Forecasting Root-Zone Electrical Conductivity of Nutrient Solutions in Closed-Loop Soilless Cultures via a Recurrent Neural Network Using Environmental and Cultivation Information

Moon, Taewon; Ahn, Tae In; Son, Jung Eek

doi:10.3389/fpls.2018.00859

Cited by 26 publications

(14 citation statements)

References 44 publications

(56 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recently, deep learning has been used to draw meaningful interpretations from complicated nonlinear data [15, 18, 43], and also showed meaningful result in agriculture. Recently, deep learning approach has been used to estimate CO 2 concentrations in greenhouses with acceptable levels of accuracy [31]. As part of deep learning, long short-term memory (LSTM) is used to analyze time-series data, such as voice recognition, video recognition and natural language processing.…”

Section: Introductionmentioning

confidence: 99%

Long short-term memory for a model-free estimation of macronutrient ion concentrations of root-zone in closed-loop soilless cultures

2019

Self Cite

View full text Add to dashboard Cite

Background Root-zone environment is considered difficult to analyze, particularly in interpreting interactions between environment and plant. Closed-loop soilless cultures have been introduced to prevent environmental pollution, but difficulties in managing nutrients can cause nutrient imbalances with an adverse effect on crop growth. Recently, deep learning has been used to draw meaningful results from nonlinear data and long short-term memory (LSTM) is showing state-of-the-art results in analyzing time-series data. Therefore the macronutrient ion concentrations affected by accumulated environment conditions can be analyzed using LSTM. Results The trained LSTM can estimate macronutrient ion concentrations in closed-loop soilless cultures using environmental and growth data. The average training accuracy of six macronutrients was R 2 = 0.84 and the test accuracy was R 2 = 0.67 with RMSE = 1.48 meq L −1 . The used values of input interval and time step were 1 h and 168 (1 week), respectively. The accuracy was improved when the input interval became shorter, but not improved when the LSTM consisted of a multilayer structure. Regarding training methods, the LSTM improved the accuracy better than the non-LSTM. The trained LSTM showed relatively adequate accuracies and the interpolated ion concentrations showed variations similar to those seen during traditional cultivation. Conclusions We could analyze the nutrient balance in the closed-loop soilless culture, the model showed potential in estimating the macronutrient ion concentrations using environmental and growth factors measured in greenhouses. Since the LSTM is a powerful and flexible tool used to interpret accumulative changes, it is easily applicable to various plant and cultivation conditions. In the future, this approach can be used to analyze interactions between plant physiology and root-zone environment.

show abstract

Section: Introductionmentioning

confidence: 99%

Long short-term memory for a model-free estimation of macronutrient ion concentrations of root-zone in closed-loop soilless cultures

2019

Self Cite

View full text Add to dashboard Cite

show abstract

“…The total concentration of nutrients in the system using CM for nutrient replenishment gradually increased with diurnal level fluctuations, and after approximately 60 days, the total concentration showed repeated fluctuations within a certain range (Figure 2a). The changes with an increasing tendency in total nutrient concentrations relative to initial values have been typically reported in most EC-based closed-loop, semi-closed-loop, and open-loop soilless culture systems [8,13,14,16,29]. Theoretically, the concentration of nutrient solutions in the substrates can be explained by the difference between the concentration of irrigated solution and the concentration of nutrient uptake when the boundary area is limited to a substrate [5].…”

Section: Theoretical Analysis: Reconsideration Of Problem and Derivatmentioning

confidence: 73%

“…Moreover, due to the intensive use of fertilizers, the threat posed to aquatic environments by repeated discharging a certain ratio of drainage is serious enough to warrant regulation by national governments [3][4][5][6]. Since a closed-loop soilless culture system reuses its drainage, the resulting variation in nutrient concentration can significantly affect the plant growth as the reuse period becomes longer [5,[7][8][9]. It is therefore difficult to intuitively explain or interpret nutrient-variation management concentration 15…”

Section: Introductionmentioning

confidence: 99%

Theoretical and Experimental Analysis of Nutrient Variations in Electrical Conductivity-Based Closed-Loop Soilless Culture Systems by Nutrient Replenishment Method

