Detecting tornado tracks using Synthetic Aperture Radar (SAR) imagery

NASA SPoRT has been working to support the NWS’s use of the Damage Assessment Toolkit (DAT) by integrating multiple satellite datasets into the DAT framework to assist in damage surveys.  Imagery from MODIS, VIIRS, and Landsat 8 are available daily within the application while imagery from higher resolution satellites, such as Terra ASTER and other high-resolution commercial imagery are facilitated by our partnership with the USGS’s Hazard Data Distribution System ( One new area being explored is the application of Synthetic Aperture Radar (SAR) imagery to detect tornado damage.


Zoomed-in section of the SAR change detection RGB generated from Sentinel-1B imagery from May 10 and May 22, 2017.  The damage indicators show preliminary track information as of June 2, 2017 and are not considered final.

On the evening of May 16th, 2017, a supercell tracked across Wisconsin producing a strong tornado. The resulting 83-mile long track tornado produced EF-3 damage.  Shown in the image above is a prototype change detection RGB using data from Sentinel-1(A/B), a European Space Agency (ESA) satellite with a SAR instrument on board. Unlike optical sensors, which observe surface reflectance and temperature, SAR instruments measure backscatter from the surface, allowing the instrument to be used at all times of the day and in any sky conditions. SPoRT has been working with the Alaska Satellite Facility, NASA’s SAR Distributed Active Archive Center (DAAC) to receive these products for evaluation and put them in the DAT to help with the identifying of damage tracks, especially in areas where damage surveys can be more challenging (i.e. forested areas, poor road network).  The RGB takes advantage of the dual polarization from the sensor, assigning the VV and VH corrected polarization from the post-event granule to the red and green channels of the RGB, respectively.  The blue channel is a difference image of the VH polarization (same as what is used in the green channel) from the before and after granules.  The resulting RGB will show any changes between the two granules in a aqua/periwinkle/purple-color.  Although the RGB will show all change between the granules over the ~12-day period (i.e. agricultural growth), tornado tracks tend to be linear, making it a possible to discern/identify the damage track.  Without the hindrance of clouds that constantly plague damage detection in optical imagery, SAR imagery offers another tool to operational forecasters for use during damage surveys.  The team is also working on other change or anomaly detection techniques to facilitate mapping of tornado and severe weather damage.

Stark contrast in Eastern U.S. soil moisture following Hurricane Matthew

Stark contrast in Eastern U.S. soil moisture following Hurricane Matthew

Major Hurricane Matthew left a trail of destruction in its wake from the Caribbean up through the U.S. East Coast.  As Hurricane Matthew tracked northward along a large portion of the U.S. Southeast Coast from Florida to North Carolina, the rainfall impacts worsened.  Figure 1 shows the weekly rainfall spanning 4-11 October, ranging from ~2-8 inches along the Florida East Coast to 10-20 inches in the eastern Carolinas.  Since antecedent soil moisture was highest in the eastern Carolinas (Fig. 2), the extreme rainfall led to the most serious flooding in this area.


Fig. 1.  Weekly rainfall totals from 4 – 11 October 2016.


Fig. 2.  Total Column (0-2 m) relative soil moisture prior to Hurricane Matthew’s impact on North and South Carolina, valid at 0000 UTC 7 October 2016.

Referring back to the precipitation totals in Fig. 1, we can see that there was a sharp rainfall gradient on the northwestern edge in the Middle Atlantic region.  Interestingly, this gradient in Hurricane Matthew’s rainfall coincided with a pre-existing transition zone between wet conditions near the Atlantic coast and drought conditions further inland from the Appalachians through New England.  The net result was to accentuate the wet-dry contrast already in place.  The animation in Fig. 3 highlights this contrast nicely by presenting the SPoRT-LIS daily total-column relative soil moisture percentiles from 1-12 October.  The percentiles are based off a 1981-2013 daily soil moisture climatology that SPoRT produced from its ~3-km resolution SPoRT-LIS simulation.  By 9 October, notice the incredible transition from excessively wet soil moisture exceeding the 98th percentile (Carolinas through the southern half of Delaware) to extremely dry soil moisture less than the 5th percentile across Pennsylvania into the Northeast (as well as much of the inland Southeastern U.S.).  In fact, total column soil moisture values are less than the 2nd percentile over a large part of Ohio, Pennsylvania, New York, and the New England states, indicative of the ongoing severe drought there.


