This study considers the relative performance of six different models to predict soil respiration from upland peat. Predicting soil respiration is important for global carbon budgets and gap filling measured data from eddy covariance and closed chamber measurements. Further to models previously published new models are presented using two sub-soil zones and season. Models are tested using data from the Bleaklow plateau, southern Pennines, UK. Presented literature models include ANOVA using logged environmental data, the Arrhenius equation, modified versions of the Arrhenius equation to include soil respiration activation energy and water table depth. New models are proposed including the introduction of two soil zones in the peat profile, and season. The first new model proposes a zone of high CO2 productivity related to increased soil microbial CO2 production due to the supply of labile carbon from plant root exudates and root respiration. The second zone is a deeper zone where CO2 production is lower with less labile carbon. A final model allows the zone of high CO2 production to become dormant during winter months when plants will senesce and will vary depending upon vegetation type within a fixed location. The final model accounted for, on average, 31.9% of variance in net ecosystem respiration within 11 different restoration sites whilst, using the same data set, the best fitting literature equation only accounted for 18.7% of the total variance. Our results demonstrate that soil respiration models can be improved by explicitly accounting for seasonality and the vertically stratified nature of soil processes. These improved models provide an enhanced basis for calculating the peatland carbon budgets which are essential in understanding the role of peatlands in the global C cycle.