INTERIM TECHNICAL REPORT

1999 Legacy Project
Project Number: 991872

Project Title:
Monitoring Bird Migrations and Movements with Radar and Landsat Imagery-II

Service/Sponsor:
DoD

CMD Division:
DoD

Report Submitted by

Sdiney A. Gauthreaux, Jr.
Professor
Department of Biological Sciences
Clemson University
Clemson, South Carolina 29634-0326

15 October 1999

INTRODUCTION

With the Partners in Flight initiative continuing to develop sound conservation strategies for migratory birds that breed in North America and winter south of the border of the United States, the need for information on important migration stopover areas is critical. Conservation organizations (e.g., National Audubon Society, Nature Conservancy) rank protection of important migration stopover areas in the United States as a very high priority. It is possible to acquire some of this information from the records of field observers (birdwatchers or birders), but this approach might overlook some important areas that are undiscovered by birders. We have recently demonstrated (Gauthreaux and Belser 1998) that imagery from new WSR-88D doppler radar (or NEXRAD for NEXt generation RADar) operated by the National Weather Service and the Department of Defense can be used to map important migration stopover areas.

Doppler weather radar can readily detect departures of migrants from stopover areas out to a range of 60 nautical miles (111 kilometers). For a brief period of time when migrants are climbing into the atmosphere and radar can begin to detect them, the echoes produced by the masses of migrants accurately reflect the stopover areas from which they have departed. The goal of this project is to define the shape and location of migration stopover areas based on radar returns and examine how these areas correspond to landcover features. of When the radar data on migration stopover hotspots are joined to Landsat and Thematic Mapper satellite data on habitat using a Geographical Information System (GIS), it is possible to determine the types of habitat used by migrants during stopover. This information is critically important to land managers that wish to preserve and increase populations of migratory birds.

PROGRESS TO DATE

The development of a self-sufficient Geographical Information System (GIS) laboratory within the Radar Ornithology Laboratory at Clemson University is complete. We have also acquired the services of Donald Van Blaricom, a GIS analyst, affiliated with the Forest Resources Systems Institute in Clemson, South Carolina. He is responsible for the development GIS layers of radar data and satellite data and subsequent spatial analyses.

To analyze migration hotspots it is necessary to have station base reflectivity files with 1 km x 1° pixel resolution. National mosaic image files that have been re-sampled to 2 km x 2 km or 4 km x 4 km pixel resolution are not usable. We have collected usable station data for several Gulf Coast stations since 1995, and these are being used in this project. This spring (1999) we began gathering the data necessary for migration stopover analysis at 50 stations in southeastern United States and this fall (1999) we expanded data gathering to all 140 WSR-88D stations in the United States.

The first objective of the radar/satellite project was to put level three WSR-88D data showing migration stopover areas (Figure 1) into a GIS layer. Such files are referred to as NIDS because they are downloaded from a NEXRAD Information Dissemination Service). Before this step could be accomplished we had to first parse the header and structure of the NIDS files containing base reflectivity and base velocity information. Alan Cunningham and Donald Van Blaricom accomplished this with assistance from Mark Logan of Unisys Corporation and Michael Wolfinbarger of Oklahoma Climatological Survey. Because NIDS files have no intrinsic projection, OCS suggested that we use polar stereographic with the latitude/longitude of the radar as the origin. Azimuths of 0° and 360° point to north in NIDS files.

Figure 1. A base reflectivity image from the WSR-88D near Houston, TX on 30 April 1994 showing the exodus of migrants from stopover areas at 01:36 UTC. The green patches are echoes produced by concentrations of migrating birds as they begin a night's migration. Note the correspondence of stopover areas to riverine floodplain areas. The yellow strobe is interference from another transmitter.

Because our protocol since 1995 was to download images every hour, we reviewed images closest to the beginning time of exodus-approximately 45-60 minutes after sunset. For an image to qualify for stopover area analysis it had to satisfy the following criteria:

1) an exodus of migrants within 60 nautical miles of radar station was occurring,
2) an input of trans-Gulf migration within last two days,
3) no daytime migration continuing into the evening and obscuring exodus events,
4) no unusual atmospherics causing radar coverage to show strobing,
5) no precipitation or smoke within 60 nautical miles that could obscure an exodus,
6) verification that the echoes were bird and not insect movements.

We compared simultaneous vertical wind profile and base velocity products from the WSR-88D with winds aloft from radiosonde to determine if an exodus in the base reflectivity image involved birds or insects. When echo velocities are close to following wind velocities, the echoes are likely insects, and when echo velocities are considerably greater than wind velocity, the echoes are likely birds. Birds show good velocities against the wind, but insect rarely do so.

After a thorough examination of the NIDS file structure, Don Van Blaricom was successful in importing the raw NIDS files into ArcInfo for further analysis. The working data sets in ArcInfo consisted of 29 files from KHGX (the Houston, TX WSR-88D) and 18 files from KLCH (the Lake Charles, LA WSR-88D) that met the inclusion criteria for the spring of 1997. In ArcInfo we set all reflectors above 30 dBZ to 0dBZ, because to date we have never recorded bird movements in excess of 30 dBZ.

