APSIM's Soil N Model Review for Future Development

The accurate simulation of soil nitrogen (N) cycling is central to the process-based model Agricultural Production Systems sIMulator (APSIM) and increasingly the focus of modellers globally on account of a growing emphasis on quantifying N losses. Recently, 24 leading APSIM model users and developers from around the world came together to write a comprehensive review of APSIM’s soil N model (https://doi.org/10.1016/j.agsy.2024.104213). The review documents in detail how the model simulates N processes and synthesizes the findings of 131 model-data comparison studies conducted over the last 26 years. Overall, the review found that APSIM’s Soil N model performs well, simulating seasonal/annual soil N uptake and loss (e.g., leaching, denitrification) accurately across a wide range of treatments/environments. A number of studies, however, noted that the model struggled to capture the daily/sub-daily N dynamics and potentially underestimated the rate of mineralisation, especially under fallow conditions. In order to remedy these shortcomings, some researchers adjusted various parameter values, but due to the disjointed manner with which these model ‘improvements’ were proposed and adopted, most have not been tested under a wider scope than the singular target process or environment of the original study. The studies often differed in their approaches to evaluating and, at times, improving model performance, with the threshold for “good” performance differing depending on the focus and scope of the study. We therefore focused on extracting the insight of the studies’ authors and revisiting their model-data evaluations in the context of the other studies, thereby seeking to delve deeper into a more comprehensive understanding of the model’s performance. Such an approach led us to uncover target areas for future model development that were not evident in singular studies. For example, it highlighted the need to revisit how fresh organic matter in the model is initialised rather than increasing the rate of turnover of other soil C pools. The review has informed ongoing work, including testing the proposed parameter changes across a range of applications to identify potential unintended consequences that exist beyond the scope of isolated studies and investigating how to better model the environmental factors that dictate daily/sub-daily N dynamics. The flexibility of APSIM’s coding allows for sensitivity analyses on the processes currently included in APSIM and the development of prototypes for processes that are beyond the current model’s capacity. Future work will look to incorporate findings from new mechanistic and field experiments across different geographic/agroecological regions. Furthermore, there is value of doing similar exercises across other process-based models. Such reviews have the potential to streamline advancements in how models are evaluated and improved, leading to the development of models with more robust predictive capabilities and broader scopes.