1 (1), 1 Journal of Operational Meteorology Article



Download 113.37 Kb.
Page1/2
Date19.10.2016
Size113.37 Kb.
#4250
  1   2

Lastname (Author #1), A. B., C. D. Lastname (Author #2), and E. F. Lastname (Author #3), 2013: Template for properly formatted Article submissions. J. Operational Meteor., 1 (1), 17.

Journal of Operational Meteorology

Article
banner2


Use of Satellite Tools to Monitor and Predict “Super Storm Sandy 2012” – Current and Emerging Products
Michael J. folmer

University of Maryland/ESSIC/CICS, College Park, MD
Mark Demaria

NOAA/NESDIS, Fort Collins, CO
RALPH FERRARO

NOAA/NESDIS, College Park, MD
JOHN BEVEN, MICHAEL BRENNAN, NOAA/NWS, Miami, FL
JAIME DANIELS, JOHN KNAFF, ROBERT KULIGOWSKI, SHELDON KUSSELSON, HUAN MENG, STEVE MILLER, SCOTT RUDLOSKY, TIM SCHMIT, CHRIS VELDEN, BRAD ZAVODSKY
(Manuscript received Day Month Year; in final form Day Month Year)
ABSTRACT

From a meteorological perspective, “Super Storm” Sandy was a “perfect storm” in terms of the convergence of several synoptic features which phased together along the mid-Atlantic coastline to create record low-pressure, a huge wind field with corresponding storm surge and copious amounts of precipitation in some areas, including record snowfall. The two features that phased – Hurricane Sandy moving northward along the Atlantic seaboard, and a strong mid-latitude winter season type disturbance – alone would have caused significant weather and disruptions in the area. But the combined impacts of Sandy to the public were nearly unthinkable for this region in terms loss of life, property damage, economic loss, and coastal flooding and erosion.
The forecast predictions by the major numerical weather centers in the United States and in Europe were generally very good, with models such as NOAA’s GFS and ECMWF’s model forecasting Sandy several days in advance. Satellite data play an important role in the initialization of these models. Additionally, NOAA/NESDIS satellite data and derived products are vital to national and regional forecasters who are responsible for issuing warnings to the public. For example, satellite imagery aids the National Hurricane Center fixing the storm center and storm intensity trends. Water vapor and precipitation estimates from satellite aid in the assessing the location, duration and trends in where the heaviest precipitation is occurring, especially when the system is offshore and out of radar range. The purpose of this paper is to illustrate many of the products that were available during Sandy, and introduce proxy products from upcoming NOAA satellite missions that will demonstrate the future capabilities.
1. Introduction

From a meteorological perspective, “Super Storm” Sandy (henceforth simply referred to as Sandy) was a “perfect storm” in terms of the convergence of several synoptic features which phased together along the mid-Atlantic coastline to create record low-pressure, a huge wind field with corresponding storm surge and copious amounts of precipitation in some areas, including record snowfall. The two features that phased – Hurricane Sandy moving northward along the Atlantic seaboard, and a strong mid-latitude winter season type disturbance – alone would have caused significant weather and disruptions in the area. But the combined impacts of Sandy to the public were nearly unthinkable for this region in terms loss of life, property damage, economic loss, and coastal flooding and erosion. Sandy caused over 250 deaths and upwards of $70 billion in damage and economic loss during its trek from the Caribbean northward to the mid-Atlantic and Northeastern United States.

The forecast predictions by the major numerical weather centers in the United States and in Europe were generally very good, with models such as the National Oceanic and Atmospheric Administration (NOAA) National Center for Environmental Prediction’s (NCEP) Global Forecast System (GFS) and the European Center for Medium-Range Weather Forecasts’ (ECMWF) global model forecasting Sandy several days in advance. Satellite data play an important role in the initialization of these models. Additionally, satellite data and derived products are vital to national and regional forecasters who are responsible for issuing warnings to the public. For example, satellite imagery aids the National Hurricane Center (NHC) with fixing the storm center and monitoring storm intensity trends. Water vapor and precipitation estimates from satellites aid in assessing the location, duration and trends of where the heaviest precipitation is occurring, especially when the system is offshore and out of radar range. Other products such as ocean surface heat content and ocean surface winds also play a vital role to the forecasters. The purpose of this paper is to illustrate many of the products that were available during Sandy, and introduce proxy products from upcoming satellite missions that will demonstrate the future capabilities.