Fig. 3. Daily animation of SPoRT-LIS total column relative soil moisture percentile from 1 to 12 October 2016.

Louisiana Flooding Captured by SPoRT-LIS

About a week ago, southern Louisiana began to experience heavy rainfall from a storm system that remained relatively stationary over the Gulf Coast.  The SPoRT-LIS, a real-time, high-resolution implementation of the the NASA Land Information System, captured trends in soil moisture that provide some insight into the evolution of this flooding, including hints at precursor conditions that may have led to the extreme nature of this event.

The 0-200 cm integrated relative soil moisture (RSM) fields have been used in the past to identify flood precursor conditions.  These fields give an indication of the total amount of water in the soil moisture column and provide information about how much additional precipitation can be accepted by the soil before all becomes runoff into nearby streams and rivers.  About 2 weeks ago (August 3 00Z; Fig. 1), Southern Louisiana showed soil moisture values in the 50% range, which are higher than other parts of the country, but likely about normal given the swampy nature of that region.  However, following a couple of precipitation events in that area on August 3, 7, 9, and 10), these integrated RSM fields bump up the 60-65% range (Fig. 2), which has become somewhat of an unofficial threshold for antecedent saturated soils that could lead to areal flooding events.  Based on various reports, it appears that the official start of the flooding event began on August 11.


Fig. 1: SPoRT-LIS valid at 00Z on 03 August 2016 showing 0-200 cm integrated relative soil moisture values around 50% over Southeastern Louisiana.


Fig. 2: SPoRT-LIS 0-200 cm integrated relative soil moisture values valid at 00Z on 10 August 2016 showing impact of multiple precipitation events since the 03 August figure above. Soil moisture values are elevated in southeastern Louisiana to values around 60-65%.



Fig. 3: SPoRT-LIS 0-200 cm integrated relative soil moisture values valid at 00Z on 14 August 2016 following the heaviest precipitation. Soil moisture values are above 90% in most areas, indicating major ongoing flooding across much of southern Louisiana.

Starting on 11 August, the 0-200cm integrated RSM begins to exhibit signs of flooding (starting to get into 70-80%; not shown).  By Aug. 12, most of SE LA is above 80% Integrated RSM with pockets above 90% (not shown).  By Aug. 14 (Fig. 3), nearly all of southern Louisiana is covered with soil moisture values above 85-90%, which indicates major ongoing flooding in this area.

These products are provided to select National Weather Service partner offices to aid in these flooding forecast challenges.  For more details on this product and to view additional days or hours, please visit the real-time SPoRT-LIS page.

April 27, 2011–Five Years Later: A Satellite Imagery Perspective

On April 27, 2011, a severe weather outbreak occurred across the southeastern United States, resulting in 199 tornadoes across the region and over 300 fatalities (NWS 2011 Service Assessment).  Alabama was among the states hardest hit, with 68 tornadoes surveyed by the National Weather Service (NWS) Weather Forecast Offices (WFOs) in Huntsville, Birmingham, and Mobile, Alabama, and over 250 reported fatalities in the state. Huntsville, home to NASA’s Marshall Space Flight Center and the Short-term Prediction Research and Transition (SPoRT) Center, lost power along with most of Madison County after tornadoes severed major utility lines.  The power outage lasted well over a week in some areas. Once power was restored, SPoRT team members were able to provide satellite imagery to our partners in the National Weather Service to help clarify some of the high-intensity tornado damage tracks that occurred throughout the state. SPoRT provided pre- and post-event difference imagery at 250 m spatial resolution from the Moderate Resolution Imaging Spectroradiometer (MODIS) and 15 m false color composites from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER). These surveys helped our NWS partners confirm their ground surveys, but also helped to correct the characteristics of several tracks (Molthan et al. 2011). Many of these products remain available through the SPoRT web page (link) and also through the USGS Earth Explorer portal (link).


