Use of a hybrid optical remote sensing classification technique for seasonal wetland habitat degradation assessment resulting from adoption of real-time salinity management practices

Abstract

Delayed seasonal wetland drawdown is being investigated as one means of improving the scheduling of saline drainage discharge to the San Joaquin River, California, as an application of the principal of basin-scale real-time salinity management. A hybrid rapid vegetation assessment methodology was derived from the California Native Plant Society Rapid Assessment Protocol for classifying the vegetation in these seasonal wetlands. A hybrid remote sensing methodology combining pixel- and object-based components was developed to apply this classification strategy to a 160 km2 region. Twenty-six different plant communities were represented in a total of 20 land cover classes. An overall mapping accuracy ranges from 60% for identification of all 26 plant communities to 100% for identification of a single plant species. Low representation of certain associations for sensor calibration resulted in lower than anticipated mapping success as measured by errors of omission and commission. This image processing methodology provides an important tool, in concert with soil salinity mapping and wetland biology surveys, to assess the long-term impact of adaptive management strategies such as real-time salinity management on the wetland resource.