Evaluation Of Nutritional Status Of Aquaponically-Grown Plants Based On Machine Learning

Aquaponics is a rapidly emerging agricultural production system,which recycles effluent from aquaculture to produce plant crops with spent nutrients by creating a symbiotic ecosystem for fish,microbes,and plants in a closed system.Aquaponics has been proposed as a sustainable solution to the current challenges in food production,as it recycles more than 98%of their water from aquaculture effluents,and therefore dramatically reduce the amount of wastewater discharged to the environment.Aquaponics deservedly contributes to solving the global issues of food,water,and energy scarcity.Recent advances in aquaponics revolve around its automation and integration with Industry 4.0 technologies.One of the significant challenges with aquaponics is that the solution doesn’t have enough bioavailable nutrients for optimal plant growth.Therefore,real-time monitoring of aquaponics plants is crucial to determining the required nutritional supplement.Traditionally,the nutritional status of crops was monitored visually,producing misleading results.The estimation of nutrients using chemical analysis is accurate,but it has many limitations,as it is destructive,costly,time-consuming,and labor-intensive.Hence,developing technologies that overcome the limitations mentioned above is vital.Therefore,this study aimed to develop reliable,automated,and rapid algorithms for monitoring and diagnosing the nutritional status of aquaponics-grown plants.Several nutritional systems have been prepared to accomplish the aim of this study,notably the aquaponics system.Data was collected using cutting-edge sensors like spectrophotometers and digital cameras.Deep learning models,machine learning strategies,and several computer vision techniques have also been applied.The initial interest involved using RGB images to diagnose symptoms of nutrient deficiencies.A potent deep learning-based model has been developed to automate the procedure of diagnosing nitrogen(-N),phosphorus(-P),and potassium(-K)deficiency symptoms in aquaponics lettuce.It was compared with some other statistical models to demonstrate its diagnostic capability.The developed deep model excelled with a total accuracy of 96.5%in diagnosing the nutritional status of lettuce(-N,-P,-K,FN)without addressing the level of deficiency.Additionally,it was demonstrated that these techniques might be included in embedded devices to control the aquaponics nutrition cycle.Since nutrients are the main component of plant photosynthesis and the essential structural component of chlorophyll in plants,the estimation of chlorophyll content is a strong indicator of the nutritional status of a plant.The most spectral vegetation indices related to the chlorophyll content of lettuce leaves were extracted to develop an automated model for estimating chlorophyll.These indices were used individually and in combination as inputs for machine learning models to build the estimation model.The green ratio vegetation index(GRVI)outperformed its counterparts in the estimation efficiency of chlorophyll content by obtaining R~2p=0.89 with the random forest(RF)model.The results of the vegetation indices combination were more promising with high predictive efficiency,as the R~2p reached 0.99with the back-propagation neural networks(BPNN)model.This study highlighted the reliability of vegetation indices as estimators of chlorophyll content in plant leaves.Balance in aquaponics necessitates monitoring plant nutritional status from transplantation to maturity.The challenge is to develop a framework based on time-series one-dimensional spectral data for the simultaneous monitoring and diagnosis of plant nutritional status in aquaponics.This objective is accomplished by encoding the spatiotemporal information of the plants into a single spectral time series model to assess the nutrient status of aquaponically-grown lettuce.Briefly,we used long-term memory(LSTM)to classify aquaponically-grown lettuce according to its nutritional status in terms of nitrogen(N),phosphorous(P),and potassium(K)using time-series spectral data.For a more thorough investigation of the classification efficiency of LSTM,three algorithms were used as comparison benchmarks:CNN,Naive Bayes(NB),and the ensemble classifier.An autoencoder was used to select the optimal wavelengths(LSTM inputs).The proposed model achieved promising results for the permanent monitoring of the nutritional status of plants.The LSTM-Autoencoder model had the highest overall classification accuracy of 0.91,0.855,and0.90 for N,P,and K,respectively.The final challenge is the precise quantification of nutrients in aquaponics plants to estimate the required nutritional supplements accurately.Firstly,the presence of disturbances in the plant was detected using colored images without determining the severity of the disorder.Subsequently,the measurement of chlorophyll content,which directly affects the color of plant leaves,was performed.After that,a time-series model was developed to diagnose the nutritional status of plants from transplantation to maturity.Now it is time to estimate the content of nutrients in plant leaves.Machine learning algorithms like partial least squares regression(PLSR),back-propagation neural networks(BPNN),and random forest(RF)were used with spectral data,specifically at selected optimal wavelengths,to develop a reliable quantitative model for the NPK content of lettuce leaves.The optimal wavelengths were used as inputs for the machine-learning models.The models were developed and obtained good and significantly correlated predictive accuracy,with R~2 of 0.97,0.94,and 0.96 for nitrogen,phosphorous,and potassium,respectively.The results demonstrated that the proposed framework provides a path toward automating aquaponics and adopting it as a precision agricultural technology.This study provided a comprehensive explanation of using the latest machine-learning techniques to monitor the nutritional status of plants in aquaponics.Promising results were obtained that contribute to the sustainability of aquaponics and adopt it as a precision agricultural technique,confirming the concept of aquaponics 4.0.More research is required to develop and automate these systems due to their importance as a sustainable food system.