Ahn

Son

2019

Agronomy

View full text Add to dashboard Cite

In closed-loop soilless culture systems, variation in nutrients can lead to instability in the nutrient management and forced discharge of nutrients and water. Total nutrients absorbed by plants are replenished in an electrical conductivity-based closed-loop system, and fluctuation in electrical conductivity within a certain range around the initial value can be expected. However, this is not always observed in systems using conventional nutrient-replenishment methods. The objectives of this study were to analyze nutrient variation in a closed-loop soilless culture system based on a theoretical model and derive an alternative nutrient-replenishment method. The performance of the derived alternative method was compared with a conventional nutrient-replenishment method through simulation analysis. A demonstration experiment using sweet peppers was then conducted to confirm whether the theoretical analysis results can be reproduced through actual cultivation. The average amounts of injected nutrients during the experimental period of four months in the conventional and alternative methods were 2257 and 1054 g, respectively. There was no significant difference in the yield of sweet peppers between the two methods. The substrate electrical conductivity in the alternative method was maintained at 2.7 dS·m −1 ± 0.5 within the target electrical conductivity value, while that in the conventional method gradually increased to 5.0 dS·m −1 ± 1.2. In a simulation study, results similar to the demonstration experiment were predicted. Total nutrient concentrations in the alternative method showed fluctuations around the target value but did not continuously deviate from the target value, while those in the conventional method showed a tendency to increase. As a whole, these characteristics of the alternative method can help in minimizing nutrients and water emissions from the cultivation system. Agronomy 2019, 9, 649 2 of 16techniques, unlike open-loop systems. In order to appropriately apply those techniques, theoretical models are required and the problems should be precisely defined [10].In both closed-loop and open-loop soilless culture systems, the electrical conductivity (EC) of the nutrient solution in the mixing tank is adjusted to a target value before the solution is applied to the plant [9,11,12]. However, unlike an open-loop system, the mixing ratio of tap water to stock solution in a closed-loop system is adjusted by considering the change in nutrient concentration due to the inflow of drainage [12]. Alternatively, in simple systems, a premixed standard nutrient solution of a certain EC is supplied based on the difference between initial and current water levels in the circulation tank, which simultaneously performs drainage collection and nutrient-solution feeding [13,14]. In an EC-based closed-loop soilless culture system, supply of stock solution or standard nutrient solution and tap water is intended to replenish nutrients and water consumed in the system [12]. For a single system in which the plants are grown ...

show abstract

“…Moreover, its versatility and model-free approach allows the modeling of underlying processes without restrictive assumptions [ 23 ]. In plant research, modeling with neural networks has been successfully used in a wide range of applications, including optimizing growth media [ 24 , 25 ], classifying cell wall architecture [ 26 ], identifying diseases [ 27 , 28 ], analyzing biophysical properties [ 29 ], forecasting pH and electrical conductivity [ 30 , 31 ], evaluating post-harvest changes and product quality [ 32 , 33 , 34 ], and characterizing and authenticating plant products [ 35 ]. Because of the unknown extent of spectral interference between the two pigments, an ANN appears to be a convenient method for modeling CMQ coordinates, using chlorophyll and anthocyanin content as input parameters and qL* , qa* , and qb* values as output parameters.…”

Section: Introductionmentioning

confidence: 99%

Predictive Modeling of a Leaf Conceptual Midpoint Quasi-Color (CMQ) Using an Artificial Neural Network

Šimko

2020

Sensors

View full text Add to dashboard Cite

The color of plant leaves is moderated by the content of pigments, which can show considerable dorsiventral distribution. Two typical examples are leafy vegetables and ornamentals, wherein red and green color surfaces can be seen on the same leaf. The proof of concept is provided for predictive modeling of a leaf conceptual mid-point quasi-color (CMQ) from the content of pigments. The CMQ idea is based on the hypothesis that the content of pigments in leaves is associated with the combined color from both surfaces. The CMQ, which is calculated from CIELab color coordinates at adaxial and abaxial antipodes, is thus not an actual color, but a notion that can be used in modeling. The CMQ coordinates, predicted from the content of chlorophylls and anthocyanins by means of an artificial neural network (ANN), matched well with the CMQ coordinates empirically found on photosynthetically active leaves of lettuce (Lactuca sativa L.), but also with other plant species with comparable leaf attributes. Modeled values of lightness (qL*) decreased with the increasing content of both pigments, while the redness or greenness (qa*) and yellowness or blueness (qb*) of the CMQ were affected more by a relative content of chlorophylls and anthocyanins in leaves. The highest vividness of quasi-colors (qC*) was modeled for leaves with a high content of either pigment alone. The model predicted a substantially duller quasi-color for leaves with chlorophylls and anthocyanins present together, particularly when both pigments were present at very high levels.

show abstract

Forecasting Root-Zone Electrical Conductivity of Nutrient Solutions in Closed-Loop Soilless Cultures via a Recurrent Neural Network Using Environmental and Cultivation Information

Cited by 26 publications

References 44 publications

Long short-term memory for a model-free estimation of macronutrient ion concentrations of root-zone in closed-loop soilless cultures

Long short-term memory for a model-free estimation of macronutrient ion concentrations of root-zone in closed-loop soilless cultures

Theoretical and Experimental Analysis of Nutrient Variations in Electrical Conductivity-Based Closed-Loop Soilless Culture Systems by Nutrient Replenishment Method

Predictive Modeling of a Leaf Conceptual Midpoint Quasi-Color (CMQ) Using an Artificial Neural Network

Contact Info

Product

Resources

About