Search

The seasonal temperature forecasting models in greenhouse based on multiple regression (MR), BP artificial neural networks (BPANN), random forest (RF) and support vector machine (SVM) were constructed using meteorological observations inside and outside the Venlo-type glass greenhouses from 27th February 2021 to 4th March 2023 at Nanjing University of Information Science & Technology, and then to validate all the models. The results showed that the fitting accuracy of the seasonal forecasting models for daily average and minimum air temperatures in the greenhouse was significantly higher than that of the seasonal forecasting models for daily maximum air temperature; the fitting accuracy of each model for air temperature inside the greenhouse in spring, summer and autumn was higher than that for winter. For the seasonal forecasting models of daily average air temperature and minimum air temperature, the fitting accuracy of the spring, summer and autumn forecasting models constructed by the four algorithms was higher, but the simulation effect of the RF model was more stable, and the coefficient of determination (R²) of the simulated values and the actual observed values were above 0.94, and the root mean square error (RMSE) and absolute error (MAE) were within 1.5 °C. For the seasonal forecast models of daily maximum temperature, the fitting accuracy of RF model for spring, summer and autumn was higher than other models in general, and its R² value was above 0.75. The MR model had higher accuracy for winter indoor air temperature and was more suitable for predicting winter greenhouse air temperature. The study concluded that it is feasible to choose the RF model to forecast the air temperature inside the glass greenhouse in spring, summer and autumn and the MR model to forecast the air temperature inside the glass greenhouse in winter, and this study can provide an important reference for the seasonal forecast of air temperature in Venlo-type glass greenhouse. [ABSTRACT FROM AUTHOR]

Objectives: Traditional remote sensing-based air temperature estimation method usually used the global models, which ignored the effects of spatiotemporal heterogeneity, especially for the researches in large regional areas. Methods: Taking the Yangtze River Economic Zone as a typical research area, this paper introduces the geographically and temporally neural networks for high-precision temperature estimation. The influence of spatiotemporal heterogeneity was considered by establishing the local models in the generalized regression neural network. Remote sensing data, assimilation data and station data were fused to obtain the spatially continuous near-surface temperature. The model performance was evaluated by the site-based ten-fold cross-validation method. Results: The results show that the geographically and temporally weighted neural network had effectively improved the estimation accuracy with the root mean square error(RMSE) =1.899 ℃, mean absolute error (MAE) =1.310 ℃ and correlation coefficient (R) =0.976. Compared with the multiple linear regression method and the traditional global neural network, the MAE value decreased by 1.112 ℃ and 0.378 ℃, respectively. Conclusions: The temperature mapping results indicated that the model used in this paper can well reflect the spatial distribution differences, which means that this study is possible to provide a new way for temperature estimation with high precision. [ABSTRACT FROM AUTHOR]

Reference crop evapotranspiration (ET0) is one of the most important hydroclimatic variables for the hydrologic and crop models, as well as the actual evapotranspiration in the irrigation schedule. The ET0 prediction a few months in advance can be beneficial to the decision-making on the long-term planning in the water management and irrigation communities. In this research, the monthly predictions of ET0 were realized over the Huaihe River basin using the model data of Beijing Climate Center second-Generation Climate Prediction System (BCC_CPSv2), and the surface meteorological observations at 172 stations during 1991 to 2020. The bilinear interpolation and quantile mapping were also selected to downscale and then correct the mean air temperature, net radiation, relative humidity, and wind speed of BCC_CPSv2. The obtained four variables were used to calculate the ET0 predictions using the Penman-Monteith equation. The correlation coefficient (r), Mean Bias Error (MBE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE) were employed to evaluate the ET0 prediction performances of the model, and the four climatic variables before and after correction. Results showed that the mean air temperature, net radiation, and relative humidity from the model were significantly smaller than those from the observations before correction in each month with the RMSE of 1.84℃, 1.70 MJ/(m2 ·d), and 15.79%, respectively. The wind speed was also smaller during March to June, but larger in other months, with the RMSE of 1.39 m/s. Errors in the climatic variables led to the lower ET0 during February to June, and the larger ET0 in January, as well as in July to December, compared with the calculations. The RMSE of ET0 before correction was ranged from 0.11 to 1.70 mm/d for each month, among which was the highest in summer. The larger RMSE exceeding 0.50 mm/d occurred in eastern coastal areas in January and October, while the southwestern mountainous areas presented the higher RMSE exceeding 0.70 mm/d in April and July. The model skills on predicting climatic variables and ET0 were improved effectively by the quantile mapping. The RMSE of air temperature, net radiation, relative humidity, and wind speed decreased to 1.32℃, 0.74 MJ/(m2 ·d), 7.38%, and 0.53 m/s, respectively. Especially, the MAPE of wind speed and net radiation was decreased by more than 19%. The RMSE of ET0 after correction was ranged between 0.05 and 1.22 mm/d for each month, indicating the decrease in 80% of the months. The performances of the model were improved significantly for the eastern coastal areas in January, and the southwest mountainous areas in April, July and October, where the RMSE decreased by more than 0.30 mm/d. Before and after the correction, the net radiation and relative humidity were the primary factors for the ET0 prediction errors in summer half year and winter half year, respectively, due to the larger errors of the model and the higher sensitivity of ET0 to them. A better performance was achieved in the monthly ET0 prediction after the model correction with the error about 10.7%. The finding can provide a strong reference for the water resources management, irrigation schedule planning, and agricultural water demand decision-making. [ABSTRACT FROM AUTHOR]

