Modelling soil methane emissions from rice production to enable scaling

Rice production is estimated to contribute around 11% of global anthropogenic methane emissions. There is consistent evidence that management practices such as drying and rewetting the soil can reduce these emissions without impacting crop yield, yet it is challenging to quantify the emission reductions in different soils and for different management practices. Robust estimates of emission reductions would enable farmers to claim carbon credits when switching from continuous flooding (CF) to alternate wetting and drying (AWD) - and advanced modelling with remotely sensed inputs can contribute to this.  Two models (DNDC and Daycent) are commonly used to simulate methane emissions, but both require large numbers of parameters making them challenging to scale to new locations where parameter values are uncertain. We have developed a new model of soil methane emissions that shows comparable performance to these more complex models but reduces the required number of parameters to around 30 (depending on management actions) by focusing only on the processes key to methane. The proposed model uses a similar approach to the Daycent methane module in which the soil redox potential after flooding or drying is modelled, but we consider the parameters for this process to be influenced by soil texture and mineralogy as well as the drainage conditions (e.g. water table). To permit an application of the model to large regions of interest, our model can be linked to remotely sensed estimates of above ground biomass. Using a Bayesian Calibration approach, we show that the model can be successfully calibrated with local data from a single season, and validated for subsequent seasons. Using the model at new sites is less consistent; we present our progress and the ongoing challenges in defining relationships between the soil properties and the model parameters to improve emission reduction estimates at new sites and enable the model to be used at scale.