Life Cycle Assessment and Estimation of Environmental Pollutants Emission in Sugarcane Production (Saccharum officinarum L.) using Artificial Neural Network

Sugarcane (Saccharum officinarum L.) cultivation as a strategic product with a high sugar content and having side products can play a major role in addressing this need and disconnecting dependence. Life cycle assessment in recent years has become an appropriate tool for assessing environmental impacts in agricultural and food industries. The purpose of this study was to evaluate the environmental impacts of sugarcane production in Imam Khomeini Sugarcane Agro-Industrial Company (IKSAIC) with the life cycle approach and predict the amount of contaminants released according to inputs using artificial neural network.

This research was carried out in IKSAIC as one of the seven companies affiliated to the sugarcane development and related industries of Khuzestan province during 2017-2018. In a life cycle assessment project, all production processes of a product, from the stage of materials extraction to disposal of the remaining waste from the product (from cradle to grave) are reviewed and the results of the reduction of environmental degradation are used.
Each life cycle assessment project has four essential steps as follows:
Goal and scope definition
Life cycle inventory
Environmental impact assessment
Interpretation of

Artificial neural network models were used to predict the environmental contamination of the product due to different inputs used in sugarcane production for plant and ratoon farms.

Assessment of environmental impacts of sugarcane production
In order to evaluate the reduction of environmental pollutants in sugarcane production, the entire life cycle of the product was investigated from primary sources extraction to harvesting in a field at plant and ratoon farms, separately. Ecoinvent databases were used to access needed information and data analysis was done with Simapro software.
In this study, the global warming potential per product in plant and ratoon farms is estimated to be equal to 126.51 and 103.95 kg CO2, respectively, which is due to the burning of sugarcane branches before harvest. The studies also show that agricultural machinery and electricity have the most impact on this sector.
In order to better interpret the results of the other functional unit, it was considered as a unit of land. This means that all indicators were calculated per hectare of production. When the functional unit is calculated on the unit basis, the differences in the products are ignored for the different performance levels in the results.
Assessment modeling of environmental emissions sugarcane production
Based on the results, in modeling for sugarcane production in plant farms, 9-10-5-10 structures with nine inputs, two hidden layer with 10 as well as 5 secret neurons and 10 output parameters were identified as the best structure. For this structure, R2, RMSE and MAPE were calculated. While the R2 coefficient for the human toxicity index and the marine aquatic eco-toxicity were obtained with the lowest values of 0.960, RMSE for these indices were calculated as 0.138 and 0.126 respectively. Furthermore, highest values of R2 coefficient for the acidification and eutrophication index of the lake were resulted about 0.992, whereas RMSE were computed about 0.106 and 0.115) respectively. Also, in modeling for sugarcane production in ratoon farms, the structure of 7-9-6-10 with seven inputs, two hidden layer with 9 and 6 secret neurons, and 10 output parameters are determined as the best structure. For this structure, R2, RMSE and MAPE were calculated and R2 coefficient was obtained with the lowest value of 0.985, while RMSE for this index was 0.164. In addition, R2 for the acidification index was calculated with the highest value of 0.995, whereas the RMSE for this index was calculated to be 0.116.

The results of the evaluation of product life cycle showed that the global warming potential in plant and ratoon farms was estimated to be 126.51 and 103.95 kg CO2, respectively. Results of artificial neural network modeling indicated that the best structure of the neural network to predict the environmental contaminants of sugarcane production, is 9-10-5-10 and 7-9-6-10 for plant farms and ratoon farms respectively.
Acknowledgements
Thanks and appreciation from the Department Agricultural Machinery Engineering, Faculty of Agricultural Engineering and Technology, University of Tehran for the credibility of this research, as well as the officials and engineers of Imam Khomeini Sugarcane Agro-Industrial Company (IKSAIC).