Short Term Prediction Model of Environmental Parameters in Typical Solar Greenhouse Based on Deep Learning Neural Network
Round 1
Reviewer 1 Report
Paper is interesting and novel methods applied in the agriculture sector, I have accepted the paper with minor revision. author please find out the some comments in attachment.
Comments for author File: 
Comments.pdf
Author Response
Response to Reviewer 1 Comments
Thank you very much for you appreciation.
Point 1: The manuscript is quite long. Should be revised to be more concise. There are to many long sentences. Consisted of three to four lines which can be split and written more precisely. Also, I noticed that there are too many incomplete sentences or a dot is missing. Fix it.
Response 1: The research targets of this manuscript are winter wheat and summer maize in five cities in the North China Plain, involving historical data such as the planting area, yield per unit area, and effective precipitation of the irrigation system, as well as the prediction method and uncertainty planning method of deep learning. So this manuscript is indeed quite long. In order to maintain the methodological and results integrity of this study, while streamlining the manuscript, we transformed the data in Figures 6 and 7 into Tables 10 and 11.
For the long sentences and incomplete sentences in the manuscript, we have revised it according to your comments, and applied for the English editing service of this journal. For details, please refer to the manuscript review section.
Point 2: The abstract needs to be improve as study area.
Response 2: Revised abstract:
Due to population growth and human activities, water shortages have become an increasingly serious concern in the North China Plain, which has become the world’s largest underground water funnel. Because the yield per unit area, planting area of crops, and effective precipitation in the region are uncertain, it is not easy to plan the amount of irrigation water for crops. In order to improve the applicability of the uncertainty programming model, a hybrid (Long short-term memory, chance-constrained programming, fuzzy possibility programming, interval parameter programming) LSTM-CPP-FPP-IPP model was developed to plan the irrigation water allocation of irrigation system under uncertainty. The LSTM (Long short-term memory) model was used to predict crop yield per unit area, and CPP-FPP-IPP programming (chance-constrained programming, fuzzy possibility programming, interval parameter programming) was used to plan the crop area and the effective precipitation under uncertainty. The hybrid model was used to crop production profit of winter wheat and summer corn in five cities, in the North China Plain. The average absolute error between the model prediction value and the actual value of the yield per unit area of winter wheat and summer maize in four cities in 2020 is controlled within the range of 14.02 to 696.66 kg/hectare. It shows that the model can more accurately predict the yield per unit area of​​crops. The planning model for the benefit of irrigation water allocation generated three scenarios of rainfall levels and four planting intentions, and compared the planned scenarios with the actual production benefits of the two crops in 2020. In a dry year, the possibility of planting areas for winter wheat and summer corn is optimized. Compared with the traditional deterministic planning method, the model takes into account the uncertain parameters, which helps decision makers seek better solutions under uncertain conditions.
Point 3: In Figures- you have to many figures inside a figure. Put the location inside each figure- It is hard to scroll down and look back the response for each location.
Response 3: The title of the figure has been modified,For details, please refer to the manuscript review section.
Point 4: revise carefully all the manuscript.
Response 4: we have revised it according to your comments, and applied for the English editing service of this journal. For details, please refer to the manuscript review section.
Reviewer 2 Report
The paper ‘Short Term Prediction Model of Environmental Parameters in Typical Solar Greenhouse Based on Deep Learning Neural Network’ provides an interesting and innovative study on prediction models of micrometeorological variables in greenhouses. I do not have any major revisions for the paper, so my recommendation was for 'Accept after minor revision'.
Additional comments
1) Data were collected over a short period in the years 2018 and 2020. What are the implications of this for the efficiency of the models? Why not evaluate throughout 2018 and 2020? It could improve the model because it would use micrometeorological variables of the summer, autumn, winter and spring seasons.
Author Response
Response to Reviewer 2 Comments
Thank you very much for you appreciation.
Point 1: Data were collected over a short period in the years 2018 and 2020. What are the implications of this for the efficiency of the models? Why not evaluate throughout 2018 and 2020? It could improve the model because it would use micrometeorological variables of the summer, autumn, winter and spring seasons.
Response 1: There are practical reasons for collecting short-term data in this paper, because in northern China, the key growth periods of solar greenhouse crops are mainly in summer and winter, and spring and autumn are not the key growth periods of solar greenhouse, so the actual demand for monitoring and management of solar greenhouse environment in these two seasons is insufficient. On the other hand, due to the relatively diversified management of crops planted in the solar greenhouse in winter, there are active heating forms in the form of hot air stove, heating pipe, warm quilt, etc., so considering the complexity of greenhouse system management in winter, this paper only collects the more important greenhouse environmental parameters in the summer and establishes a prediction model.