The MODIS Band 1 difference image above shows some of the scars left behind by the April 27, 2011 tornado outbreak. Radar snapshots were taken from various times to identify the supercell thunderstorms associated with each track.  Reproduced from Molthan et al. 2011.

Follow-on studies examined the capability of various NASA sensors for detecting and measuring the length and width of scars visible when using the Normalized Difference Vegetation Index, or NDVI, a measurement of vegetation greenness and health commonly derived from multiple satellite imaging platforms.  SPoRT examined NDVI products from MODIS (250 m), Landsat-7 Enhanced Thematic Mapper Plus (ETM+, 30m) and ASTER (15 m) collected in May and June 2011. Possible tornado tracks were identified, mapped, and were then measured to compare against the official NWS damage surveys.  In general, many of the major tornadoes (defined here with maximum intensity EF-3 and greater) were at least partially visible at resolutions of 15-250 m, though weaker tornadoes or those that occurred in complex terrain were more difficult to detect using NDVI and a single snapshot in time. As tornadoes initiated and increased in intensity, or dissipated and decreased in intensity, some of their characteristics became more difficult to detect.  However, some weaker tornadoes were also apparent in Landsat-7 imagery (30 m) in well-vegetated areas.  A summary of the study is available as a publication in the National Weather Association’s Journal of Operational Meteorology. In 2013, SPoRT received support from NASA’s Applied Sciences: Disasters program to partner with the NWS and facilitate the delivery of satellite imagery to their Damage Assessment Toolkit (DAT).  The DAT is used by the NWS to obtain storm survey information while in the field. Satellite imagery from NASA, NOAA, and commercial sensors (acquired in collaboration with USGS and the Hazards Data Distribution System) helps to supplement the survey process by providing an additional perspective of suspected damage areas.

Many of the damage scars apparent from the April 27, 2011 outbreak exhibited signs of recovery and change in the years following the outbreak.  Other tornado events also brought additional vegetation damage and scarring to the region. With five years passing since the 27 April 2011 tornado outbreak, annual views of cloud-free imagery have been obtained from the Landsat missions, operated and managed as a collaboration between the USGS and NASA.  In the viewer linked below, SPoRT has collaborated with the USGS Earth Resources Observation Systems (EROS) Data Center to acquire 30 m true color and vegetation index information from Landsat 5, Landsat 7, and Landsat 8 during the late spring and summer months when local vegetation is at its greenest, allowing the greatest contrast between damaged and undamaged areas. Users can take a look at these images in a web viewer that allows toggling between different products and years, view some of the tornado tracks surveyed by the NWS following the April 27, 2011 event, and zoom into areas of interest to examine how some of the affected areas have evolved over time:

Tuscaloosa, AL


The above animation shows the year before and years after the EF-4 tornado impacted the Tuscaloosa area. The tornado track has seen a significant recovery, but a scar still remains in 2015. In addition to seeing how the landscape as recovered from tornado, development in and around Tuscaloosa is also apparent.  Missing pixels in 2012 are due to an issue with the Landsat-7 imager.

Hackleburg-Phil Campbell


Similar to the Tuscaloosa animation, this animation shows the recovery of the EF-5 tornado that moved through Hackleburg and Phil Campbell, before tracking northeast across the Tennessee River.  Missing pixels in 2012 are due to an issue with the Landsat-7 imager.