Section 2 summarizes the life cycle of Sandy, section 3 describes how satellite data and products contributed to the operational analysis and forecasting of Sandy’s track, intensity, structure and precipitation, and section 4 presents some emerging satellite capabilities. A summary and conclusions are presented in section 5.



2. The Life Cycle of Sandy
Fig. 1 shows the track and intensity of Hurricane Sandy. The genesis of Hurricane Sandy was as complex as its evolution and eventual track. As Hurricane Rafael moved away from the Northern Leeward Islands on 14 October 2012, a pocket of higher moisture in the “tail” was left behind in the eastern Caribbean Sea. This area of moisture interacted with a rather innocuous easterly wave that moved into the eastern Caribbean Sea around 19 October 2012 and became trapped under a developing upper-level ridge. Meanwhile, a trough passing through the southeast U.S. led to the development of a weak upper low in the southeast Gulf of Mexico on 21 October 2012. Convection increased in organization in the central Caribbean Sea in response to the upper-low development and associated diffluence aloft leading to the development of a mid-level mesoscale convective vortex (MCV) that slowly drifted west-southwest. By 1200 UTC on 22 October 2012, the MCV had spawned a surface circulation in the southwest Caribbean Sea that was organized enough to be classified by the NHC as a tropical depression.

The depression rapidly organized in the next six hours and was classified as Tropical Storm Sandy around 2100 UTC on 22 October 2012. Steady strengthening occurred over the next 42 hours and Sandy became a 70 knot hurricane at the 1500 UTC NHC advisory on 24 October 2012 just before landfall in eastern Jamaica. After crossing Jamaica, Sandy continued to strengthen up until landfall in eastern Cuba as a 100 knot, Category 3 hurricane. At this time, the upper-low in the southeastern Gulf of Mexico drifted southwest of Sandy allowing for an enhancement of the pole-ward outflow channel which extended as far north as the offshore waters of North Carolina. The hurricane’s interaction with eastern Cuba was minimal considering the rugged terrain and Sandy maintained hurricane intensity through the central Bahamas until dry air was entrained from the upper-low to the southwest. Sandy took on a hybrid-like structure from 26 October 2012 to 27 October 2012 as the storm tried to fight off the dry air. Meanwhile a series of shortwaves to the north and west of Sandy started to exert a more north-northeast track to the storm. Early on 28 October 2012, the hurricane was able to wrap its outflow around the southwest quadrant and the effects of the upper-low diminished. This allowed Sandy to re-intensify over the warm Gulf Stream waters east of North Carolina.

Late on 28 October 2012, the final in a series of shortwaves dropped from the Upper Plains of the U.S. towards the Carolinas and started to cutoff an upper-low from the jetstream just west of Hurricane Sandy. This allowed the upper-low to capture Sandy early on 29 October 2012 which led to a west-northwest motion of Sandy towards the Mid-Atlantic coastline. Sandy was considered a hurricane up until the 2100 UTC 29 October 2012 NHC advisory when it was considered a post-tropical storm. At this time, the upper-low had merged with the hurricane and the storm had acquired extratropical characteristics, including a warm seclusion structure surrounded by much colder air. What was now dubbed “Superstorm” Sandy would continue west and make landfall in southern NJ with winds along the coast near hurricane strength with stronger wind gusts experienced well north of the landfall point. “Superstorm” Sandy moved into south-central Pennsylvania and fully occluded on 30 October 2012, slowly spinning down before being picked up by an upstream shortwave around 01 November 2012.
3. Satellite contributions to the analysis and forecasting of Sandy
NHC provided forecasts of the track and maximum sustained surface winds of hurricane Sandy out to 5 days, and wind structure forecasts in terms of the radii of 34 and 50 kt winds out to 3 days and 64 kt winds to 36 h. These forecasts were updated every 6 h. General guidance on precipitation was also provided in coordination with the NCEP Weather Prediction Center (WPC). Location specific information was provided by the local National Weather Service (NWS) Weather Forecast Offices (WFOs) in the regions impacted by the storm. In this section the contributions of satellite data and products to these operational forecasts are described.


  1. Track forecast guidance

One of the most important uses of satellite data is for assimilation into forecast models. Tropical cyclones typically spend most of their lifetimes over tropical and subtropical oceans where conventional data is sparse, so satellite observations are fundamental to tropical cyclone forecasts. In some cases tropical cyclones transition to extra-tropical cyclones if they move into higher latitudes, like Hurricane Sandy did. Satellite observations are also very important for model initialization in those cases as well.

