200-6 Delineating Dambo Landscape Elements Using Aerial Gamma-Ray Data As the Sole Predictor.

Poster Number 1111

See more from this Division: ASA Section: Climatology & Modeling
See more from this Session: General Airborne and Satellite Remote Sensing: II (includes graduate student competition)

Tuesday, November 5, 2013
Tampa Convention Center, East Exhibit Hall

Jerome Sebadduka, School of Environment, Washington State University, Pullman, WA
Abstract:
Dambos are seasonally saturated grassy valleys that characterize the central African plateau. Seasonality of precipitation and topography interact to influence pedogenesis and plant distribution patterns, so that in a dambo landscape, soil and floral characteristics vary along a given cross-profile. It is on this basis that four dambo landscape elements or classes – bottom, floor, margin and upland – are usually distinguishable. Being able to determine and delineate these landscape elements is beneficial to surveys of soils in dambo terminated catenae. Hansen et al. (2009) demonstrated this when they used a suite of SPOT-4 imagery, terrain and field data, to disaggregate a dambo landscape in central Uganda. However, in tropical Africa, use of multispectral optical imagery is limited by costs and the generally cloudy conditions which affect the quality of sensor data. The alternative is to use the freely available aerial gamma-ray data. We assessed the suitability of this data by mapping a subset of a catchment in central Uganda, which was originally studied by Hansen et al. (2009). A multinomial regression approach was used, where dambo landscape elements were considered to be the nominal response variable, while predictors included the signals and derivatives of gamma activity. The model with %K and eU as the only predictors was found to be the most suitable. This model was implemented in ENVI 5.0 to create dambo class-based probability surfaces. Using the maximum likelihood classifier (MLC) in ENVI 5.0, the probability surfaces were classified to generate a landscape map of the study area. The probability that dambo floors and margins are predictable based on only gamma-ray activity was noted to be low (< 0.7), owing to the marginal difference in %K activity between both classes and the bottom. On the other hand, the level of producer accuracy attained for bottoms and uplands – 55.3% and 64.6%, respectively – is largely due to the high contrast in %K and eU gamma activity between the two classes. This is so because gamma energy recorded over bottoms is generally dimmed by the ever present vegetation, which is not the case with the uplands.

See more from this Division: ASA Section: Climatology & Modeling
See more from this Session: General Airborne and Satellite Remote Sensing: II (includes graduate student competition)