Backscatter-brained: Using Radar Imagery to Determine Wetland Inundation Patterns

John R. O'Connell, Michael W. Eichholz, Heath M. Hagy

Research output: Contribution to conferenceOtherpeer-review

Abstract

Monitoring wetland inundation is critical to the management and conservation of waterfowl. Different inundation regimes impact the biogeochemical processes and thus productivity of wetlands. Inundation, or lack thereof, can have an even more direct impact on waterfowl, waterbirds and shorebirds; if a wetland is not inundated during the period when one of those functional groups are in the area, then it is not a vailable to that group. Relying on the National Wetlands Inventory (NWI) alone and lacking the data necessary to consider wetland availability during biologically important periods, managers may substantially underestimate the amount of wetlands needed to support waterfowl, waterbirds and shorebirds. Determining the timing and extent of Illinois wetland inundation is an important step towards making informed decisions in wetland management in the state. Ground surveys are labor intensive and optical remote sensing techniques struggle to accurately resolve wetland inundation where there is heavy vegetative cover. Synthetic aperture radar (SAR) has been used to detect inundation in multiple wetland cover types and may be an efficient method of recording inundation patterns in Illinois. A pilot study testing the use of SAR imagery to these ends was conducted on imagery of the Cache River watershed prior to the commencement of a state-wide survey in order to assess feasibility and identify potential pitfalls. Archival PALSAR L-band imagery was attained for four seasons over a one year period and was classified without ground survey to determine inundated wetlands. The results were compared to NWI and were compared among seasons. Inundated wetlands were detected to cover a much smaller area than the total NWI polygon area (46%), but seasonally matched the hydrologic record of the Cache River during the same time frame (R2 = 0.83). Only a small portion of detected inundation persisted throughout spring, summer and fall (17%), highlighting the ephemeral nature of inundation in the watershed during the period evaluated. Further analysis of SAR imagery with in situ validation is necessary to determine the accuracy of the technique. Beginning in spring of 2015, ground surveys crews have been mapping the extent of wetland inundation in ~90 wetland plots throughout Illinois. The ~50ha plots cover multiple cover types including forested, scrub/shrub, emergent and open water wetlands. Surveys are timed to coincide with the spring waterfowl migration, the summer waterbird nesting season and the autumn shorebird migration. The resulting thematic maps will be used to train and validate a random forest classification model of inundation from newly -acquired PALSAR-2 L-band imagery. Several image resolutions will be compared to determine the most cost-efficient method before the model is expanded to cover the entire state. Successful implementation of the model will allow managers to monitor inundation in the state’s wetland at a level that was previously unattainable.
Original languageEnglish (US)
StatePublished - 2016

Keywords

  • INHS

Fingerprint

Dive into the research topics of 'Backscatter-brained: Using Radar Imagery to Determine Wetland Inundation Patterns'. Together they form a unique fingerprint.

Cite this