Estimation of CO2 and Carbon Exchange Rate of Strawberry in Greenhouse Cultivation with 1D Convolutional Neural Network and Temperature for Application to Agriculture

Abstract

Estimates of CO2 and Carbon Exchange Rate (CER) is vital tasks for agriculture challenges and hydrological and climate science studies. In the situation that farmers decrease and climate change, applying artificial intelligence techniques to agriculture challenges are desired as innovative approaches to overcome these difficulties in Society.  This study aim to approach following four agriculture challenges (1)limited available data (2) real-time estimate (3) reliable estimate (4) user-friendly, using deep learning models and environment datasets of strawberry in greenhouse cultivation. In this study, using deep learning models, 1D Convolutional Neural Network(1D CNN), Long-Short-Term-Memory(LSTM) and 1D CNN+LSTM, the estimation of CO2 and Carbon Exchange Rate of Strawberry in greenhouse cultivation at one minutes or ten minutes time duration are evaluated. 1D CNN performed the best with R2 0.67 for CO2 in one minute time duration from Temperature, Vapour Pressure Deficit. The applying 1D CNN for estimate CER in ten minutes duration performed with R2 0.77 from Temperature. These results suggest that applying 1D CNN success for estimate CO2 and CER of strawberry in greenhouse cultivation with limited available data and its applicability in agriculture system in society.