Cabezudo Vázquez, Onix H.

Profile Picture

Filters

20221
Reset filters

Settings

Citations

Publication Search Results

Now showing 1 - 1 of 1
  • Thumbnail Image
    Publication
    Classification of coastal lowland acid sulfate soils using digital soil mapping techniques for Puerto Rico and U.S. Virgin Islands
    (2022-12-13) Cabezudo Vázquez, Onix H.; Pérez Alegría, Luis R.; College of Agricultural Sciences; Martínez Rodríguez, Gustavo A.; Macchiavelli, Raúl E.; Department of Crops and Agro-Environmental Sciences; Parés Matos, Elsie I.
    Oxidation of soil pyrite due to hydrologic disturbances is responsible for severe soil acidification and the production of noxious greenhouse gases. Coastal lowland acid sulfate soils (CLASS) are an environmental hazard worldwide due to the associated high acidity (pH < 4), and threats posed by runoff discharge. We use digital soil mapping (DSM) techniques to device a more effective, and least time- and resource-consuming alternative than traditional field sampling schemes to determine the probability of occurrence of CLASS in Puerto Rico and U.S. Virgin Islands. Soil samples collected from coastal areas in Puerto Rico were classified based on anaerobic incubated pH results after 7 weeks. Out of 58 samples at a depth of 0-12 in, 20 reached pHfinal < 5 after incubation, and hence were labeled as positive for CLASS. Sulfur content was analyzed in positive samples resulting in a 0.61 inverse correlation. Two setups of predictive models, Model A and Model B, were considered using DSM techniques and a Random Forest (RF) hyper-tuning parameters classifier. Sentinel 2 sensor indices and SCORPAN model data were used as covariates for the region of interest, established by Holocene correlated data with sulfur deposition. Model A resulted best with an out-of-bag error of 19%, test error of 23%, and a spatial prediction for CLASS of 104.18 km2 in Puerto Rico, with a mean probability of 65%. Also, for the U.S. Virgin Islands, model A resulted with an out of bag error of 19%, test error of 75%, and a spatial prediction for CLASS of 0.82 km2 in U.S. Virgin Islands with a mean probability of 59%. Resulting CLASS mapping shows the first assessment of this technique in the Caribbean region.