The NOAA Coastal Services Center (CSC) requested the creation of benthic habitat data along the southern Texas coast to support the Texas Seagrass Monitoring Program. The benthic habitat map was created from 1m ADS40 digital airborne imagery collected along the Texas coast during 2004 for the National Agriculture Imagery Program. The original raw images were reprocessed into 3-band true color and color-IR orthos. The benthic habitat map was created from resampled 2m mosaicked orthos. Habitat classification was performed through segmentation of the imagery using Definiens Professional and habitat labeling through Classification and Regression Tree (CART) Analysis. The minimum mapping unit is 100m2. This map covers Lower Laguna Madre, which is approximately 800mi2.
The NOAA CSC worked cooperatively with the Texas Parks and Wildlife Department (TPWD) and the Texas A&M University, Corpus Christi, Center for Coastal Studies to have benthic habitat data, primarily submerged aquatic vegetation, created for eight bay systems along the southern Texas coast. This project, the phase 2 project, consists of the two remaining bay systems, Lower Laguna Madre and San Antonio/Espiritu Santo Bays. This benthic habitat data will support the recently adopted Texas Seagrass Monitoring Program which calls for the monitoring of seagrass beds along the Texas coast for assessment of status and trends.
The geographic extent of the project area is ~800mi2. Benthic habitat data was generated from 2004 NAIP imagery for all estuarine lands data was required for the marine side of the barrier beaches.
Publication Date
None
7320 Executive Way
NOAA Coastal Services Center
Horizontal accuracy of the reprocessed source imagery was verified to be better than 5 meters at 90% confidence level in accordance with National Map Accuracy Standards for a 1-meter GSD. The accuracy of the contractor's final map is 89% and the accuracy of the final map is 90%. Overall final map deterministic accuracy is 88% (with fuzzy accuracy at 90%) which exceeds the contract standard of 85%. Additionally, 14 of the 16 final map class producer's and consumer's accuracies are at 85% which exceeds the contract standard of 80%. However, the deterministic producer's accuracy for emergent marsh is 77%, and the user's accuracy for unconsolidated sediments is 72%.
For the reprocessed imagery, compliance with the accuracy standard was ensured by the placement of photo identifiable ground control points. A total of 18 photo identifiable ground survey points was used for the calculations. An RMS value was calculated based on the imagery reprocessed for this project by comparing the aerotriangulated X and Y coordinates. This value represents an estimate of the accuracy of the horizontal coordinate measurements in the tile expressed in meters. For the final map product Initial Map accuracy assessment was used as a tool to prioritize areas for further field examination and after field investigation to prioritize those areas where additional modeling or interpretation was needed. Error matrices showing both deterministic and fuzzy accuracies were compiled for the initial map. Based on the results compiled from the assessment, the team visit any classes exhibiting inaccuracy and addressed the classes through modeling, additional analysis or manual editing.
Compliance with the accuracy standard for the reprocessed imagery was ensured by the placement of photo identifiable ground control points and the collection of airborne GPS data. Compliance with the accuracy standard for the final map product was ensured by field checks and manual editing.
Accuracy assessment determined by evaluating the horizontal accuracy obtained during the aerotriangulation process for each lift for the reprocessed imagery and by field verification for the completed map product.
TerraSurv, Inc. of Pittsburgh, PA was contracted by EarthData International of Frederick, MD to perform a geodetic control survey in support of mapping an area along the southeasterly coast of Texas between Port Lavaca and Brownsville. Thirty-eight photo identifiable locations were surveyed to provide ground control and quality assurance checks for the mapping. Twenty of the stations were used for mapping control and eighteen of the stations were used for quality checks. The horizontal datum was the North American Datum of 1983, CORS adjustment (NAD 1983 CORS). The vertical datum was the North American Vertical Datum of 1988 (NAVD 1988).
