Improving model skill in simulating phenology in rice – a case study in the U.S. Mississippi delta

Authors: LI, SANAI - Forest Service (FS); Fleisher, David; Timlin, Dennis; Reddy, Vangimalla

Submitted to: BARC Poster Day
Publication Type: Abstract Only
Publication Acceptance Date: 4/4/2019
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: Warmer temperatures, such as that predicted under future climates, will greatly impact the rate of crop development which can have negative consequences on U.S. crop yields and grain quality. Crop models can be used to assess current and potential climate impacts and identify adaptation responses. However, the ability of these tools to accurately simulate high temperature impacts on developmental rates, including that of rice, needs to be rigorously tested. ORYZA (2000/v3) is one of the most popular physiological-based crop models, and has been calibrated and validated in countries primarily using the indica sub-species of Asian rice. In the present study, we used over 20 years of experiment station data for two cultivars located in the Mississippi Delta to evaluate and improve the model’s temperature responses for U.S. production conditions. First, we observed that ORYZA’s default base and optimal temperatures needed to be adjusted. Improving cardinal temperatures resulted in improved model accuracy and reduced systematic error where the correlation coefficient between observed and simulated growth duration increased by 0.24 and root mean square error reduced by 0.75-2.2 days for cultivar Lemont from 1983 to 1998. Second, we evaluated ORYZA’s bi-linear, thermal time approach for simulating rice developmental rate. We show that the bilinear model can successfully reproduce the temperature response for rice under historical climate in the U.S. However, this approach results in systematic error in simulated phenology under high and low temperatures, which are increasingly more prevalent in U.S. rice growing regions. A smooth curve, such as a beta-distribution function, showed a better performance in simulating the growth duration under future climate conditions, especially when temperature is above the optimal value for development. Using a modified version of ORYZA, our studies showed that the warmer temperatures expected by 2050 in RCP4.5 and RCP8.5 climate scenarios hasten phenological development and shortened growth duration by 3-20 days compared to 1990s. Under the RCP85 scenario in 2080, growth duration was reduced by 13-24 days at relatively cool site such as Stuttgart, LA, but may actually increase growth duration by 7-8 days at a warmer site such as Stoneville, MS as a result of high temperature stress. Our analysis showed that the bilinear model was more sensitive to temperature increases as compared to the beta function method, and thus was more likely to underestimate or overestimate rice developmental responses. These temperature response improvements to ORYZA are a necessary step to increase the accuracy of the model to support decision support options for U.S. rice growers.