|Abstract or Summary
- Denitrification, the biological or chemical reduction of ionic nitrogen oxide or dinitrogen, has not been widely studied in forest ecosystems despite widespread interest in other facets of the forest nitrogen economy. This study had three main objectives: to determine whether potential for denitrification exists in forest riparian and hill slope soils in the western Oregon Cascade Mountains; to test the hypothesis that denitrification potential within riparian forests varies with slope position, soil depth, and season; and to contrast denitrification potential in old-growth Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco.) and secondary-successional red alder (Alnus rubra Bong.) forests. Denitrification potential was assessed by anaerobically incubating freshly-collected soils in the presence of acetylene and measuring nitrous oxide emission with electron capture gas chromatography. Soils were amended with deionized water, but not with nitrate or carbon sources. Soil moisture, pH, nitrate, ammonium, and total nitrogen, carbon, and phosphorus were measured. Soils were collected during May-June 1984 and August-September 1983 and 1984. After 1 h of incubation, average N-losses from the upper mineral soil (0-15cm) ranged from 0-11.3 ng N g-1 h-1 for the Douglas-fir soils, and from 3.4-24.4 ng N g-1 h-1 for the red alder soils. N-loss was significantly less (p < 0.001) from the lower mineral soil (15-30cm); averages ranged from 0-0.4 ng N g-1 h-1 for the Douglas-fir soils, and from 1.1-16.0 ng N g-1 h-1 for the red alder soils. Flood plain soils usually lost more N (3.6-24.4 ng N g-1 h-1) than did hill slope soils (0-2.3 ng N g-1 h-1); intermediate N losses were measured in toe slope soils (0.8-4.2 ng N g-1 h-1). The effect of slope position was significant (p < 0.001) only in 1983. Seasonal variation was not significant. Multiple regression of N-loss on soil nitrate, total N, and pH accounted for 61% of the N-loss rate variation. Denitrification potential was highly variable, with coefficients of variation ranging between 53 and 281%, and appeared to fit a lognormal distribution.