The digital orthophotography was developed from imagery acquired as part of the 2004 overflight of the State of Texas developed for the USDA National Agricultural Imagery Program (NAIP). In order to achieve a horizontal accuracy of 5 meters, CE90 it is necessary to reprocess the imagery incorporating new GPS field control. It should be noted that the imagery was not tide coordinated so tidal variation may exist between sorties. The imagery was acquired between November 3, 2004 and November 7, 2004.
The original 1m DOQQs for the project area were resampled to 2m and mosaicked. For habitat classification, the mosaicked imagery was divided into two processing areas for Lower Laguna Madre, resulting in a total of four 2m mosaicked orthos: one set of two mosaics for true color and one set of two mosaics for color-IR. Lower Laguna Madre was divided into two processing areas. Image segmentation was performed in Definiens Professional using the blue, green, red, and near-infrared bands for each of the processing areas. The classification of the habitat segments (as ESRI polygon shapefiles) was performed using CART analysis. The habitat maps for each area was refined with the aid of field data collected during July, August, and October of 2007. The two processing area shapefiles were edgematched to one another and clipped to the final project area boundary. Adjacent areas do not overlap, and each polygon, within and across all processing areas and bay systems, has a unique polygon identification number. Each shapefile was checked for proper topology and to insure that each polygon has a correct habitat label, habitat code, modifier label if present, unique identification number, and an area calculation. Polygons below the 100m2 minimum mapping unit (MMU) were eliminated, though some polygons <100m2 were retained if their area changed to below the MMU due to the polygon boundary smoothing process. The habitat data also went through a NOAA independent validation review. Polygons in the habitat map labeled as Patchy SRV (seagrass) were used to mask the 2m image mosaics for further classification of these areas. Pixels in the imagery falling within the Patchy SRV polygons were classified into a "percent seagrass" cover category. For each Patchy SRV polygon from the habitat map, the average percent seagrass coverage was calculated based on the coverage values of each pixel within the polygon. Accuracy assessment was performed on seven classes with Patchy SRV and Continuous SRV being combined into a single accuracy class. For field data collection, non-random sites in the form of polygons were chosen by analysts with an attempt to sample all available image signatures. These sites were visited in the field and data on each site was collected directly into digital format (ESRI shapefile) using a laptop or onto a paper form that was later entered into digital format. Sites were navigated to primarily using a Garmin GPS 76 unit connected to a Panasonic Toughbook laptop displaying the project imagery and polygons in ArcMap v9.1 or using the GPS unit alone. Habitat classification was estimated as accurately as possible using different methods or combination of methods which included above water observation, snorkeling, wading, and underwater video. This data was entered into an ESRI shapefile via a digital field form in ArcMap specifically developed for this type of field data collection. More sample polygon sites were collected in-office based on the in-field collected sites in order to meet the 30 sites per class accuracy assessment requirement. For each class, a random selector macro in ArcMap was used to randomly select 30 sites for accuracy assessment. The entire pool of accuracy sites was kept separate from the remaining sites and only used for accuracy assessment during the project. Anonymity of the accuracy sites was maintained throughout the project because it was unnecessary to ever visually review these sites in order to perform the accuracy analysis. More accuracy assessment sites were collected in a later field collection trip to add to the analysis. These sites were chosen by randomly selecting polygons within specific regions that were pre-determined to be visited. Information for these sites was collected using the same methods for the other sites. Accuracy information was compiled using ArcMap. The Zonal Statistics tool in ArcToolbox was used to collect accuracy information from the habitat map using the accuracy polygons. The "mean" statistic was used to determine the map value for each accuracy polygon. An accuracy assessment error matrix was generated using this information by importing it to Microsoft Excel and building the matrix. Both deterministic and fuzzy accuracy assessment were performed. The accuracy analysis and error matrices are presented and discussed in the project final report entitled Coastal Bend of Texas Benthic Habitat Mapping Phase 2 Final Report.
Not applicable
2234 South Hobson Avenue
Users must assume responsibility to determine the usability of these data.
2234 S Hobson Ave.
