Applications of artificial intelligence and deep learning for sustainable water management in Prince Edward Island

Precision agriculture evaluates and quantifies the input needs of crops for their optimum yield and sustainable production. Growth of potato plants is highly sensitive to drought conditions, which drastically reduce tuber yield if precision supplemental irrigation (SI) is not provided. The hypothesis of this study, that the rainfall in Prince Edward Island is not enough for sustainable potato production in the island, was tested under three specific objectives including i) to model evapotranspiration with artificial intelligence for precision water resource management, ii) to determine the effects of different irrigation systems\n(sprinkler, drip, fertigation and control; rainfed) on potato tuber yield, quality, payout returns, and iii) to model the groundwater levels of Prince Edward Island using deep learning methods to ensure sustainability of water balance in Prince Edward Island.\n\nThis study used deep learning, artificial neural networks (ANNs) and the standard hydrology models to estimate components of water cycle for their use and impact on potato production in Prince Edward Island. Reference evapotranspiration was estimated with recurrent neural networks (RNNs) namely long short term memory (LSTM) and Bidirectional LSTM. Four representative meteorological sites (North Cape, Summerside, Harrington and Saint Peters) were selected across the island. Crop specific evapotranspiration (ETc) was calculated from reference evapotranspiration (ETO) using Penman Monteith equation, FAO-56 method, ANNs, and RNNs, and LSTMs. Based on subset regression analysis, the highest contributing climatic variables namely maximum air temperature and relative humidity were selected as input variables for RNNs’ training (2011-2015) and testing (2016-2017) runs. The results suggested that the LSTM and idirectional LSTM are suitable methods to accurately (R2 > 0.90) estimate ETO for all sites except for Harrington. No major differences were observed in the accuracy of LSTM and Bidirectional LSTM. The potential gap between ETO and rainfall were highlighted for\nassessing agriculture sustainability in Prince Edward Island. Analyses of the data highlighted that the cumulative ETO surpassed the cumulative rainfall potentially affecting yield of major crops in the island. Therefore, agriculture sustainability requires viable options such as SI to replenish the crop water requirements as and when needed. Results suggested that July, August, and September are relatively drier months of the study years and SI may be required to meet the crop water requirements.\n\nIn order to evaluate impact of SI, pressurized irrigation systems including sprinkler, fertigation and drip irrigation were installed at small-scale to offset deficit in soil moisture as compared to conventional practice of rainfed conditions, i.e., no irrigation practice (control). Significant differences in potato yield were observed between control and irrigation methods used in this study. A two-way ANOVA was run to examine the effect of irrigation methods and year on potato tuber yield, water productivity, tuber quality, and payout. In term of payout returns the sprinkler treatment performed significantly better than control, drip, and fertigation in 2018. However, in terms of water productivity, the fertigation treatment performed significantly better than the control and sprinkler treatments during both growing seasons. The lower water productivity of sprinkler irrigation was due to higher water consumption in comparison with drip and fertigation systems.\n\nNeeds of SI for potato production in Prince Edward Island can be met from groundwater pumping. This necessitates the budgeting of water cycle components for efficient management of water resources. In areas where groundwater pumping is common for SI or for domestic use, the inventory control of groundwater resources could become more convenient with the use of deep learning, ANNs, and RNNs namely a multilayer perceptron (MLP) and LSTM. The analysis of two watersheds namely Baltic and Long creep showed that the deep learning methods used in this study are accurate to simulate groundwater levels. Input variables for this watershed-scale modelling investigation included stream level, streamflow, precipitation, relative humidity, mean temperature, heat degree days, dew point temperature, and ETo. Using a hit and trial approach and various hyperparameters, all ANNs were trained from scratch (2011–2015) and validated (2016–2017). The stream level was the major contributor to GWL fluctuation for the Baltic River and Long Creek watersheds (R2 = 0.508 and 0.491 respectively). The MLP performed better in validation for Baltic River and Long Creek watersheds (RMSE = 0.471 and 1.15,\nrespectively). The deep learning techniques introduced in this study to estimate GWL fluctuations are convenient and accurate as compared to collection of periodic dips based on the groundwater monitoring wells for groundwater inventory control and management.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.