Global models are the backbone of tropical cyclone track forecasting. A wide variety of satellite data are assimilated to help determine the global model initial conditions. The satellite data are assimilated in two ways. When possible, the satellite radiances are included using variational data assimilation systems. In these assimilation systems, the model meteorological fields such as temperature and moisture are used as input to “forward” models that convert them to a form that is equivalent to satellite radiances. The meteorological model fields are then adjusted to minimize the difference between the observed radiances and model-derived radiances. A wide variety of satellite radiances are assimilated in this way. These include data from microwave and infrared sounders from geostationary and low-earth orbiting (LEO) satellites from the U.S. and the International community.

The second method for including satellite data is to assimilate derived products. The most common example of these are atmospheric motion vectors (AMVs), which are determined by tracking features in visible and infrared satellite imagery. Geostationary satellites are used over most of the globe (Daniels et al, 2010), although , in polar regions, LEO satellite coverage is frequent enough to estimate AMVs as well (Key, et al. 2010).

Ozone is an important atmospheric trace gas that affects the absorption of incoming solar radiation. Global forecast models also assimilation satellite radiances from specific channels that are sensitive to ozone to properly initialize its concentrations.

NHC uses several operational global models to forecast tropical cyclone tracks. These include the U.S. GFS, the global models from the ECMWF and U.K. Meteorological Office, and the Navy’s Operational Global Atmospheric Prediction System (NOPGAPS). Two regional coupled ocean-atmospheric prediction systems, the Geophysical Fluid Dynamics Laboratory (GFDL) model, and the Hurricane Weather Research and Forecast (HWRF) model, are also used for track forecasting. At the present time, the regional models do not directly assimilate much satellite data, but the satellite data still impacts the regional model forecasts through the use of the global models fields in initializing the outer grids and through the lateral boundary conditions.

Because of its proximity to the U.S. east coast, Sandy was extremely well sampled through the use of satellite data, supplemental aircraft reconnaissance data, as well supplemental rawinsondes that were launched over the U.S. for the several days before the final landfall along the New Jersey coast. Despite the non-climatological left-hand turn at high latitudes described above, the NHC track forecast errors were much smaller than the previous 5-year average, as can be seen in Fig. 2. These smaller than average NHC track errors were primarily due to the excellent track forecast models available for Sandy, where satellite data made an important contribution to determining the initial conditions.
b. Intensity forecast guidance
Because of the complexity of the processes involved in tropical cyclone intensity changes, a hierarchy of intensity forecast models is utilized by NHC. These range in complexity from coupled ocean-atmosphere models such as the HWRF model to much simpler statistical-dynamical models such as the Logistic Growth Equation Model (LGEM), which use statistical post-processing techniques to make intensity forecasts (DeMaria et al 2013). Similar to the track errors, the NHC official intensity errors for Sandy were 25 to 50% smaller than those from the past 5 years. The primary reason for the low intensity errors was that, except for the initial period of rapid intensification south of Cuba, the maximum wind changed fairly slowly for Sandy during much of its lifecycle, making the intensity forecasts somewhat easier than usual. The satellite data also contributed to these lower errors through their direct use in the global models, which helped to improve the regional models and the simpler statistical dynamical models, which use input from the global models. Satellite microwave, infrared, and altimetry are also used in the analysis of the sea surface temperature and sub-surface ocean structure.

The satellite data also contributes in a more direct way to the statistical-dynamical intensity models, which include predictors from the Geostationary Operational Environmental Satellites (GOES) imagery and ocean heat content estimates from satellite altimetry. Forecasting rapid intensity changes (often defined as an increase in the maximum wind of 30 kt or greater in 24 h) is an especially challenging problem, and neither the dynamical nor statistical models have shown much skill for these cases. For this reason, a specialized statistical model called the rapid intensification index (RII) has been developed, which estimates the probability of rapid intensification using model and satellite input (Kaplan et al 2010). The satellite input has a much larger influence on the RII than the other statistical techniques. Fig. 3 shows the observed maximum wind of Sandy and the probability of rapid intensification predicted by the RII. Sandy underwent a period of rapid intensification beginning at 1800 UTC on 23 October, which lasted until the storm made landfall in western Cuba early on 25 October. The RII predicted probabilities of rapid intensification of up to 50% during this period, which is 10 times higher than climatological mean for the Atlantic basin. These very high probabilities were due to the very high sea surface temperatures and large ocean heat content in the Caribbean in an atmospheric environment of low shear. The GOES data indicated that Sandy was very convectively active at this time, further increasing the probability of RI. Once Sandy moved north of Cuba, the oceanic and atmospheric environments were much more hostile for intensification, and the probability of rapid intensification remained very low for the rest of Sandy’s life cycle. There was a brief period of re-intensification early on 29 October, but it did not satisfy the criterion for rapid intensification. Thus, the satellite data input to the RII provided very valuable information for helping to forecast the intensity changes of Sandy.

