Habitat Suitability Index for Alligator Gar Spawning within the Lower Mississippi River Corridor

A landscape level AG spawning suitability data layer was developed to screen for landscape level features indicating locations that may be suitable for AG spawning within the lower Mississippi river corridor. Suitability was determined using a combination of inundation frequency, land cover, and thermal characteristics derived from related landscape level analyses (see below). The result for each evaluation was coded to display the predicted suitability for AG spawning . Habitat suitablility was informed by an alligator gar telemetry project on the St. Catherine Creek National Wildlife Refuge south of Natchez, MS. Approximately 20 fish were tagged in 2010, 2012 and 2013 and movement patterns on the floodplain were tracked using an array of fixed station Vemco receivers. Imagery from six Landsat scenes (Path 23; Rows 34-39; n=133 images total) were used to determine inundation frequency and average thermal properties for locations along the lower Mississippi River corridor that are still subject to direct influence by the River. The lower Mississippi River corridor was defined using USGS watershed boundaries (HUC 12), federal levee boundaries, elevation and obvious barriers to flow based on inundation frequency results. Inundation Frequency: Ideal inundation frequency was defined to fall in an intermediate range. Thresholds for each Landsat scene were established based on comparison of the distribution of river levels at a key gage location associated with each scene. The distribution of river levels on all dates was compared with the distribution of river levels on the dates associated with each landsat image. Appropriate thresholds were selected that excludes locations high in the floodplain that are inundated less than 50% of the time during spawning season of April, May and June. A lower threshold was also established tthat ranks the most persistently inundated locations low - these locations may be more indicative of deep water refugia. All locations within the mainstem of the Mississippi River were also ranked low because flow conditions in the river were considered to be unsuitable for successful spawning. Land Cover:The location was evaluated for suitable habitat structure. The 2012 Cropland data layer was used as a basis for habitat coding because it offered: the most current landcover evaluation, details about low cover croplands that may be important in defining spawning habitat suitability, and also includes standard national land cover categories. Each habitat type was assigned either a one or zero based on results from telemetry which indicated that AG are most frequently found spawning in open fields with low brushy cover and were not frequently found using forested areas with heavy canopy cover. For example, active croplands, pasturelands, fallow croplands, and herbaceous wetlands were all designated as sutiable spawning habitat ; woody wetlands and deciduous forest were designated as unsuitable. Thermal characteristics: Telemetry observations also showed that AG were found spawning in locations that, on average, had the highest temperature difference from the Mississippi River. For each Landsat scene, thresholds were established to categorize locations as poor, acceptable, and good based on their average difference from the mainstem Mississippi River river temperature. Details of the inundation frequency and thermal analysis are below. Climate Data Records (CDR) from Landsat 5 and Landsat 7 imagery was downloaded from USGS (http://earthexplorer.usgs.gov/). CDR products reflect the most current georectified, atmospherically corrected and thermally calibrated surface reflectance products available. Landsat 5 and 7 sensors have acquired images of the Earth nearly continuously since 1984, with a 16-day repeat cycle. Under ideal conditions, satellite imagery can give a nearly instantaneous whole-basin snapshot of water distribution. Cloud-free images captured from December-March were considered for use in the inundation frequency analysis to minimize the obscuring effects of tree canopy and floating vegetation on water classification. These images were generally captured under a wide variety of river levels. Each scene was associated with a key gaging station on the Mississippi River and each image was associated with a Mississippi River stage at the time of image capture. For path/rows 23/36, 23/37, 23/38 and 23/39 the distribution of Mississippi River stages observed on all imagery dates was similar to the distribution of all observed river stages from 1984-2011. For path/rows 23/34 and 23/35 the distribution of river levels observed on all dates was somewhat lower when compared with the distrbution of river levels observed on dates having associated Landsat imagery. For each image, simple thresholding of Landsat band 5 (1.55 - 1.75 mm) was used to classify categories of dry land, flooded land and open water. This technique worked well for the focus habitats of this study - floodplain areas that are dominated by deciduous woody vegetation - but misclassified shadows and dense pine forests which also have very low band 5 reflectivity. For this reason, NDVI was also calculated for each image and pixels having NDVI > 0.42 were recoded as dry land. For the current data layer, categories of flooded land and open water were collapsed to dry vs. wet. To determine the potential thermal refuge offered in inundated floodplain areas, thermal habitat was characterized using the same images and the associated thermal band from Landsat. Landsat thermal band results from 2012 were within 0.5 degrees compared with in situYSI measurements in Butler Lake collected at the same time. For each historical image, average mainstem river temperature was determined by averaging the temperature in a zone that is inundated in 95% of all images. Temperature for each inundated pixel in each image was then compared with the average mainstem river temperature for that image. The resultant "difference from mainstem temperature" layers were then averaged over all images to define the "average thermal difference from river temperature". The resulting composite was then masked to show only pixels that were inundated at least 50% of the time during spring.