Reviewer 3 Report
Comments on „Short term prediction model of environmental parameters in typical solar greenhouse based on deep learning neural net- work”
The title reflect the content of the manuscript and convey to the readers the scope, design, and goal of the research.
The abstract is a brief summary of the manuscript complete with concise description of key methodological features of the study and important research findings.
The introduction provides a good, generalized background of the topic that quickly gives the reader an appreciation of the wide range of applications for this technology.
Abstract
The study has been conducted so verbs should be in "past tense" (also throughout the manuscript)
Introduction
The background section provides an introduction to the single-slope solar greenhouse indoor environment which is influenced by the outdoor weather parameters.
According to current evidence, architecture and hyperparameters are the main factors affecting the prediction accuracy of the deep learning models. Thus, deep learning prediction models of environmental parameters and the hyperparameters determined by orthogonal experimental design (OD), could have a high impact on irrigation decision-making and water allocation in single-slope solar greenhouses.
I think the motivations for this study need to be made clearer. The purpose of the research emerges from the background, but by the end of the paragraph, the purpose is defined only by the novelty brought by the applied research methods and not by clear definition of the research scope which otherwise is mention in the abstract.
Lines 91-94: „Although there have been many related studies on solar greenhouses, it is difficult to apply the above models to an actual solar greenhouse. Since the interior space of the solar greenhouse is small and the change of the greenhouse internal environmental parameters is gradual” - merge the sentences for better understanding.
Line 162: delete „ from two typical greenhouses, respectively” (readers already know that research develops in two typical greenhouses).
Lines 173-174: „ The current and historical outdoor meteorological parameters downloaded from the 173 website (www.data.cma.cn/)”. The time of the verb is missing (were downloaded).
Material and methods
The methods section identifies the nature of the data analyzed in the study and answers the question how was the study conducted.
The overall research method of the study includes experimental, data collection, model training and optimization, prediction, and evaluate.
Figure 8 provides a very clear illustration of the entire work methodology that is simple to follow.
Results
The results clearly explained and presented in an appropriate format.
The figures and tables show essential data are easy to interpret and no data is duplicated in the text.
The explanation of the abbreviations at the bottom of the tables is missing (even if they are mentioned in the text). For figures 4, 5, 6 and 7, 8 the same observation.
Discussions
The findings are not properly described in the context of the published literature!
The paragraph should be rewritten with reference to the positive or negative results of similar research.
Conclusions
The conclusions of the study are supported by appropriate evidence.
Literature cited
The literature cited is relevant to the study, but there are several instances, which have been noted above (Discussion paragraph), in which the author doesn’t makes assertions concerning the results of similar research substantiating them with references.
Maintain uniformity in the writing of cited bibliographic titles.
Comments for author File: Comments.pdf
Author Response
Response to Reviewer 3 Comments
Thank you very much for you appreciation.
Point 1: The study has been conducted so verbs should be in "past tense" (also throughout the manuscript).
Response 1: The type of single-slope solar greenhouse is mainly used for vegetable production in China. The coupling of heat storage and release courses and the dynamic change of the outdoor weather parameters affect the indoor environment momentarily. Due to the high cost of small weather stations, the environmental parameters monitored by the nearest meteorological stations are usually used as outdoor environmental parameters in China. In order to accurately predict the solar greenhouse and crop water demand, this paper proposes three deep learning models, in-cluding neural network regression (DNNR), long short-term memory (LSTM), and convolutional neural network- long short-term memory (CNN-LSTM), and the hyperparameters of three models were determined by orthogonal experimental design (OD). The temperature and relative humidity monitored by the indoor sensors and outdoor weather station were taken as the inputs of models, the temperature and relative humidity 3, 6, 12 and 24 hours in advance were taken as the output, 16 combinations of input and output data of two typical solar greenhouses were trained separately by three deep learning models, those models were trained 144, 144 and 288 times, respectively. The best model of three type models at four prediction time points were selected, respectively. For the forecast time point of 12 hours in advance, the errors of the best LSTM and CNN-LSTM models in two greenhouses were all smaller than the DNNR models. For other three time points, the results show that the DNNR models have excellent prediction accuracy among three models. The max-imum and minimum temperature, relative humidity, and ETo were also accurately predicted using the corresponding optimized models. In sum, this study provided an optimized deep learning prediction model for environmental parameters of greenhouse and provide technical support for irrigation decision-making and water allocation.
Point 2: Introduction, the background section provides an introduction to the single-slope solar greenhouse indoor environment which is influenced by the outdoor weather parameters.
According to current evidence, architecture and hyperparameters are the main factors affecting the prediction accuracy of the deep learning models. Thus, deep learning prediction models of environmental parameters and the hyperparameters determined by orthogonal experimental design (OD), could have a high impact on irrigation decision-making and water allocation in single-slope solar greenhouses.