The next generation GOES satellites beginning with GOES-R in late 2015 will include a Geostationary Lightning Mapper (GLM). The GLM will provide nearly continuous measurements of total lightning times and locations over most of the North Atlantic and eastern North Pacific tropical cyclone basins. Because lightning activity is related to cloud updraft speed in the mixed phase region of convection (Black and Hallet 1999), it has the potential to provide information about TC intensity changes. To prepare for GOES-R, an experimental version of the RII with lighting input from a ground-based network has been run as part of the Proving Ground since 2010. DeMaria et al (2012) have shown that enhanced lightning in the rainband region (100-200 from the center) of TCs is a predictor of RI, and that enhanced lighting in the eyewall region indicates that the RI is near its end. Fig. 4 shows the lightning locations over a 6 h period from the World Wide Lightning Location Network (WWLLN) during the time when Sandy was beginning its RI phase. There was considerable lightning activity away from the storm center, especially to the east of the storm, but very little at the storm center, consistent with the relationship described by DeMaria et al. (2012).
c. Wind structure prediction
The NHC official forecast includes several parameters that describe the storm structure. The size of the surface wind field is represented by the radius of 34, 50 and 64 kt winds to the NE, SE, SW and NW of the storm center. Forecasts of the 34 and 50 kt wind radii are provided out to 72 h and 64 kt wind radii estimates are provided to 36 h. The radii values represent the maximum extent in each quadrant and are reported in units of n mi (1 n mi = 1.85 km). Estimates and forecasts of the wind structure are important as the relative size of a TCs wind field has direct implications for the potential damage a given storm may cause (Powell and Reinhold 2007; Maclay et al. 2008). Both the potential direct wind damage and the potential coastal inundation increases with the size of the wind field (e.g., Irish, et al. 2008, Lin et al. 2013). In addition, the onset of gale and hurricane force winds occurs earlier in larger tropical cyclones, which can hamper pre-storm mitigation efforts. The assessment of the wind field of Sandy was particularly important given its overall size and its forecast to make landfall in a highly populated and vulnerable location.

The NHC forecast also provides an estimate of the storm type. Official forecasts are provided only for cyclones classified as tropical or subtropical. However, the forecast also provides guidance on when dissipation or transition to an extra-tropical cyclone will occur. As described above, Sandy had a complex interaction with mid-latitude weather systems so the structure prediction was especially challenging. A number of satellite data and products are used as guidance for the structure analysis and prediction to complement the in situ and aircraft data. Once Sandy moved north of the Bahamas it became a very large storm, and much of the storm wind field extended well beyond the standard length of the aircraft reconnaissance flight legs. Also, the cyclone moved through a region with relatively sparse ocean buoy coverage, so the satellite data and products were especially useful during Sandy. These included ocean surface winds, satellite atmospheric motion vectors, the Advanced Microwave Sounding Unit (AMSU) products, a multi-platform tropical cyclone surface wind analysis system, multi-channel imagery products and an objective storm type classification algorithm.





  1. atmospheric motion vectors

Forecasters rely on subjective interpretation of satellite imagery and satellite-derived Atmospheric Motion Vectors (AMVs) to aid in analyzing meteorological conditions over oceanic regions and in the vicinity of tropical cyclones like Sandy (Weldon and Holmes, 1991; Velden et al, 2005). Fig. 5 shows a 3-day loop of GOES-13 infrared window imagery with associated AMVs in the vicinity of hurricane Sandy as it moved up the east coast and made landfall in New Jersey.

