Sources of Carbon Dioxide on the Forest Floor of Mt. Kinabalu

Global warming will potentially increase rates of natural CO₂-emitting processes, such as soil carbon respiration (R_s) and litter decomposition, and consequently, these have become the focus of much research in the neotropics and elsewhere. However, studies in the Indomalayan region remain scarce. Tropical montane forests (TMFs) provide large temperature gradients over relatively small geographical areas which facilitate the study of temperature effects on such complex ecological processes. Relatively aseasonal climate, high rainfall and similar site histories within a continuous TMF system reduce the effects of confounding factors that often convolute the interpretation of results from latitudinal gradient studies. I conducted a 1.5-yr long litter bag study (using 2 local leaf species and a common lowland secondary forest species) and a series of three soil respiration studies along a 1.8 km elevation gradient (1614 – 3412 m a.s.l.) on Mt. Kinabalu, Sabah, Malaysia to ascertain the main drivers of litter decomposition and soil respiration, and to determine how they are influenced by changes in temperature. I also determined the temperature sensitivities, Q₁₀, of the two processes. Four 20 x 20 m plots were chosen at 1614 (lower montane forest), 2127 (upper montane forest), 3106, and 3412 (subalpine forests) m a.s.l.. The three SCR studies included: 1. Baseline on-site assessment of soil respiration and potential diurnal effects (Baseline Study); 2. In situ soil translocation study (In Situ Study); 3. Ex situ laboratory soil translocation study (Lab Study). A meta-analysis of tropical litter bag studies involving 68 publications (619 observations) was also determined to determine the main factors driving leaf litter decomposition in tropical regions.
   
   Lignin content was the best predictor of leaf decomposition rates in the litter bag study, explaining 41.9% of the variation in decomposition rates, exceeding soil temperature. However, when the lowland, secondary forest leaf litter species with low lignin content, Macaranga tanarius, was excluded, the effect of temperature surpassed lignin content by 3-fold. Soil temperature explained 99% of the variation in mean decomposition rate via an exponential growth function. Macroinvertebrate feeding did not significantly affect leaf litter decomposition rates (F_{1, 357} = 4.1, p = 0.045; α_adj = 0.05/2³). Mean Q₁₀ of litter decomposition was found to be 6.88, higher than reported in other studies (usually 2-3), and increased in the species order of M. tanarius, S. houttuynii, and X. montanum. Over time, the temperature sensitivity of litter decomposition, Q₁₀, increased as litter quality, B, decreased. Ln B significantly explained 78% of the variation in Ln Q₁₀, in line with the temperature-quality hypothesis.
   
   The Baseline Study measured natural soil respiration levels diurnally in situ and revealed the lowest R_s rates in the warmest site, owing to poorly drained soils with a thin soil organic layer that were subject to flooding. Daytime R_s was higher only in the two subalpine sites. The In Situ Study was fully reciprocal, removing soil from each site to be incubated in the four forest plots. The Lab Study was the same, but performed in a controlled environment where only incubation temperature was altered to mirror the native temperatures of the study plots. Temperature explained 47.7% of the variation in R_s in the Lab Study, compared to only 32.3 % in the In Situ Study, and I attribute the difference to the removal of the influence of site factors other than average temperature in the laboratory study, since soil depths and the incubation setups were identical. SOC content was not a good predictor of R_s, but percentage soil macroaggregate (w/w; >250 µm), a, was found to be an increasingly effective indicator of intrinsic SOM quality for soil respiration as sources of variation was progressively removed from the Baseline, to the In Situ and Lab Studies (R² increased from 0.22-0.99). In the In Situ Study where soil was exposed to rainfall, a positive trend was detected for Q₁₀ vs a, which we attribute to enhanced rates of soil respiration at warmer sites when macroaggregates are being routinely broken down and reformed through heavy rain events. Q₁₀ values were higher in the In Situ study (3.15-5.39) compared to the Lab study (1.84-3.25), but no effects due to elevation were detected. The discrepancy between the Lab and In Situ study results highlight the urgency of soil respiration studies performed in situ to capture more realistic responses of organic matter to global warming under natural conditions.
   
   In the meta-analysis, mean annual temperature (MAT), mean annual precipitation (MAP), and leaf litter nitrogen concentration were found to be important (w+ > 0.5) in two out of three of the data subsets, making them the most important predictors of leaf litter decomposition rates across the dataset. MAT reliably predicted tropical litter decomposition rates in the range of 10-20°C, as similarly found in the litter bag study in Mt. Kinabalu. Nitrogen scored the highest R² values (= 0.74) overall, followed by latitude. Latitude was a good predictor of decomposition rates, but is prone to deviate from the regression model when considering tropical montane sites and may represent an artefact of the strong positive correlation between latitude and MAT that is present in our dataset (r = 0.72). Further, we found the acid-unhydrolyzed fraction (AUF):N to be a reliable predictor of litter decomposition rates in the AUF:N range of 40-100. According to the AIC model selection procedure, the candidate model containing latitude, nitrogen, MAP, and AUF represents the best model permutation in the complete data subset containing all tested variables (R² = 0.83). However, we are inclined to replace latitude with MAT, which results in the third-ranked model in the Complete data subset (ie. Nitrogen, MAP, AUF, MAT; R² = 0.80) due to the direct and well-documented ecophysiological effect of temperature on organic matter decomposition. This is supported by the strong correlations between MAT, latitude, and elevation in our dataset.