Empirical Evaluation of Artificial Intelligence Based Customised Weather Forecasting and Monitoring Model for the Agriculture Sector


Student thesis: Doctoral Thesis


As a result of the evolution of agriculture from 1.0 to 4.0, modern-day agriculture is driven by smart systems and smart devices. The evaluation of the Internet of Things (IoT) and specific approaches in edge computing contributed towards understanding and predicting the weather conditions with a demonstrated impact on precision farming. Non-predictive or inaccurate weather forecasting can severely impact the community of users such as farmers. Farmers depend on the weather forecast so that various farming activities can be undertaken such as ploughing, cultivation, harvesting, and others. An inaccurate forecast directly impacts the farmer’s ability to engage in these activities, negatively influencing their capability of managing the resources related to these operations. In addition, there are significant risks to life and property loss due to unexpected weather conditions. Numerical Weather Prediction (NWP) models run in major weather forecasting centres with several supercomputers to solve simultaneous complex non-linear mathematical equations. This requires considerable computing power to obtain a forecast based on current weather conditions. Such models provide the medium-range weather forecasts, i.e. every 6 hours up to 18 hours with a grid length of 10-20 km. However, a community of users often depend on more detailed short-to-medium-range forecasts with higher resolution regional or local forecasting models. Moreover, the regional or local weather forecasting may not be accurate based on the geographical appearance of location, such as the top of a mountain, land covered by several mountains, the slope of the land, etc. The first part of this research is to determine the competence of using neural networks for weather forecasting. The weather forecasting model is developed by exploring the set of models, namely Multi-Input and Multi-Output (MIMO) and Multi-Input and Single-Output (MISO). The proposed MIMO and MISO models are experimented with the state-of-the-art deep neural network approaches such as Long Short-Term Memory (LSTM) and Temporal Convolutional Networks (TCN), which are based on Artificial Neural Networks (ANN). The accuracy of the proposed model is compared to classic machine learning approaches as well as the well-established Weather Research and Forecasting (WRF) NWP model. The proposed model is also experimented using different deep learning configurations and controls for effective weather forecasting. The second part of this research is to apply similar neural network techniques aimed at developing and evaluating a lightweight, fine-grained and novel weather forecasting model, which consists of one or more local weather stations. The proposed model can be used as an efficient localised weather forecasting tool for the community of users, and it could be run on a standalone personal computer.
Date of Award8 Jul 2020
Original languageEnglish
Awarding Institution
  • Edge Hill University
SupervisorARDHENDU BEHERA (Director of Studies), MARCELLO TROVATI (Supervisor) & ELLA PEREIRA (Supervisor)


  • Long Short-Term Memory
  • Temporal Convolutional Networks
  • Weather Prediction
  • Weather Research and Forecasting model
  • Time-series Data Analysis
  • Neural Networks
  • Precision farming

Cite this