Satellite AMVs are often the only available wind observations in these situations and provide forecasters with key information on the wind structure, especially with respect to the depiction of upper level wind features and their evolution, which can play a role in tropical cyclone formation and motion. A continuing challenge to forecasters is the interaction of tropical cyclones with upper level troughs, which often affect the storm motion and intensity. Satellite AMVs, which are made available to forecaster display systems, help characterize these interactions for the forecasters and they prepare their forecasts. Forecasters also rely on Numerical Weather Prediction (NWP) model guidance when preparing their forecasts. Satellite AMV observations are critical observations that contribute to improved NWP model forecasts. Today, AMVs are used more effectively within NOAA NWP model suite (NCEP’s GFS/GDAS; GFDL’s hurricane model) and have been demonstrated (Soden et al, 2001; Goerss et al, 1998) to improve the initialization of the wind field, model forecasts of the storm recurvature, and improve the vorticity gyres in the environmental flow. In the case of Hurricane Sandy, NOAA model forecasts correctly captured the strong westward curvature of Sandy and helped forecasters issue very accurate warnings as it made landfall in New Jersey. AMVs are also used by forecasters to gauge the extent of shear affecting a tropical cyclone or in the case of Sandy, the encroaching mid-level storm system that would initiate the extratropical transition.


  1. amsu products

One of the key characteristics defining a tropical cyclone is a deep warm core structure in the thermal field. As a TC undergoes transition to a more extra-tropical type system, the warm core structure typically erodes from the lower-mid troposphere as cold air advection near the surface intrudes into the circulation. This can often be depicted in AMSU-derived x-sections of thermal anomalies. One such product that NHC often relies on is derived in near real-time by CIMSS and provided on a web site.

A good example of the utility of the AMSU data to depict this transitional phase occurred just prior to and after the landfall of Sandy. About 5 hrs before landfall, the AMSU data showed a deep warm core still existed, particularly strong in the mid levels, indicative of tropical origin (Fig. 6). About 7 hrs after landfall, the warm core has eroded considerably (Fig. 7). By this time the NHC had declared Sandy post-tropical. The issue of whether to keep Sandy a hurricane or transition to an extratropical storm just before landfall was quite contentious. The AMSU data supported other aircraft-based observations in making this key decision.

The AMSU data is also used to provide wind estimates in the storm environment. Temperature retrievals are used as input to the hydrostatic equation to estimate the geopotential height field at the standard pressure levels from 100 to 1000 hPA. The nonlinear balance equation is the used to estimate the wind field (Bessho et al. 2006). These winds are routine generated by NCEP operations, and are used as input to the operational surface wind product described in the next subsection. The horizontal resolution of the AMSU data is too coarse (50 km at nadir) to resolve the tight height gradients near the storm center, the method provides reasonable wind estimates in the outer portions of the storm.




  1. multi-platform tropical cyclone surface wind analysis product

As described above, remotely sensed data is very useful for the assessment of hurricane wind fields. However, satellite-based surface and near-surface wind estimates come from a variety of sources and each source has its own distinct strengths and weaknesses. The interpretation required to combine this information requires extra time – time that is often not available in the operational setting. To overcome some of the issues associated with combining and the interpretation of multiple satellite-based surface and near-surface wind information in the hurricane environment, the National Environmental Satellite, Data, and Information Service (NESDIS) produces the operational Multi-satellite-platform Tropical Cyclone Surface Wind Analysis (MTCSWA), which combines many of the satellite–based surface and near-surface wind fields for all active global tropical cyclones (Knaff et al. 2011). Inputs include previously described AMSU non-linear balance winds, atmospheric motion vectors, scatterometry, and infrared (IR)-based flight-level proxy winds described in Mueller et al. (2006) and produces wind estimates with generally smaller errors than the raw input data. These analyses were available in time for consideration by NHC forecasters for the preparation of their advisories and forecasts. An example of this product and the four inputs produced on 29 October at 18UTC is shown in Fig. 8 when Sandy was off the coast of Delaware and its large wind field was beginning to be felt along the mid-Atlantic coast. FOLMER - (OSCAT mention)


  1. multi-channel imagery products

Visible, infrared and water vapor imagery from the GOES satellites as well as microwave imagery from LEO satellites were used during Sandy for analysis of convective structure and organization and interactions with synoptic features in the storm environment. The next generation of GOES, beginning with GOES-R, will include a 16 channel Advanced Baseline Imager (ABI). The current GOES imager includes 5 channels. With this large increase in available channels, forecasters will not have time to fully utilize all the information from the imager. For this reason, multi-channel imager products that are designed to highlight features of interest will become increasingly important. Several of these products are already being utilized by NHC, as described in more detail below.

