Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Applying Time Series Deep Learning Models to Forecast the Growth of Perennial Ryegrass in Ireland

Oluwadurotimi Onibonoje, Vuong M. Ngo, Andrew McCarre, Elodie Ruelle, Bernadette O-Briend, Mark Roantree

TL;DR

The paper investigates time-series deep learning for forecasting univariate perennial ryegrass growth in Ireland, benchmarking ARIMA, LSTM, GRU, MLP, and particularly Temporal Convolutional Networks (TCN) on a long Cork dataset (1982–2015). It finds that TCN offers the best predictive accuracy with RMSE $2.74$ and MAE $3.46$, while traditional ARIMA remains a competitive baseline in some settings; deeper DL models yield diminishing returns beyond modest complexity. The study highlights the importance of model configuration (layer depth, sequence length) and dataset scale in time-series forecasting for grass growth, and discusses generalization as evidenced by training/validation curves. Practical implications include enabling data-efficient, forecast-driven pasture management to support sustainable dairy farming in Ireland. Future work proposes extending to multivariate forecasting with climate features and leveraging domain knowledge through ontologies and knowledge graphs to improve performance and computational efficiency.

Abstract

Grasslands, constituting the world's second-largest terrestrial carbon sink, play a crucial role in biodiversity and the regulation of the carbon cycle. Currently, the Irish dairy sector, a significant economic contributor, grapples with challenges related to profitability and sustainability. Presently, grass growth forecasting relies on impractical mechanistic models. In response, we propose deep learning models tailored for univariate datasets, presenting cost-effective alternatives. Notably, a temporal convolutional network designed for forecasting Perennial Ryegrass growth in Cork exhibits high performance, leveraging historical grass height data with RMSE of 2.74 and MAE of 3.46. Validation across a comprehensive dataset spanning 1,757 weeks over 34 years provides insights into optimal model configurations. This study enhances our understanding of model behavior, thereby improving reliability in grass growth forecasting and contributing to the advancement of sustainable dairy farming practices.

Applying Time Series Deep Learning Models to Forecast the Growth of Perennial Ryegrass in Ireland

TL;DR

The paper investigates time-series deep learning for forecasting univariate perennial ryegrass growth in Ireland, benchmarking ARIMA, LSTM, GRU, MLP, and particularly Temporal Convolutional Networks (TCN) on a long Cork dataset (1982–2015). It finds that TCN offers the best predictive accuracy with RMSE and MAE , while traditional ARIMA remains a competitive baseline in some settings; deeper DL models yield diminishing returns beyond modest complexity. The study highlights the importance of model configuration (layer depth, sequence length) and dataset scale in time-series forecasting for grass growth, and discusses generalization as evidenced by training/validation curves. Practical implications include enabling data-efficient, forecast-driven pasture management to support sustainable dairy farming in Ireland. Future work proposes extending to multivariate forecasting with climate features and leveraging domain knowledge through ontologies and knowledge graphs to improve performance and computational efficiency.

Abstract

Grasslands, constituting the world's second-largest terrestrial carbon sink, play a crucial role in biodiversity and the regulation of the carbon cycle. Currently, the Irish dairy sector, a significant economic contributor, grapples with challenges related to profitability and sustainability. Presently, grass growth forecasting relies on impractical mechanistic models. In response, we propose deep learning models tailored for univariate datasets, presenting cost-effective alternatives. Notably, a temporal convolutional network designed for forecasting Perennial Ryegrass growth in Cork exhibits high performance, leveraging historical grass height data with RMSE of 2.74 and MAE of 3.46. Validation across a comprehensive dataset spanning 1,757 weeks over 34 years provides insights into optimal model configurations. This study enhances our understanding of model behavior, thereby improving reliability in grass growth forecasting and contributing to the advancement of sustainable dairy farming practices.

Paper Structure

This paper contains 18 sections, 20 equations, 7 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Dataset
  4. Time Series Forecasting Models
  5. AutoRegressive Integrated Moving Averge
  6. Long Short Term Memory
  7. Gated Recurrent Unit
  8. Multi-Layer Perceptron
  9. Temporal Convolutional Networks
  10. Experimental Setup
  11. Dataset Processing
  12. Evaluation Metrics
  13. Hyperparameter Tuning
  14. Results
  15. Effect of Network Layers and Input Sequence Length
  16. ...and 3 more sections

Figures (7)

  • Figure 1: County Cork in the South of Ireland
  • Figure 2: Weekly Grass Height at Moorepark From 1982-2013
  • Figure 3: Weekly Height of Ryegrass
  • Figure 4: Monthly Height of Ryegrass
  • Figure 5: TCN model with 2 layers, a kernel size of 2 and dilations of [1, 2, 4]
  • ...and 2 more figures