In ArcInfo we clipped all data beyond 60 nautical miles from the radar (111 km) and also beyond 4 km of the coastline to eliminate targets over water. We also prefiltered pixels by dBZ bins. We first removed all pixels below 5 dBZ, then removed all pixels below 10 dBZ, and then removed all pixels below 15 dBZ. This procedure helped reduce background reflectors and enhanced the stopover areas. The dBZ values were then rescaled to Z values where Z= antilog (dBZ/10) (see Gauthreaux and Belser in press) for additional details of converting Z values (reflectivity) to birds km^{3 -1}. The resulting statistical coverages consisted of mean Z values for each pixel, the variance in Z values for each pixel, and the total count of instances when each pixel was above the dBZ filter.

Figures 2 and 3 show the distribution of migration stopover hotspots based on 29 images from the KHGX (Houston) WSR-88D and 18 images from the KLCH (Lake Charles) WSR-88D from the spring of 1997. The data displayed are "count" data that have been prefiltered by 10 dBZ. The white areas represent the stopover areas within 60 nautical miles of the radar. We are currently processing similar data for the spring seasons of 1995, 1996, 1998, and 1999. Within a couple of months we will have completed our analysis for Houston, TX and Lake Charles, LA, and then we will extend our analysis to include data from other radar stations along the Gulf coast. Our work has already stimulated similar analyses for the Great Lakes Region.

In an effort to match stopover areas with landcover types, we have been acquiring satellite landcover data. We began our analysis by using the USGS GIRAS data set generated from 1978-1980 Thematic Mapper satellite imagery. This imagery has 30 m resolution so the shapes of forested areas can be seen clearly. Figures 4 and 5 show landcover types of interest for the Houston, TX and Lake Charles, LA areas. The blue-green is forested wetlands, the dark green is floodplain forest, the middle green is brushland, and the light green is pine forest. We are also in the process of getting more recent raw Thematic Mapper satellite data. The images must be cloud free and we are searching metadata to select appropriate data sets. In addition to purchasing imagery, we are consulting with investigators from states engaged in GAP analyses to see if they can share satellite imagery data in return for information on migration stopover areas. Figure 6 and 7 show Thematic Mapper landcover with migration stopover areas superimposed.

Figure 2. Migration stopover areas within 60 nautical miles of the Houston WSR-88D. The white areas indicate locations where the greatest number of exoduses above 10dBZ occurred in the spring of 1997.

Figure 3. Migration stopover areas within 60 nautical miles of the Lake Charles, LA WSR-88D. The white areas indicate locations where the greatest number of exoduses above 10dBZ occurred in the spring of 1997.

Figure 4. Thematic Mapper satellite imagery showing forested habitat in the Houston, TX area.

Figure 5. Thematic Mapper satellite imagery showing forested habitat in the Lake Charles, LA area.

Figure 6. Thematic Mapper satellite imagery showing forested landcover with stopover areas superimposed in red for the Houston, TX area.

Figure 7. Thematic Mapper satellite imagery showing forested landcover with stopover areas superimposed in red for the Lake Charles, LA area.

Although there is general agreement between the location of migration stopover hotspots based on radar and forested habitat, the agreement is far from perfect. There are several reasons for this.

The radar images we have used thus far in our analyses were taken once an hour and some of the images were beyond the time of optimum exodus. This causes a considerable amount of smearing and fuzziness because the concentrations of birds have moved well beyond the departure areas and the birds are less concentrated as they move farther away. With imagery closer to the peak time of exodus, we can eliminate this source of variance.

It is extremely hard to avoid occasions when trans-Gulf migrants are not arriving on the northern Gulf coast in spring during the time of a major exodus. On most occasions when trans-Gulf flights are arriving, the flights continue in some measure after dark. One can still detect areas of exodus through the mass of arriving migrants, but the results are less clear cut. This will not be a problem in areas where daytime migration terminates before nightfall.

Another source of error is related to the fact that the radar beam increases in altitude as a function of distance. As birds climb they are moving away from the stopover area, and departing migrants at 50-60 nautical miles from the radar station may not enter the radar beam until they are some distance from their departure areas. Consequently stopover areas at greater ranges based on radar imagery will be displaced geographically in the direction of flight. This also means that stopover areas based on radar imagery will be less distinct and perhaps less frequently detected because of the height of the radar beam. We are currently examining ways of correcting for the displacement and detection rate of distant stopover areas. Our 60 nautical mile limit was imposed because beyond that range we rarely detect departures unless atmospheric inversions cause the radar beam to bend downward and stay closer to the earth's surface.

CONCLUSION

Our progress to date is very encouraging and we are eager to complete analysis of the five years of spring data sets for the northern Gulf coast. We will also be looking at the fall data sets, because preliminary analysis shows that the same migration stopover areas may be equally important during the fall migration period when birds are moving south. We feel that this approach to mapping migration stopover areas or migration "hotspots" has great potential, and the technique will produce valuable data that land managers can use not only to protect existing stopover areas, but to restore degraded stopover areas and possibly construct new ones.

REFERENCES

Gauthreaux, S. A., Jr. and C. G. Belser. 1998. The use of weather surveillance radar to map important migration stopover areas. Paper presented at the North American Ornithological Conference in St. Louis, 6-12 April 1998.

Gauthreaux, S. A., Jr. and C. G. Belser. 1999. Reply to "Comments on 'Display of bird movements on the WSR-88D: Patterns and quantification'".  Weather and Forecasting 14:1041-1042.

Go to Radar Ornithology Introduction

Department of Defense Partners in Flight Program Email:DoD PIF Program Manager