Quantifying nitrogen flows from agricultural land using an ABM to inform future land use change for ecohydrological modelling

In agricultural catchments, the decisions made by farmers regarding crop choices and field management practices influence both water flow paths through the landscape and the amounts of reactive nitrogen (N) applied to fields. Agent-based models (ABMs) capture regionally specific, complex, and dynamic interactions between farmers and natural systems across multiple scales, enabling the simulation of physical landscape changes. In this study, we applied a novel methodology using an ABM to generate annual land use layers to inform a hydrological model. This approach allows us to spatially identify specific crops contributing to excess nitrogen flows in the landscape and into water bodies. Additionally, we link the temporal sequence of crops within their crop rotations to weather patterns, enabling us to examine nitrogen transport pathways under future change scenarios. We used the SECLAND ABM to inform the SWAT ecohydrological model of annual land use change at the field scale.

Using SECLAND, three agricultural land use scenarios were developed: a business-as-usual (BAU) scenario, an extensification scenario (based on SSP1), and an intensification scenario (based on SSP5). We integrated the respective annual land use layers derived from the ABM into SWAT, with and without climate change scenarios, to quantify the resulting impacts on crop yields, water balance components, nitrogen concentrations in surface flows, nitrogen leaching, and nitrogen in groundwater.

The ABM results for the BAU scenario over the next 35 years show a decrease in the number of active farms, accompanied by a loss of agricultural areas. The results also indicate a transition toward organic farming and a shift in intensity toward extensification. In all land use scenarios, less corn is grown. As well, the area of forested land increases in the future.

For all three land use scenarios, implementing the land use layers with their annual crop rotations in SWAT led to significant differences in nitrogen losses (kg/ha) to surface and subsurface water bodies, compared to using a static land use approach. All three land use scenarios consistently showed lower nitrogen losses per area to the environment, particularly for crops requiring high levels of nitrogen fertilizer (e.g., corn and winter rapeseed).

When the land use scenarios were implemented in conjunction with climate change simulations in SWAT, lower N loads in lateral flow and groundwater was simulated, and hence in reduced nitrogen losses per area. Implementing crop rotations in the SWAT model also reduced the number of water and nitrogen stress days for the crops.

Our findings underscore the importance of including detailed spatial crop rotations when assessing regional water quality, particularly in conjunction with climate change scenarios. Furthermore, we found that using static land use in hydrological modeling generally leads to an overestimation of nitrogen losses, especially from crops with high fertilizer applications.