Nitrous Oxide Emission Fluxes in Coffee Plantations during Fertilization: A Case Study in Costa Rica

The main source of N2O emissions is agriculture, and coffee monocultures have become an important part of these emissions. The demand for coffee has increased in the last five decades. Thus, its production in agricultural fields and the excess of fertilizers have increased. This study quantified N2O emissions from different dose applications and types of nitrogen fertilizer in a region of major coffee production in Costa Rica. A specific methodology to measure N2O fluxes from coffee plants was developed using Fourier-transform infrared spectroscopy (FTIR). Measurements were performed in a botanical garden in Germany and plots in Costa Rica, analyzing the behavior of a fertilizer in two varieties of coffee (Catuai and Geisha), and in a field experiment, testing two types of fertilizers (chemical (F1) and physical mixture (F2)) and compost (SA). As a result, the additions of synthetic fertilizer increased the N2O fluxes. F2 showed higher emissions than F1 by up to 90% in the field experiment, and an increase in general emissions occurred after a rain event in the coffee plantation. The weak levels of N2O emissions were caused by a rainfall deficit, maintaining low water content in the soil. Robust research is suggested for the inventories.

Evaluation of LandscapeDNDC Model Predictions of CO2 and N2O Fluxes from an Oak Forest in SE England

Process-based biogeochemical models are valuable tools to evaluate impacts of environmental or management changes on the greenhouse gas (GHG) balance of forest ecosystems. We evaluated LandscapeDNDC, a process-based model developed to simulate carbon (C), nitrogen (N) and water cycling at ecosystem and regional scales, against eddy covariance and soil chamber measurements of CO2 and N2O fluxes in an 80-year-old deciduous oak forest. We compared two LandscapeDNDC vegetation modules: PSIM (Physiological Simulation Model), which includes the understorey explicitly, and PnET (Photosynthesis–Evapotranspiration Model), which does not. Species parameters for both modules were adjusted to match local measurements. LandscapeDNDC was able to reproduce daily micro-climatic conditions, which serve as input for the vegetation modules. The PSIM and PnET modules reproduced mean annual net CO2 uptake to within 1% and 15% of the measured values by balancing gains and losses in seasonal patterns with respect to measurements, although inter-annual variations were not well reproduced. The PSIM module indicated that the understorey contributed up to 21% to CO2 fluxes. Mean annual soil CO2 fluxes were underestimated by 32% using PnET and overestimated by 26% with PSIM; both modules simulated annual soil N2O fluxes within the measured range but with less interannual variation. Including stand structure information improved the model, but further improvements are required for the model to predict forest GHG balances and their inter-annual variability following climatic or management changes.

KGML-ag: A Modeling Framework of Knowledge-Guided Machine Learning to Simulate Agroecosystems: A Case Study of Estimating N₂O Emission using Data from Mesocosm Experiments

Agricultural nitrous oxide (N2O) emission accounts for a non-trivial fraction of global greenhouse gases (GHGs) budget. To date, estimating N2O fluxes from cropland remains a challenging task because the related microbial processes (e.g., nitrification and denitrification) are controlled by complex interactions among climate, soil, plant and human activities. Existing approaches such as process-based (PB) models have well-known limitations due to insufficient representations of the processes or constraints of model parameters, and to leverage recent advances in machine learning (ML) new method is needed to unlock the “black box” to overcome its limitations due to low interpretability, out-of-sample failure and massive data demand. In this study, we developed a first of its kind knowledge-guided machine learning model for agroecosystems (KGML-ag), by incorporating biogeophysical/chemical domain knowledge from an advanced PB model, ecosys, and tested it by simulating daily N2O fluxes with real observed data from mesocosm experiments. The Gated Recurrent Unit (GRU) was used as the basis to build the model structure. To optimize the model performance, we have investigated a range of ideas, including: 1) Using initials of intermediate variables (IMVs) instead of time series as model input to reduce data demand; 2) Building hierarchical structures to explicitly estimate IMVs for further N2O prediction; 3) Using multitask learning to balance the simultaneous training on multiple variables; and 4) Pretraining with millions of synthetic data generated from ecosys and fine tuning with mesocosm observations. Six other pure ML models were developed using the same mesocosm data to serve as the benchmark for the KGML-ag model. Results show that KGML-ag did an excellent job in reproducing the mesocosm N2O fluxes (overall r2 = 0.81, and RMSE = 3.6 mg N m−2 day−1 from cross-validation). Importantly KGML-ag always outperforms the PB model and ML models in predicting N2O fluxes, especially for complex temporal dynamics and emission peaks. Besides, KGML-ag goes beyond the pure ML models by providing more interpretable predictions as well as pinpointing desired new knowledge and data to further empower the current KGML-ag. We believe the KGML-ag development in this study will stimulate a new body of research on interpretable ML for biogeochemistry and other related geoscience processes.