As part of the GOES-R Proving Ground activities at the NHC, the WPC, the Ocean Prediction Center (OPC), and the NESDIS Satellite Analysis Branch (SAB), the Red Green Blue (RGB) Air Mass product has been introduced to forecasters to monitor tropical cyclones and potential extratropical transitions (ET). The complicated ET of Sandy was captured with this product using the GOES-Sounder, the Moderate Resolution Imaging Spectroradiometer (MODIS) on NASA’s Aqua and Terra satellites, and the combined Visible Infrared Imaging Radiometer Suite (VIIRS) and Cross-track Infrared Sounder (CrIS) on the new Suomi-National Polar-orbiting Partnership (S-NPP) satellite. Forecasters had been introduced to this product earlier in the year and found it useful in identifying potential vorticity (PV) anomalies and possible stratospheric intrusions. As the complex barotropic-baroclinic interaction began to unfold, some interesting features were noted in the RGB Air Mass product, shown in Fig. 9.

Shortly after crossing eastern Cuba on 26 October, Sandy started to acquire an atypical hybrid hurricane appearance as dry air was entrained into the center of the cyclone. Some of this was due to the interaction with higher terrain over Cuba, but more notably was a persistent upper-level low to the west-southwest of Sandy that provided the drier air at mid-upper levels. This was analyzed in the RGB Air Mass product with an orange coloring that moved towards the center of the agitated hurricane. By 1800 UTC on 26 October, Sandy exhibited a comma-shape as the dry air, possibly stratospheric in nature, overtook the eastern quadrant of the cyclone, while any convection was displaced to the north-northwest portion of what was once an eye. This was an unusual structure and may have indicated the first signs of the approaching ET.

As early as 1800 UTC on 27 October, much of the dry, possibly stratospheric air was starting to become mixed out near the cyclone center, and the deeper orange-coloring was becoming more displaced to the east-northeast. This indicated that the first upper-low had weakened and most likely become part of the overall circulation of Sandy, which may have assisted in the expanding wind field and overall areal extent of the storm. Meanwhile, upstream, a series of modest shortwaves were diving into the center of the country helping to carve out a deep trough along the east coast in response to a strong blocking high near Greenland. These disturbances are evident in the RGB Air Mass product as light orange/red colored features moving towards the Carolinas on 28 October.

By 1800 UTC on 28 October, the first evidence of a transfer of energy to Sandy becomes apparent over the eastern Ohio Valley as shortwave energy (orange/red coloring) becomes more enhanced and the cloud shield from Sandy begins to expand farther west. Incidentally, Sandy begins to look more tropical again around 0600 UTC on 29 October as the storm crosses the Gulf Stream. Meanwhile, the aforementioned shortwave trough begins to overtake the western portion of Sandy as the orange coloring expands northwest to southeast, enhancing the southwest ouflow channel of Sandy. This indicates that more stratospheric drying is taking place and becoming increasingly entrained into the hurricane’s circulation.

As early as 1500 UTC on 29 October, there is increasing evidence of the ET underway as the orange coloring becomes more apparent in the imagery with a very sharp gradient developing over NC and points east. At this time, ASCAT high resolution winds show a secondary wind maximum has developed in the southwest quadrant, near a bent-back front structure. At this point, it appears the hurricane is still a somewhat separate part of the overall storm structure. A full ET had been observed by 2100 UTC on 29 October as indicated by surface, aircraft, and satellite data. At this point, the imagery shows a very amorphous looking cyclone with a very strong gradient of green to orange off the Delmarva peninsula. Most likely, the stratospheric intrusion is deeper into the atmosphere than indicated by the orange coloring alone due to the limited view by the product (~300 hPa to 500 hPa).

By 0000 UTC on 30 October, post-tropical storm Sandy has made landfall in southern NJ and a very strong push of stratospheric drying is racing north around the eastern quadrant towards Long Island and the northern NJ coastline. This is part of the shortwave energy that was deepening over the eastern Ohio Valley. The secondary wind maximum at this time had moved to the eastern quadrant in association with the deep orange coloring in the imagery and wound up producing very strong wind gusts in many locations along the coast, coupled with the storm surge. By 1200 UTC on 30 October, post-tropical storm Sandy is full occluded and the oranges have intruded into the center of the now vertically stacked cyclone. There is even some evidence of blue coloring mixed in the cloud shield near the Appalachians were the incredible snow amounts were observed.


