| 英文摘要: | To the Editor —
We thank Houlton et al.1 for presenting a comparison between modelled (CLM4.5) and observationally inferred (using δ15N) estimates of the proportion of denitrification nitrogen loss (fdenit) versus total nitrogen emissions (denitrification plus hydrological losses) in natural terrestrial ecosystems. We agree that terrestrial models must represent these losses accurately if they are to credibly estimate emissions of nitrogen- and carbon-containing greenhouse gases important in climate change predictions.
They demonstrated that CLM4.5 predicted unrealistic frequency and spatial distributions of fdenit. Our recent work with CLM4.5 (and the identical land model ALM; ref. 2) indicates that the failure in this regard is due primarily to unrealistic assumptions regarding nitrogen competition and also to poor numerical representation of advective fluxes. These models assume a sequential competitive structure: first, plants and free-living decomposing and nitrifying microbes use available soil ammonium (scaled by their relative demands); second, denitrifiers use the available nitrate; and finally hydrological processes (that is, leaching and runoff) access the (often depleted) residual nitrate. In this approach, hydrological nitrogen losses are usually unrealistically small compared with denitrification losses.
To address this problem, we modified the nitrogen competition module with the equilibrium chemistry approximation (ECA) approach3, 4 and improved the representation of leaching fluxes. ECA represents the competition between multiple substrates (NH4+ and NO3−, for example) sharing one consumer and multiple consumers (plants, decomposing microbes, denitrifiers and so on) sharing one substrate (such as NO3−). These changes improved the model comparison with the probability, latitudinal, and longitudinal distributions of δ15N-inferred fdenit (Fig. 1). However, the improved model has larger spatial heterogeneity. The δ15N-inferred fdenit estimates are extrapolated from observed temperature and precipitation and are sensitive to isotope effects during denitrification5. In contrast, modelled denitrification and hydrological nitrogen losses are primarily controlled by hydrological dynamics, soil oxygen content, and soil nutrient competition, which tend to be more heterogeneous than air temperature and precipitation. |