I think the motivations for this study need to be made clearer. The purpose of the research emerges from the background, but by the end of the paragraph, the purpose is defined only by the novelty brought by the applied research methods and not by clear definition of the research scope which otherwise is mention in the abstract.
Response 2: In general, in order to predict short-term indoor environmental parameters and crop water consumption of solar greenhouse, this paper optimizes three deep learning prediction models through orthogonal experimental design to obtains the best prediction model.
Point 2: Lines 91-94: „Although there have been many related studies on solar greenhouses, it is difficult to apply the above models to an actual solar greenhouse. Since the interior space of the solar greenhouse is small and the change of the greenhouse internal environmental parameters is gradual” - merge the sentences for better understanding.
Response 2: Although there have been many studies on the environmental parameters of solar greenhouses, the models of these studies are difficult to apply to the actual solar green-house. Because the internal space of solar greenhouse is small, the changes of its internal environmental parameters is gradual. Therefore, the cost of environmental parameter sensor inside the greenhouse is low, and the monitoring accuracy is higher. In contrast, the outdoor environmental parameters of solar greenhouses are usually monitored by small weather stations, the monitoring coverage of weather stations is relatively small, and the outdoor environment of the solar greenhouse is complex and changeable, and the accuracy of environmental parameters collected by small weather stations is relatively poor.
Point 3: Line 162: delete „ from two typical greenhouses, respectively” (readers already know that research develops in two typical greenhouses).
Lines 173-174: „ The current and historical outdoor meteorological parameters downloaded from the 173 website (www.data.cma.cn/)”. The time of the verb is missing (were downloaded).
Response 3: These sentences have been deleted. “were” were added.
Point 4: The explanation of the abbreviations at the bottom of the tables is missing (even if they are mentioned in the text). For figures 4, 5, 6 and 7, 8 the same observation.
Response 4: The explanation of the abbreviations at the bottom of the tables are mentioned in the text, and its easy to understand for readers.
Point 5: The paragraph should be rewritten with reference to the positive or negative results of similar research.
Response 5: The indoor temperature and relative humidity are keys to the healthy growth of solar greenhouse crops. In order to implement precise control and make irrigation decisions in advance, it is necessary to construct models to predict the indoor environmental parame-ters and ETo [1,46]. Accurate prediction of indoor and outdoor environmental parameter and ETo of solar greenhouses requires complete indoor and outdoor environmental pa-rameters of the greenhouse.
Although many kinds models have been developed for indoor environmental pa-rameters of greenhouses, has all have their own drawbacks. Previously, Imran et al. [12] utilizes artificial neural networks for prediction of hourly mean values of ambient tem-perature. Morteza et al. [17] compare mathematical models with artificial neural network and select the best prediction models. Yu et al. [20] present a novel temperature prediction model based on a least squares support vector machine model with parameters optimized by the improved particle swarm optimization. Yue et al. [18] proposed a model to predict the temperature and humidity of a greenhouse based on improved LM-RBF. Wang et al. [47] propose the model of greenhouse temperature and parameter states. All these models mainly consider the environmental parameters inside the greenhouse and the change trend of time series. Jung et al. [8] proposed three deep learning models to predicting en-vironmental parameters change. Although these models considering the external and in-door environmental parameters of the greenhouse, the prediction time steps are only from 5 to 30 minute. Most of the indoor environmental parameters (air temperature and air humidity) of solar greenhouses in developing countries can be monitored completely and accurately [13,48]. Agricultural activities, especially crops production in greenhouse, are sensitive to the prediction time steps of the prediction model. Although the shorter the prediction time, the higher the accuracy of the deep model. In the actual greenhouse crop irrigation decision and management process, the prediction time steps from 5 to 30 min are actually relatively short.
The outdoor environmental parameters monitored by small weather stations, due to the high cost of outdoor environmental parameter acquisition equipment and the small monitoring coverage, the environmental parameters prediction models using outdoor en-vironmental parameters and indoor environmental parameters for solar greenhouses in China are still not entirely reliable, and a universal model has not yet been reported. In this study, three deep learning time series models were built, and the hyperparameters of these models were designed using the orthogonal design. In order to verify the precision of three models under long time steps, indoor and external environmental parameters from the nearest weather station were predicted using three models in 16 datasets, the opti-mized model was used to predict environmental parameters under four forecast time points, 3h, 6h, 12h and 24h, respectively. The result show that the DNNR model shows the best performance for the prediction of short-term greenhouse environmental parame-ters. The optimized models can help users realize prediction of water consumption with-out the recent environmental parameters of the solar greenhouse, which can reduce de-pendence of prediction of greenhouse crop water consumption on environmental detec-tion sensors.