In this study, the influence of the process parameter setting of the hot-air leaf moisturizer on the quality index of the exit leaf during the secondary leaf conditioning of threshing and redrying is studied, and the corresponding prediction model is established. A BP neural network prediction model is established based on the characteristics of the secondary leaf conditioning process data. The current popular neural network writing framework TensorFlow’s high-level API interface is called to construct the neural network structure. The activation function, optimizer, number of hidden layer neurons and other key parameters are gradually adjusted in the neural network structure to make the prediction result of the test set reach the best state. By inputting the parameters of the front steam nozzle pressure, front-end water flow rate, hot air temperature, return air temperature, feed blade temperature, and feed blade moisture combination, the two key tobacco leaf evaluation indicators, namely, outlet leaf moisture and temperature, are predicted. According to the mean square error, root mean square error, and average absolute error of the prediction results, it is concluded that when the number of neurons in the hidden layer is 7, the activation function selects ReLU, and the optimizer selects RMSprop, the effect is the best. [ABSTRACT FROM AUTHOR]

Reference evapotranspiration (ET0) is an important variable to estimate all the forms of evaporation and transpiration from crops and vegetation. An accurate estimation of ET0 plays a critical role in crop water management. The Food and Agriculture Organization of the United Nations (FAO) has recommended the Penman Monteith (PM) model as the standard to estimate the ET0. However, the PM model requires a large number of input parameters, including the solar radiation (Rs) or sunshine hours (n), the maximum and minimum air temperatures (Tmax and Tmin), Relative Humidity (RH), and wind speed at 2m high (u2). In addition to temperature data, most datasets are often unavailable, incomplete, or of uncertain quality in practice. Thus, the ET0 estimation using temperature data has drawn much attention in recent years. In this study, the recent estimation models were compared to improve the high accuracy of ET0 using the temperature data. The traditional Hargreves-Samani (HS) model (HS1), the improved HS model (HS2), and the PM temperature approach (PMTPP) were recently proposed by Paredes and Pereira. Some latest models were also introduced, including the ea estimation model proposed by Qiu et al., and the temperature-based Rs model (Rs-K) proposed by Korachagaon to improve the PM temperature approach (PMTNew). All models were calibrated and tested using the dataset collected from 51 radiation stations in China during the period 1967-2016. The results showed that the Rs-K model performed a better accuracy for the ET0 estimation, compared with the Rs model proposed by Paredes and Pereira (Rs-HS). For instance, the values of modelling efficiency (EF) were much closer to 1.00, the correlation coefficient (r) increased from 0.92-0.99 to 0.94-0.99, the values of Root Mean Square Error (RMSE), and the Relative Root Mean Square Error (RRMSE) were reduced by 0.08-0.25 mm/d and 0.03-0.09, respectively, at different climate zones. Except for the arid and hyper-arid climate zones, the average values of RMSE and RRMSE in the HS2 model decreased by 0.05-0.09 mm/d and 0.02-0.03, respectively, and the average values of EF increased by 0.03-0.10, compared with the HS1 model. Although the PMTPP model produced a similar r, the mean values of RMSE and RRMSE from the dry sub-humid and humid climate zones were reduced by 0.01-0.15 mm/d and 0-0.05, respectively, while the EF values increased by 0.01-0.06, compared with the HS2 model. Therefore, the PMTPP model presented a better accuracy than the HS model in estimating the ET0 at various climate zones. Except for the arid and semi-arid climate zones, the PMTNew model further reduced the RMSE and RRMSE by 0.04-0.12 mm/d and 0-0.04, respectively, indicating the higher EF and r values, compared with the PMTPP. In addition, the HS2 model presented a higher Global Performance Index (GPI) value than the HS1 model, indicating a better GPI value in the humid zone. Except for the humid climate zones, the PMTPP model presented a higher GPI value at various climate zones, the mean value of GPI range -0.06-0.42, compared with the HS2 model. The PMTNew model also presented a better GPI of each station, 90% of which ranked the first. Therefore, it is strongly recommended to use the PMTNew model to estimate the ET0 of each station, when the limited data is available. [ABSTRACT FROM AUTHOR]