
• Development of a Tropical Rainfall Nowcaster (TRaN) for storms over land

Let's take a moment to look at some key trap initiatives and improvements. First SAB would like to continue building on its partnerships with NWS River Forecast Centers.

Second, SAB plans to continue its validation efforts, comparing TRaP outlooks with combined Eta model output and rainfall observations.

The third item addresses output on the Web. SAB would like to begin offering 6-hourly TRaPs and organize those products by worldwide storm basin.

For the fourth item, SAB plans to continue its evaluation on the NESDIS Hydro-Estimator. This is an example of a TRaP that was evaluated for Isabel, using one-hour rain rates produced by the Hydro-Estimator. This evaluation will be continued for the 2004 Western Hemisphere hurricane season.

The fifth item on the list addresses an effort ot develop a Tropical Rainfall Nowcaster (TRaN) for storms over land.

8.2 An Example of Digital Output to Users

This past summer, a TRaP graphic from the 2002 hurricane season was successfully sent to the NCEP TPC. They were able to convert the digital file to an image product and display it on their N-AWIPS workstation. In the future the plan is to take TRaPs that are generated and delivered to the NCEP TPC, display them on their N-AWIPS workstations and then make them available to the NCEP HPC through their N-AWIPS workstations. Since the TRaP product will not be on AWIP’s at NWS forecast offices, hopefully those offices interested in obtaining TRaP guidance will be able to somehow display it in their office so they can use it as guidance for rainfall prediction when a storm is within 24 to 36 hours of landfall.

8.3 Tropical Rainfall Nowcaster (TraN)

Rod Scofield of NESDIS has started development on a tropical rainfall nowcaster for storms that are over land, referred to as the TRaN.

As mentioned earlier, the current microwave sensor and algorithms tend not to do as well with identifying rain over land. This slide provides a preview of what the tropical rainfall nowcaster will look like for storms over land. The slide is from one of Rod Scofield's latest presentations that he has shown at various conferences.

The product is a nowcasting tool that uses the GOES Hydro-Estimator hourly rain rates as input for when tropical disturbances are within 6 hours of landfall, to when they dissipate over land. Possibly further in the future, the Hydro-Estimator or a blended GOES-Microwave Rain Rate product can be used to produce projections of rainfall like the TRaN.

9.0 Summary

9.1 TRaP Summary

• TRaP is an objective satellite-based technique for forecasting rainfall from tropical cyclones expected to make landfall within 24 to 36 hours
• TRaP is most useful while a storm is offshore, where few rainfall observations are available
• Users of the TRaP product include a broad community of domestic and international forecasters

To summarize, TRaP is an objective, satellite-based technique that provides guidance for rainfall totals from tropical cyclones that are expected to make landfall within a 24 hour period. In some cases the product is generated out to 36 hours if SAB thinks the storm's rainfall will be relatively stable as the storm approaches the coast. TRaP is most useful offshore where there are few additional rainfall observations. HPC forecasters use an integrated approach including numerical and statistical guidance products, radar, satellite data, and the TRaP product in producing their QPFs for landfalling tropical cyclones. There is significant collaboration between scientists, modelers, and forecasters in a coordinated effort to produce the best rainfall total estimates.

Users of the TRaP product include a broad community of domestic and international forecasters.

TRaP initiatives include:

• Providing digital TRaP output to users
• The continuation of validation exercises
• Development of a TRaN Product for storms over land

Polar Satellite Microwave Sensing Capabilities for Observing Rainfall

Near-term:

• Continued microwave observation of global rain rates with existing POES satellites, offering local coverage up to six times daily
• Global coverage thru 2010 with NOAA-AMSU, MetOp-AMSU,DMSP-SSM/I, and DMSP–SSMIS

Long-term:

• Enhanced microwave sensor technologies with NPP (expected 2006) and NPOESS (expected 2010), offering:


• Continued local coverage with a three satellite operational configuration


• More timely availability of data and derived products (95% within 28 minutes)


• Increased spatial coverage and elimination of orbital gaps


• Increased spatial resolution and improved diagnosis of convective rain rates


• Enhanced spectral coverage and polarization capabilities for improvement in rain rate estimates
• MetOp is expected to continue mid-morning AMSU coverage from 2005 thru 2019

9.2 Contacts, Resources, and Internet Links

This slide lists some contacts, resources, and verification links for TRaP. SAB is not afraid to show its verification data even if it does not always meet expectations. Forecasters know how difficult rainfall predictions are. Obtaining quality observational rainfall data can be difficult and collecting that data is often a chore. This is especially true in tropical cyclone situations with the wind blowing in excess of 50 mph (43 kt, 22 m/s). At NESDIS SAB, personnel have made an effort to verify TRaPs with total rainfall amounts observed over land. Anyone is more than welcome to browse the verification FTP sites shown here. They will give you a clearer sense of how NESDIS is verifying the TRaP product.

