JIA-2019-11

2526 ZHOU Tian-mei et al. Journal of Integrative Agriculture 2019, 18(11): 2521–2533 Independent different data sets were then used for further validation of these models. 3.3. Validation of the photothermal simulation model Independent data sets were used to validate the model, and RMSE was calculated to determine the accuracy of the models. Validation of the photothermal simulationmodels in tomato To determine the accuracy of the models, the simulated values were compared with the measured values through calculating the RMSE. The simulated G value and healthy indexes were calculated using the corresponding model equations and the TEP s . The results showed that the RMSE values were small (0.0018 and 0.0004 ( n =14), respectively) for G value and HI1, which indicated that both models have high prediction accuracy, and the trends between the simulated and measured values were similar (Fig. 4-A and B). By contrast, larger deviation between simulated and measured values for HI2 and HI3 resulted in higher RMSE values, 0.0089 for HI2 model and 0.0099 for HI3 model ( n =14), indicating that the models have a lower prediction accuracy (Fig. 4-C and D). For the cultivar Jinlingmeiyu, the R 2 of the simulated and measured G value based on the 1:1 line was 0.9076, and the RMSE was 0.0020, indicating that the fitting degree and reliability of G value model was low (Fig. 5-A). The R 2 of the simulated and measured HI1 based on the 1:1 line was 0.9642, and RMSE was only 0.0006, suggesting that the HI1 model had a high fitting degree and reliability (Fig. 5-B). For the cultivar Hezuo 903, similar results were observed. The fitting degree and reliability of the HI1 model were higher than those of the G value model (Fig. 5-C and D). Validation of the photothermal simulation models in cabbage The RMSE for cabbage simulation growth models were calculated using the simulated G value and healthy indexes that were generated according to the corresponding model and the TEPs. Similar to tomato models, RMSE for models developed with the G value and HI1 of cabbage were significantly low with 0.0007 for G value model and 0.0004 for HI1 model ( n =14), indicating that the G value and HI1 models have a higher prediction accuracy for greenhouse cabbage seedlings (Fig. 6-A and B). The RMSE values for HI2 and HI3 models were 0.0098 and 0.0037 ( n =14), respectively, which were higher than those for G value and HI1 models; this was mainly due to larger deviations between the simulated and measured values (Fig. 6-C and D). The higher RMSE values indicated that these HI2 and HI3 models are less accurate compared to G value and HI1 models. For the cultivar Sugan 27, the R 2 of the simulated and measured G value based on the 1:1 line was 0.9283, and the RMSE was 0.0008 (Fig. 7-A). The low R 2 indicated that the G value model may have a biased prediction. By contrast, the R 2 of the simulated and measured HI1 based on the 1:1 line was 0.9897, and RMSE for this model was 0.0004 (Fig. 7-B), suggesting that the HI1 model has a high fitting degree and prediction reliability. Similarly, we also found that G value model for Bochun has lower R 2 and higher RMSE compared to its HI1 model (Fig. 7-C), indicating a relatively lower prediction reliability of the G value model; The R 2 and RMSE of the HI1 model were 0.9821 and 0.0004, respectively (Fig. 7-D), indicating that this model can accurately simulate the change of HI1 in cabbage. Table 3 Growth characters and seedling quality indexes of tomato and cabbage seedlings 35 days after sowing, and the corresponding TEP 1) Sowing date/ Cultivar 2) TEP Seedling height (cm) Stem diameter (mm) Shoot fresh mass (mg) Root fresh mass (mg) Shoot dry mass (mg) Root dry mass (mg) Whole fresh mass (mg) Whole dry mass (mg) Root cap ratio G value HI1 HI2 HI3 14/9/18JL 82.85 24.04 b 3.73 b 5 148.31 a 514.74 b 509.48 ab 65.65 b 5 663.05 a 575.13 a 0.1302 b 0.0164 ab 0.0090 b 0.0882 b 0.0832 b 14/9/28JL 80.31 27.85 a 3.74 b 5 218.34 a 506.14 b 523.83 a 67.64 b 5 724.48 a 591.47 a 0.1299 b 0.0169 ab 0.0080 b 0.0771 b 0.0845 b 14/9/28HZ 80.31 20.20 c 3.66 b 4 132.20 b 452.03 b 426.94 b 61.61 b 4 584.22 b 488.55 a 0.1421 b 0.0139 b 0.0090 b 0.0842 b 0.0888 b 15/4/23HZ 91.89 18.30 d 4.17 a 4 266.68 b 749.97 a 462.11 ab 84.53 a 5 016.65 ab 546.64 a 0.1824 a 0.0182 a 0.0125 a 0.1146 a 0.1128 a 15/4/23SG 57.51 8.95 b 2.90 a 3 793.50 a 409.45 a 380.70 a 58.87 a 4 202.94 a 439.57 a 0.1562 a 0.0126 a 0.0142 a 0.1361 a 0.0827 a 15/5/12SG 44.26 9.18 ab 2.94 a 3 660.85 a 165.64 c 300.74 b 34.95 c 3 826.49 a 335.69 b 0.1172 b 0.0096 b 0.0108 b 0.1231 ab 0.0499 c 15/4/23BC 57.51 8.85 b 2.56 b 3 727.17 a 291.96 b 396.42 a 46.51 b 4 019.13 a 442.94 a 0.1218 b 0.0127 a 0.0132 a 0.1211 ab 0.0656 b 15/5/12BC 44.26 10.24 a 2.92 a 3 694.06 a 191.97 c 315.57 b 31.46 c 3 886.03 a 347.03 b 0.1019 c 0.0099 b 0.0100 b 0.1109 b 0.0447 c 1) TEP stands for product of thermal effectiveness (TE) and photosynthetically active radiation (PAR). Root-shoot ratio, G value and seedling healthy indexes (HI1, HI2 and HI3) reflect seedling qualities, and the calculations were performed according to eqs. (4)–(8). 2) JL is tomato cultivar Jinlingmeiyu; HZ is tomato cultivar Hezuo 903; SG is cabbage cultivar Sugan 27; BC is cabbage cultivar Bochun.