SPoRT Disasters Team Provides Post-Storm Imagery to WFO Des Moines, IA Office (DMX)

Following the 10 May 2015 tornado (EF1) near Lake City, IA, SPoRT Disasters team members began to acquire commercial satellite imagery through our partnership with the USGS. Once a request for imagery was made by folks in NWS Central Region, images hosted on USGS’s Hazards Data Distribution System (HDDS) web portal were obtained for processing and distribution to the Des Moines, IA Weather Forecast Office (WFO). Concurrently, teams from the DMX WFO were deployed to assess and survey damaged areas. As survey teams collected information from the field, damage points and polygons were uploaded to the NWS Damage Assessment Toolkit’s (DAT) beta damage editor as early as 12 May 2015.

By 14 May 2015, the Disasters team was able to process and disseminate commercial satellite imagery (in this case SPOT-6 panchromatic) to the DAT application for post-event damage assessment and verification by the office. Ground scarring and swirling (see red inset in the image below) from the tornado were evident as a linear feature moving SW to NE through the image. Once the imagery was delivered, members of the DMX office were able to corroborate their original damage polygon with ground scarring in the imagery. It was in this corroboration that surveyors noted divergence in the two paths, as highlighted by the figure below, and subsequently revised their outputs.


SPOT-6 Panchromatic image (1.5 m resolution) of an EF1 tornado track near Lake City, IA. Image acquired on 12 May 2015.

Kevin Skow, one of the surveyor’s from the DMX office who worked the event, provided the Disasters team positive feedback on the usefulness of this imagery with regards to their work, “…the modifications you are seeing in the DAT are a direct result of what we found in the satellite data. A second storm produced wind damage within a few miles of the actual tornado path that same night. The ground survey team thought that this damage might be from the tornado. However, the satellite data showed that the path was further to the NW. The satellite data also helped us fine tune the path north of Lake City. The satellite data has proven once again to be a great asset for our storm surveying operations.”

Even with some latency, this event demonstrates the importance of providing satellite imagery to forecast offices during (and following) severe weather. While satellite imagery cannot replace the work performed during ground surveys, this type of data has the ability to provide yet another source of information for surveyors to utilize during their response efforts.

Illinois Tornado Track Observed by Landsat-8

On April 9, 2015 a powerful storm system moved through the Midwest, producing numerous severe thunderstorms and reports of damaging wind, large hail, and tornadoes. One of these thunderstorms produced a long-track tornado that moved through north central Illinois. The National Weather Service in Chicago has assigned a preliminary EF-4 rating to this tornado.

Landsat-8 Panchromatic Band at 15 m resolution showing portion of the EF-4 tornado track northwest of Rochelle, Ill.

Landsat-8 Panchromatic Band at 15 m resolution showing a portion of the EF-4 tornado track northwest of Rochelle, Ill. Image acquired on April 10, 2015.

NASA SPoRT has established a collaboration with the USGS, National Weather Service, and NASA Applied Sciences: Disasters Program to provide Earth remote sensing imagery to supplement other data sets available during their storm damage assessments. Imagery is made available to the NOAA/NWS Damage Assessment Toolkit, a geographic information system (GIS) application operated on mobile devices and web browsers, which aids in the collection of photos and other damage indicators collected during a survey. Satellite imagery can be helpful in some cases, helping to identify affected areas where road networks are limited, or there are other access restrictions.  Image sources include moderate spatial resolution (250-375 m) imagery from the NASA MODIS and NASA/NOAA VIIRS sensors, higher resolution (15-30 m) satellite imagery from NASA’s Landsat-7, Landsat-8, EO-1, and Terra ASTER, and fine-scale (1-4 m) imagery from commercial vendors that provide disaster support in collaboration with USGS and other federal agencies.

Landsat-8 True Color image at 30 m resolution showing a portion of the EF-4 tornado track northwest of Rochelle, Ill.

Landsat-8 True Color image at 30 m resolution showing a portion of the EF-4 tornado track northwest of Rochelle, Ill. Image acquired on April 10, 2015.