Contacts for More Information:

• Sheldon.Kusselson@noaa.gov.................for TRaP info


• Ralph.R.Ferraro@noaa.gov......................for Microwave info


• Antonio.Irving@noaa.gov........................for Automated TRaP

Resource Links:

TRaP Homepage at NESDIS
Site provides access to the latest TRaP (Tropical Rainfall Potential) information and products. Includes a link to TRaP graphics and rainrate imagery archived for the past year's tropical cyclones covering both Western and Eastern Hemispheres.

Polar Satellite Learning Resources
Learning resources for polar meteorological satellites. Links provide information on polar satellites, imagery, derived products, data access, and related training.

Atlantic and Caribbean Tropical Satellite Imagery at NESDIS
Site provides access to real-time imagery for tropical regions from NOAA, DMSP, TRMM-TMI, GOES, METEOSAT and GMS satellites. Visible and infrared imagery are included from GOES, GMS and METEOSAT and microwave image products (brightness temperature, precipitable water, rain rate, ocean surface winds (DMSP-SSM/I only)) from NOAA, DMSP and TRMM-TMI. Links to polar satellite-based sea surface temperatures are also provided.

East and Central Pacific Satellite Imagery at NESDIS
Site provides access to real-time imagery for tropical regions from NOAA, DMSP, TRMM-TMI, GOES, METEOSAT and GMS satellites. Visible and infrared imagery are included from GOES, GMS and METEOSAT and microwave image products (brightness temperature, precipitable water, rain rate, ocean surface winds (DMSP-SSM/I only)) from NOAA, DMSP and TRMM-TMI. Links to polar satellite-based sea surface temperatures are also provided.

Verification Links:

TRaP Verification Statistics for 2002 Atlantic Basin Hurricane Season
Access to a table and charts of various statistical performance measures applied to the TRaP technique throughout the 2002 Atlantic hurricane season.

Select TRaP Rainfall Forecast Verification Graphics
ftp://gp16.wwb.noaa.gov/pub/volcano/Kusselson/JHTTRaPObsModelExamples.ppt
Access to TRaP rainfall forecast verification graphics for a selection of landfalling tropical cyclones during the 2001, 2002 and 2003 Atlantic Hurricane Seasons.

9.3  References

There have been a number of references, especially in the last four years on the microwave portion of the TRaP technique. This page lists those references and where they appear in the literature.

Ferraro, R.R., P. Pellegrino, S.J. Kusselson, M. Turk, and S.Q. Kidder, 2002: Validation of SSM/I and AMSU derived tropical rainfall potential (TRaP) during the 2001 Atlantic hurricane season. NOAA Tech. Report NESDIS 105, 43 pp.

Kidder, S.Q., S. Kusselson, J. Knaff, R. Kuligowski, 2001: Improvements to the experimental tropical rainfall potential (TRaP) technique. Preprints, 11th Conf. on Satellite Meteorology and Oceanography, Madison, WI, Amer. Meteor. Soc., 375 - 378.

Kidder, S.Q., M. Goldberg, R. Zehr, M. DeMaria, J. Purdom, C. Veldon, N. Grody, and S. Kusselson, 2000: Satellite analysis of tropical cyclones using the advanced microwave sounder unit (AMSU). Bull. Amer. Meteor. Soc., 81, 1256-1257.

Scofield, R.A., R. Kuligowski, and C. Davenport, 2003: From satellite quantitative precipitation estimates (QPE) to nowcasts for extreme precipitation events. Preprints, 17th Conference on Hydrology, Amer. Meteor. Soc.

Spayd, L.E., Jr., and R. A. Scofield, 1984: A tropical cyclone precipitation estimation technique using geostationary satellite data. NOAA Tech. Memo. NESDIS 5, 36 pp.

Turk, M.A., S.J. Kusselson, R.R. Ferraro, P. Pellegrino, S. Kidder, and J. Yang, 2003: Validation of a microwave-based tropical rainfall potential (TRaP) for landfalling tropical cyclones. Preprints, 12th Conference on Satellite Meteorology and Oceanography, Amer. Meteor. Soc.