  1. objective storm type classification algorithm

As described above, Sandy was an especially challenging case because of the complex interactions with mid-latitude systems, and the transition to an extra-tropical cyclone right before landfall in a heavily populated region of the U.S. Forecasters use a variety of information sources to make their subjective determination of the storm type. These include global and regional model, and variables diagnosed from the model fields. Hart (2003) showed that thermal wind and circulation asymmetries estimated from global model fields are especially useful in determining whether a cyclone is tropical, subtropical or extra-tropical. These parameters are used in three-axis phase space diagrams that are available in real time from http://moe.met.fsu.edu/cyclonephase/ and often utilized by NHC foreacasters. The only input required for these diagrams is the geopotential height field. To provide a purely observation based version of these diagrams, a version is provided that uses the geopotential height analyses from the operational AMSU wind product previously described.

To provide objective guidance for estimating the storm type, a classification algorithm was added to the operational SHIPS model, beginning with the 2011 hurricane season. The product uses a linear discriminant technique to classify a cyclone as tropical, subtropical or extra-tropical. The input to the algorithm includes parameters from the GFS that are similar to those in the Hart (2003) phase space diagrams, ocean input, and inner core convective features from GOES data. This algorithm worked very well during Hurricane Sandy and correctly predicted the transition from tropical to extra-tropical.
c. Precipitation estimates and forecasts
Satellite derived moisture and precipitation fields serve as complimentary sources of information to weather forecasters because they provide information over regions where in-situ information is limited, in particular, over the open oceans. In this section we provide descriptions of several NESDIS derived products.


  1. blended total precipitable water (bTPW)

In terms of passive microwave products, Total Precipitable Water (TPW) is perhaps the most widely used to support real-time forecasting applications, as it accurately depicts tropospheric water vapor and its movement. In particular, it has proven to be extremely useful in determining the location, timing, and duration of “atmospheric rivers” which contribute to and sustain flooding events (Ralph et al. 2004). Such phenomena are detectable across the globe, with the source of the “river” typically emanating from the low-latitudes.

Early uses of TPW were developed with the Special Sensor Microwave/Imager (SSM/I) (Kusselson 1993), however, the method has been expanded to include data from all passive microwave sensors, including AMSU. Recently, a multi-sensor approach has been developed and implemented at NESDIS in which passive microwave estimates from multiple satellites and sensors are merged to create a seamless TPW product that is more efficient for forecasters to use (see http://www.osdpd.noaa.gov/bTPW/). Additionally, ground-based estimates over land that are derived from the Global Positioning System (GPS) are also included to anchor the satellite estimates, especially for product continuity across coastlines (Gutman et al. 2004). Details on the blended TPW (bTPW ) can be found in Kidder and Jones (2007).

The time sequence of the bTPW product for Sandy is presented in Fig. 10. Forecasters at NESDIS/SAB who provide national precipitation forecasts in conjunction with WPC which are then collectively used by the local NWS WFOs, began to examine the bTPW (and a companion percent of normal product – not shown) to examine the rainfall potential from the system as it began its northward track from the Caribbean Sea towards the U.S. east coast. Interestingly, several days prior to Sandy, atmospheric moisture was streaming northward from the system which interacted with a frontal system that moved offshore. After a brief drying period, extremely moist air associated with Sandy slowly worked northward. Note the extremely high values near the storms core. Also, in regions of extreme rainfall, the retrieval algorithm becomes unreliable and thus the missing TPW values. As Sandy made landfall, very moist conditions persisted for several days and contributed to rainfall in excess of 8 inches in the mid-Atlantic. As Sandy became absorbed into the mid-latitude cyclone, the atmospheric moisture slowly decreased.



