Search for: All records

Abstract The response of highly productive croplands at northern mid-latitudes to climate change is a primary source of uncertainty in the global carbon cycle, and a concern for future food production. We present a decadal time series (2007 to 2019) of hourly CO 2 concentration measured at a very tall tower in the United States Corn Belt. Analyses of this record, with other long-term data in the region, reveal that warming has had a positive impact on net CO 2 uptake during the early crop growth stage, but has reduced net CO 2 uptake in both croplands and natural ecosystems during the peak growing season. Future increase in summer temperature is projected to reduce annual CO 2 sequestration in the Corn Belt by 10–20%. These findings highlight the dynamic control of warming on cropland CO 2 exchange and crop yields and challenge the paradigm that warming will continue to favor CO 2 sequestration in northern mid-latitude ecosystems.

Abstract. Agricultural nitrous oxide (N2O) emission accounts for a non-trivialfraction of global greenhouse gas (GHG) budget. To date, estimatingN2O fluxes from cropland remains a challenging task because the relatedmicrobial processes (e.g., nitrification and denitrification) are controlledby complex interactions among climate, soil, plant and human activities.Existing approaches such as process-based (PB) models have well-knownlimitations due to insufficient representations of the processes oruncertainties of model parameters, and due to leverage recent advances inmachine learning (ML) a new method is needed to unlock the “black box” toovercome its limitations such as low interpretability, out-of-sample failureand massive data demand. In this study, we developed a first-of-its-kindknowledge-guided machine learning model for agroecosystems (KGML-ag) byincorporating biogeophysical and chemical domain knowledge from an advanced PBmodel, ecosys, and tested it by comparing simulating daily N2O fluxes withreal observed data from mesocosm experiments. The gated recurrent unit (GRU)was used as the basis to build the model structure. To optimize the modelperformance, we have investigated a range of ideas, including (1) usinginitial values of intermediate variables (IMVs) instead of time series asmodel input to reduce data demand; (2) building hierarchical structures toexplicitly estimate IMVs for further N2O prediction; (3) using multi-tasklearning to balance the simultaneous training on multiple variables; andmore »(4)pre-training with millions of synthetic data generated from ecosys and fine-tuningwith mesocosm observations. Six other pure ML models were developed usingthe 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 mesocosmN2O fluxes (overall r2=0.81, and RMSE=3.6 mgNm-2d-1from cross validation). Importantly, KGML-ag always outperformsthe PB model and ML models in predicting N2O fluxes, especially forcomplex temporal dynamics and emission peaks. Besides, KGML-ag goes beyondthe pure ML models by providing more interpretable predictions as well aspinpointing desired new knowledge and data to further empower the currentKGML-ag. We believe the KGML-ag development in this study will stimulate anew body of research on interpretable ML for biogeochemistry and otherrelated geoscience processes.« less

Abstract. We apply airborne measurements across three seasons(summer, winter and spring 2017–2018) in a multi-inversion framework toquantify methane emissions from the US Corn Belt and Upper Midwest, a keyagricultural and wetland source region. Combing our seasonal results withprior fall values we find that wetlands are the largest regional methanesource (32 %, 20 [16–23] Gg/d), while livestock (enteric/manure; 25 %,15 [14–17] Gg/d) are the largest anthropogenic source. Naturalgas/petroleum, waste/landfills, and coal mines collectively make up theremainder. Optimized fluxes improve model agreement with independentdatasets within and beyond the study timeframe. Inversions reveal coherentand seasonally dependent spatial errors in the WetCHARTs ensemble meanwetland emissions, with an underestimate for the Prairie Pothole region butan overestimate for Great Lakes coastal wetlands. Wetland extent andemission temperature dependence have the largest influence on predictionaccuracy; better representation of coupled soil temperature–hydrologyeffects is therefore needed. Our optimized regional livestock emissionsagree well with the Gridded EPA estimates during spring (to within 7 %) butare ∼ 25 % higher during summer and winter. Spatial analysisfurther shows good top-down and bottom-up agreement for beef facilities (withmainly enteric emissions) but larger (∼ 30 %) seasonaldiscrepancies for dairies and hog farms (with > 40 % manureemissions). Findings thus support bottom-up enteric emission estimates butsuggest errors for manure; we propose that the latter reflects inadequatetreatment of managementmore »factors including field application. Overall, ourresults confirm the importance of intensive animal agriculture for regionalmethane emissions, implying substantial mitigation opportunities throughimproved management.« less