改进CERES-Rice模型模拟覆膜旱作水稻生长

摘要

覆膜旱作是节水稻作生产体系的重要措施之一,采用CERES-Rice模型模拟覆膜旱作水稻生长需另外考虑覆膜的增温效应和根系层土壤水分布差异及由此所带来的影响.该文借鉴部分旱地作物的相关研究成果,对原CERES-Rice模型中的积温和土壤温度、蒸发和土壤水分胁迫等模拟计算过程进行了改进,并进一步通过2个水稻生长季的田间试验予以验证.试验于2013、2014年在湖北房县进行,共涉及淹水(对照)、覆膜湿润栽培和覆膜旱作共3个水分处理,采用原模型和改进模型分别对2个生长季、2个覆膜处理的生育期、叶面积指数与地上部干物质质量的变化过程及产量进行模拟.结果表明:原CERES-Rice模型难以准确刻画覆膜旱作水稻的生长发育过程,经改进后,模拟效果大大改善,可有效反映环境变化(水分、温度)对覆膜水稻生育进程的影响和产量形成,维持生育期与产量模拟的相对误差在15%以内;覆膜水稻叶面积指数的动态模拟基本满足要求,其均方根差≤1.54 m2/m2、相对均方根差≤27%、建模效率≥0.85;对覆膜水稻地上部干物质质量变化过程的模拟也呈现出较好的效果,均方根差和相对均方根差分别小于1490 kg/hm2、16%,建模效率则高于0.95.总体而言,经改进后的CERES-Rice模型基本可满足要求,较好地用于模拟覆膜旱作水稻的生长发育规律.%The ground cover rice production system (GCRPS) is a potential alternative to the traditional paddy rice production system (TPRPS) by irrigating soil beds mulched with film and maintaining soils under predominately unsaturated condition, and it has become one of the most promising water-saving technologies for rice. The increase of soil temperature effected by film mulching and the unsaturated root-zone condition should be taken into consideration when CERES-Rice (a software package widely and successfully applied in TPRPS) is used to simulate rice growth in a GCRPS. In this study, the sub-modules of soil temperature and soil water in original CERES-Rice model were improved (through changing soil temperature and water conditions based on the relevant research results of the dryland crops) to evaluate the simulation on rice growth in the GCRPS. A 2-year field experiment (2013 and 2014) with 3 treatments. The treatment W1 referred to the traditional treatment with a 2-5 cm water layer on the soil beds but without plastic film mulching, W2 was the film mulching treatment keeping soil moisture in root zone near the saturated content by filling the furrows with water completely but without water layer on the soil beds, and W3 was also the film mulching treatment that was managed as the same way as the W2 before mid-tillering stage and then kept the soil moisture in root zone at 80%-100% of field water capacity. The experiment was conducted in Fang county of Hubei province, located at 32°7′N and 110°42′E to test the feasibility and rationality of the model improvement. Each treatment was replicated 3 times. A total of 9 plots were arranged and each plot was 9 m in wide and 10 cm in length. A seepage-proof material was laid around each plot under the depth of 80 cm to avoid lateral percolation from the neighbor plots. Five soil beds (156 cm wide) in each plot were built for planting rice, 6 lines of rice were planted for each soil bed with the fixed spacing (26 cm between lines and 18 cm between plants). The small furrows (15 cm in width and depth, respectively) were dug around each soil bed. Among the 2 growth seasons, the experimental data (obtained in 2013 and 2014) were used to rectify the simulation models and verify the rectified models, respectively. Based on the measured meteorological data (air temperatures/solar radiation/precipitation etc.), soil data (soil water contents/soil physical parameters/soil organic matter contents etc.) and field management data (irrigation amount/displacement/fertilizing amount by field), the changing processes of rice growth in the W2 and W3 treatments were simulated using the rectified models. The original and improved CERES-Rice models were also used to simulate the change of leaf area index, the aboveground dry weight, and the rice yield during the 2 growth seasons. The results of the comparison showed that the improved CERES-Rice model had remarkable superiority in delineating the effects of changing environments (e.g. soil temperature and soil water) on rice growth and production in the GCRPS. Both of the estimation of the phenological phases and yields were in good agreement with the measured values, and the relative error was not more than 15%. The root mean squared errors (RMSE) between the simulated and measured leaf area index was not higher than 1.54 m2/m2, the correspondingly normalized root mean squared errors (NRMSE) was not higher than 26.59% and the values of modeling efficiency (EF) were not less than 85%. Moreover, the simulated dynamics of aboveground dry weight were compared well with the measured values (RMSE was smaller than 1490 kg/hm2, NRMSE was smaller than 16%, but EF was not less than 0.95). Therefore, the improved CERES-Rice model is rational and reliable to simulate rice growth and production in GCRPS.