Calibration of the DayCent Model for Native Pastures in South-Eastern Queensland

Soil organic carbon dynamics are strongly influenced by soil and climate conditions, as well as management practices including grazing and cropping. Over the past two decades, biogeochemical models have been widely used for analysing the effect of different environmental and management variables on soil carbon, including potential change under hypothetical future climate and management scenarios. The DayCent model, which is a daily implementation of the Century model, considers the impacts of soil texture, climate, historical vegetation cover, and land management practices, including crop type, fertilizer additions, and cultivation events on soil carbon dynamics.

In this study, we calibrate the DayCent model for two long-term (38-year) native pasture exclosures at one location in south-eastern Queensland. These sites have had similar management, being ungrazed and burnt or mown at the beginning of each pasture growing season but differ with respect to soil type (texture and depth) and species composition. One site is dominated by kangaroo grass (Themeda triandra), which represents the species composition prior to the introduction of tree clearing and grazing by cattle in the late 1800s. The other site is dominated by black spear grass (Heteropogon contortus) which has become the dominant species in the region since that time. To reflect the long-term species composition changes in the region, kangaroo grass crop parameters were used to run the model to equilibrium from year 1 AD to the year 1900 for both sites, and spear grass parameters were introduced in 1901 for the spear grass site.

The model calibration concentrated on the key ‘crop’ parameters governing potential production, root to shoot ratio, and plant carbon to nitrogen ratio. The calibrated DayCent model accounted for only 21 percent of the observed year-to-year variability in end-of-season above-ground biomass at the kangaroo grass site and 58%  at the spear grass site. The observed biomass production for the two sites was most strongly correlated with simulated evapotranspiration during the growing season (R² = 0.43 and 0.58 for kangaroo grass and spear grass respectively) and we found a strong correlation between simulated and observed soil water content to a depth of 50 cm at both sites (R² = 0.64 and 0.6 for kangaroo grass and speargrass respectively).

Whilst year-to-year variability was not well simulated, the long-term average production of each site is the main driver of soil carbon. For both sites, the model overestimated the average observed above-ground biomass at the end of the growing season by approximately 15 percent. By this time of year, the plants have flowered and lost biomass through the detachment of seeds and seed heads as well as some dead leaves. The timing of this detachment process is difficult to simulate in DayCent and it is therefore likely that DayCent simulated the annual biomass production quite closely. It remains to validate the DayCent simulations against similar long-term production data at a further six long-term study sites at this location and to evaluate how well DayCent simulates observed soil carbon across soil types, both under grazed and ungrazed conditions.