  1. Hydroestimator (H-E)

Until the late 1990’s, operational satellite precipitation estimates from NESDIS were produced by a largely manual technique called the Interactive Flash Flood Analyzer (IFFA). Because of manpower constraints, estimates could only be produced for a limited number of relatively small regions at a time. In response, NESDIS scientists developed the Auto-Estimator (A-E), which automated many IFFA functions and allowed real-time production of satellite precipitation estimates over the entire CONUS with a much shorter lead time than was possible with manually-produced estimates (Scofield and Kuligowski, 2003). The second generation of the A-E is the Hydro-Estimator (H-E), which improved over the A-E in distinguishing non-raining cirrus clouds from convectively active, precipitating clouds. All of these products have been the cornerstone of operational satellite precipitation estimation by NESDIS/SAB for the past three decades.

As was described in the previous section, SAB and WPC generated national precipitation nowcasts and forecasts that are utilized by the NWS WFOs. As Sandy approached the U.S. mainland, rapid update rainfall estimates were made to help the local offices monitor rainfall and issue necessary flood warnings; a total of 11 Satellite Precipitation Estimate Statements (SPENES) were issued by SAB. Fig. 11 shows both the 24-h and 96-h totals for the H-E and Stage IV ending at 0600 UTC on 29 October 2012 and 0000 UTC on 31 October 2012, respectively. As can be seen, the H-E did a reasonable job with the tropical parts of Sandy (i.e. in NC), but underestimated the rainfall in the areas of tropical transition in the Mid-Atlantic States when the cloud tops warmed and the H-E performance weakens. Future techniques under development which fuse satellite microwave estimates with the rapid update GOES visible and IR estimates have shown promise to improve the rainfall estimates (Kuligowski 2002)




  1. Ensemble Tropical Rainfall Prediction (eTRaP)

Heavy rains associated with land falling tropical cyclones (TCs) frequently trigger floods that cause millions of dollars of damage and lost lives. To provide observations-based forecast guidance for TC heavy rain, Kidder et al. (2005) developed the Tropical Rainfall Potential (TRaP), an extrapolation forecast generated by accumulating rainfall estimates from microwave sensors over a 24 hour period as the storm is translated along the forecast track (Kidder et al. 2005). TRaP aims to predict the maximum rainfall at landfall, as well as the spatial pattern of precipitation, and has been shown to have similar or better skill than short-range numerical weather prediction models (Ferraro et al. 2005, Ebert et al. 2005)

A recent innovation has been to combine the TRaP forecasts from multiple sensors and various start times into an ensemble TRaP product known as eTRaP (Ebert et al. 2011). The ensemble approach provides not only more accurate quantitative precipitation forecasts, including more skillful maximum rainfall amount and location, it also produces probabilistic forecasts of rainfall exceeding various thresholds that decision makers can use to make critical risk assessments. Ebert et al. (2011) showed that eTRaP probabilistic forecasts have useful skill, but the grid-scale probabilities are biased high when compared to observations and should be interpreted in a broader spatial context. Efforts to calibrate the probabilistic forecasts from eTRaP are underway.

Shown in Fig. 12 is the 24-hour eTRaP estimate ending 0600 UTC on 29 October 2012 as well as the probability of rainfall exceeding 100 mm. The maximum estimate was approximately 200 mm; the region of predicted rainfall exceeding 100 mm matches very well with the Stage IV estimate on the Del-Mar-Va coast, but appears to have overestimated the rainfall further inland where the characteristics of the rain were more stratiform in nature.


  1. Microwave snowfall rates

After several years of operating in an experimental mode, a new passive microwave snowfall rate algorithm (SFR) was implemented into NESDIS operations in the fall of 2012 (Meng et al. 2012). Although the product is best suited for use in sparse surface data and radar coverage regions, it is interesting to evaluate its utility and performance during Sandy because of the diverse nature of the storm and the excessive snowfall that was experienced in the higher altitudes of the mid-Atlantic region.

Currently, the AMSU SFR is retrieved using data from three satellites: NOAA-18, -19, and Metop-A. Fig. 13 shows the NOAA-18 SFR at 1924 UTC on 29 October and the NEXRAD radar reflectivity images with both 0 min and 180 min time lag. The light blue in the SFR image represents the AMSU retrieved rain area. West Virginia is marked in the NEXRAD images to provide a geographic reference. Comparisons with ground observations showed reasonable agreement in the overall snowfall rates. Because Sandy was so early in the snowfall season and the ambient conditions were relatively warm compared to winter conditions, the SFR product did have some problems in properly detecting the precise area of the snowfall.



Download 113.37 Kb.

Share with your friends:
  1   2




The database is protected by copyright ©ininet.org 2024
send message

    Main page