Assessing the sensitivity of modelled estimates of N₂O emissions and yield to input uncertainty at a UK cropland experimental site using the DailyDayCent model

Abstract

Biogeochemical models such as DailyDayCent (DDC) are increasingly used to help quantify the emissions of green-house gasses across different ecosystems and climates. For this use they require parameterisation to represent a heterogeneous region or are site specific and scaled upwards. This requires information on inputs such as climate, soil, land-use and land management. However, each input has an associated uncertainty, which propagates through the model to create an uncertainty in the modelled outputs. To have confidence in model projections, an assessment of how the uncertainty in inputs propagated through the model and its impact on modelled outputs is required. To achieve this, we used a pre-defined uncertainty range of key inputs; temperature, precipitation, clay content, bulk density and soil pH, and performed a sensitivity and uncertainty analysis, using Monte Carlo simulations. This allowed the effect of measurement uncertainty on the modelled annual N₂O emissions and crop yields at the Grange field experimental site to be quantified. Overall the range of model estimates simulated was relatively high and while the model was sensitive to each input parameter, uncertainty was driven by the sensitivity to soil pH. This decreased as the N fertiliser application rate increased, as at lower N application rates the model becomes more sensitive to other drivers of N mineralisation such as soil and climate inputs. Therefore, while our results indicate that DDC can provide a good estimate of annual N₂O emissions and crop yields under UK conditions, reducing the uncertainty in the input parameters will lead to more accurate simulations.