Location: Hydrology and Remote Sensing Laboratory
Project Number: 8042-66000-002-006-S
Project Type: Non-Assistance Cooperative Agreement
Start Date: Aug 1, 2021
End Date: Jul 31, 2025
This project aims to quantify environmental impacts (i.e., global warming, eutrophication, and acidification) of agricultural systems with winter cover crops (WCC). The impacts of WCC on reducing agricultural releases are not fully understood. The performances of WCC are often spatially heterogeneous and climate dependent. Life Cycle Assessment (LCA) is a systematic research approach for quantifying environmental impacts of agricultural systems and identifying environmentally preferred farming practices. However, spatially explicit and climate predictive LCAs of agricultural systems with winter cover crops are lacking. To addresses this important knowledge gap, the project will integrate spatially explicit life cycle assessment, machine learning and SWAT modeling approaches to quantify environmental impacts of the WCC systems under future climate. Particularly, the project will focus on the WCC systems in Tuckahoe and Greensboro watersheds. This assessment is critical for growers, producers, agribusiness managers, and policy makers to quantify the environmental impacts of WCC systems under various climate scenarios and to design environmentally friendly and climate resilient management practices. Objective 1. Quantify spatially explicit life cycle environmental impacts of the WCC systems at farm scale. Objective 2. Predict life cycle environmental impacts of the WCC systems under future climate.
To achieve Objective 1, we will utilize an integrative modeling approach developed by The Research Foundation of the State University of New York (RFSUNY). This approach features a novel combination of spatially explicit life cycle assessment, SWAT modeling and boosted regression tree methods. Despite being complex, this approach presents three unique merits including capabilities of 1) simultaneously capturing spatially explicit on-farm and supply-chain environmental releases, 2) determining the top influential factors among weather, soil and farming practice, and 3) assessing environmental impacts at multiple spatial scales. We will determine the spatially explicit on-farm environmental impacts of the WCC systems based on the SWAT modeling. Subsequently, spatially explicit on-farm releases will be fed into a process-based LCA model, where regionalized supply chain emissions are computed and combined to generate the overall life cycle releases of WCC systems in Tuckahoe and Greensboro watersheds. Furthermore, the relative importance of weather, soil and farming practices on life cycle releases will be ranked by using boosted regression tree method. Additionally, model uncertainty will be assessed by a Monte Carlo assessment approach, as demonstrated in the RFSUNY's previous work. To achieve Objective 2, we will build, train and deploy semi-supervised sequence learning models to predict life cycle environmental impacts of the WCC systems under future climate. First, we will build and compare semi-supervised sequence learning models like vanilla recurrent neural network, long short-term memory and gated recurrent units to detect all hidden relationships between the features of the in-house datasets. Then, the best performing machine learning model will be used to efficiently predict the life cycle releases of the WCC systems under future climate conditions. To capture the influences of uncertain climate, the future life cycle releases will be assessed under four representative scenarios identified by the Intergovernmental Panel on Climate Change, including Representative Concentration Pathway scenarios (RCP) 2.6. 4.5, 6.0, and 8.5. The temperature, precipitation and relative humidity datasets up until the Mid-21st century under the RCP scenarios will be used as model inputs for prediction. The above effort will provide efficient and accurate models for estimating life cycle environmental impacts in future years, serving as scientific basis for planning mitigation efforts.