Comparison of simulation accuracy of potato phenology models and their sensitivity to rising temperature

Potatoes are the fourth largest staple food crop worldwide. The accurate simulation of potato phenology is the basis for assessing the impacts of climate change on potato growth and development. Selecting the optimal potato phenology models across China’s potato planting regions, and exploring the difference in the simulated impact of historical climate change on the potato growth period among these phenology models, as well as their sensitivity to temperature rise (1−5℃), could help provide direction for model improvement. However, there have been few studies on the integrated comparison of potato phenology models across different climate types and cultivars. In this study, three temperature response functions and two day-length response functions were combined into three temperature-based and six light-temperature-based models. The accuracy of the nine models in simulating the length of the potato growth period at eight sites across China’s potato planting regions was compared with the derived model parameters based on a simulated annealing algorithm. The study results showed that: 1) The simulation accuracy of linear function-based temperature with a negative exponential function-based photoperiod-temperature model (M3), logistic function-based temperature with a negative exponential function-based photoperiod-temperature model (M6), and beta function-based temperature with a negative exponential function-based photoperiod-temperature model (M9) had optimal simulation accuracies among the nine phenology models. The root mean square errors (RMSEs) between the observed and simulated lengths of the potato growth period, simulated by M3, M6, and M9, were 7.7 d, 6.8 d, and 7.1 d, respectively (R²>0.90). 2) Based on the three optimal phenology models, the simulation results at the four representative sites showed that historical climate change had decreased the length of the potato growth period, although there were differences between the models at different sites. 3) When the temperature was increased by 1℃ to 5℃, the simulated length of the potato growth period by M3 and M6 were linearly shortened with the decline rate of 3.0–8.2 d per 1℃ rise. With increasing temperature, the shortening trend of the simulated length of the potato growth period by M9 first increased and then decreased. Our study found that the three optimal models could effectively simulate potato phenology across China’s potato planting regions under current climate conditions. All the simulated results, based on the three optimal phenology models, showed that historical climate change shortened the length of the potato growth period, but with different rates of decrease simulated by different models. The photoperiod response function in the phenology models changed the sharpness of the temperature response function and increased the temperature sensitivity of the models. The differences between the models increased with increasing temperature. Our study provides an important reference for the selection and improvement of potato phenology models.