NASA Next-Generation Satellite Observations of Hurricane Odile

NASA Next-Generation Satellite Observations of Hurricane Odile

A unique weather event is unfolding this week as Hurricane Odile, now a tropical storm, is impacting Baja California Sur, bringing heavy rain and high winds to the region and causing tourists to evacuate resorts. The National Hurricane Center reports that Odile ties Olivia (1967) as the strongest hurricane to make landfall in the satellite era in Baja California Sur**. NASA SPoRT provides specialized satellite products to National Weather Service Forecast Offices as well as National Centers such as the National Hurricane Center to aid forecasting high impact events such as Hurricane Odile.

Below is an example of Passive Microwave RGB imagery created from the NASA Global Precipitation Measurement (GPM) mission as part of The Core Observatory satellite launched on 27 February 2014. The images are in N-AWIPS (National Centers for Environmental Prediction Advanced Weather Interactive Processing System) format and are an example of products available to forecasters at the National Hurricane Center.  Forecasters use the 89 GHz RGB product to look for areas of strong convection which show up as deep red as seen in Fig. 1 which captures Hurricane Odile a few hours before landfall.

89 GHz RGB 0121 UTC 15 September 2014. Areas of deep convection appear red and can be seen surrounding the eye wall and within the rainbands of Odile in this image a few hours before landfall.

Figure 1. GMI 89 GHz RGB 0121 UTC 15 September 2014. Areas of deep convection appear red and can be seen surrounding the eye and within the rainbands of Hurricane Odile in this image a few hours before landfall.

The 37 GHz can additionally be used to distinguish areas of deep cloudiness (light blue) from more active convection (pink) as well as open water (green) or land (cyan).  Note the areas of pink or active convective in Fig. 2 surrounding the eye and within the rainbands.


Figure 2. GMI 37 GHz RGB 0121 UTC 15 September 2014. Areas of active convection appear pink and can be seen surrounding the eye and within the rainbands of Hurricane Odile in this image a few hours before landfall.

Figure 3 and 4 show similar observations from the legacy NASA Tropical Rainfall Measurement Mission (TRMM) as Hurricane Odile made landfall near Cabo San Lucas around 445 UTC 15 September. TRMM is expected to run out of fuel by February 2016 and will no longer be available to collect valuable observations. We are well prepared for a replacement with GPM in orbit and already collecting observations.

TRMM 89 GHz RGB 0307 UTC 15 September 2014

Figure 3. TRMM 89 GHz RGB 0307 UTC 15 September 2014.  Areas of deep convection appear red and can be seen surrounding the eye and within the rainbands of Hurricane Odile in this image a little over one hour before landfall.


Figure 4. TRMM 37 GHz RGB 0307 UTC 15 September 2014.  Areas of active convection appear pink and can be seen surrounding the eye and within the rainbands of Hurricane Odile in this image a little over one hour before landfall.

Additionally the Visible Infrared Imaging Radiometer Suite (VIIRS) Day-Night Band Radiance imagery from the next generation NASA Suomi National Polar-orbiting Partnership (NPP) satellite shows an impressive picture of Hurricane Odile approximately one day before landfall (Fig. 5). Note the city lights that can be seen through the clouds in Fig. 5 as well as lightning within the area of convection in the rainband. This imagery can be used to support disaster response and help emergency managers identify the areas where conditions have caused power outages. Local knowledge of city light patterns can allow users to identify where the most significant power outages are and determine where to begin relief efforts.

VIIRS Day-Night Band Radiance

Figure 5. VIIRS Day-Night Band Radiance 0904 UTC 14 September 2014. City lights and lightning observed approximately one day before Hurricane Odile made landfall.

As the community transitions from legacy instruments such at TRMM and MODIS, NASA SPoRT will continue to develop unique products from Next-Generation missions such as GPM and Suomi NPP to aid National Weather Service Forecast Offices and National Centers in forecasting high impact events such as Hurricane Odile.

**see archived National Hurricane Center forecast discussion at