KGML-ag: a modeling framework of knowledge-guided machine learning to simulate agroecosystems: a case study of estimating N<sub>2</sub>O emission using data from mesocosm experiments DOI Creative Commons
Licheng Liu, Shaoming Xu, Jinyun Tang

и другие.

Geoscientific model development, Год журнала: 2022, Номер 15(7), С. 2839 - 2858

Опубликована: Апрель 7, 2022

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

Язык: Английский

Simultaneous quantification of N2, NH3 and N2O emissions from a flooded paddy field under different N fertilization regimes DOI
Longlong Xia, Xiaobo Li, Qianqian Ma

и другие.

Global Change Biology, Год журнала: 2019, Номер 26(4), С. 2292 - 2303

Опубликована: Дек. 13, 2019

Gaseous nitrogen (N) emissions, especially emissions of dinitrogen (N2 ) and ammonia (NH3 ), have long been considered as the major pathways N loss from flooded rice paddies. However, no studies simultaneously evaluated overall response gaseous losses to improved fertilization practices due difficulties directly measure N2 paddy soils. We quantified (using membrane inlet mass spectrometry), NH3 nitrous oxide O) a field in southern China over an entire rice-growing season. Our experiment included three treatments: control treatment (no addition) two fertilizer (220 kg N/ha) application methods, traditional surface incorporation into soil. results show that season, cumulative accounted for 13.5% 19.1% 0.2% 32.8% (total loss) applied fertilizer. Compared with treatment, soil decreased , O by 14.2%, 13.3% 42.5%, respectively. Overall, significantly reduced total 13.8%, use efficiency 14.4%, increased yield 13.9% intensity (gaseous loss/rice yield) 24.3%. indicate is effective agricultural management practice ensuring food security environmental sustainability ecosystems.

Язык: Английский

Процитировано

73

N2O emissions of low input cropping systems as affected by legume and cover crops use DOI

Céline Peyrard,

Bruno Mary,

Pierre Perrin

и другие.

Agriculture Ecosystems & Environment, Год журнала: 2016, Номер 224, С. 145 - 156

Опубликована: Апрель 9, 2016

Язык: Английский

Процитировано

70

Dynamic biochar effects on soil nitrous oxide emissions and underlying microbial processes during the maize growing season DOI
Joseph D. Edwards, Cameron M. Pittelkow, Angela D. Kent

и другие.

Soil Biology and Biochemistry, Год журнала: 2018, Номер 122, С. 81 - 90

Опубликована: Апрель 14, 2018

Язык: Английский

Процитировано

70

Greenhouse gas fluxes from agricultural soils of Kenya and Tanzania DOI Open Access
Todd S. Rosenstock, Mathew Mpanda, David E. Pelster

и другие.

Journal of Geophysical Research Biogeosciences, Год журнала: 2016, Номер 121(6), С. 1568 - 1580

Опубликована: Май 7, 2016

Abstract Knowledge of greenhouse gas (GHG) fluxes in soils is a prerequisite to constrain national, continental, and global GHG budgets. However, data characterizing from agricultural Africa are markedly limited. We measured carbon dioxide (CO 2 ), nitrous oxide (N O), methane (CH 4 ) at 10 farmer‐managed sites six crop types for 1 year Kenya Tanzania using static chambers chromatography. Cumulative emissions ranged between 3.5–15.9 Mg CO ‐C ha −1 yr , 0.4–3.9 kg N O‐N −1.2–10.1 CH depending on type, environmental conditions, management. Manure inputs increased ( p = 0.03), but not O or emissions. Soil cultivation had no discernable effect any the three gases. Fluxes were 54–208% greater < 0.05) during wet versus dry seasons some, all, types. The heterogeneity seasonality suggest that available describing soil Africa, based measurements limited duration only few agroecological zones, inadequate use as basis estimating impact A targeted effort understand magnitude mechanisms underlying African necessary accurately estimate influence this source climate system determining mitigation strategies.

Язык: Английский

Процитировано

67

Soil carbon dioxide and methane fluxes from forests and other land use types in an African tropical montane region DOI Creative Commons
I. Wanyama, David E. Pelster, Klaus Butterbach‐Bahl

и другие.

Biogeochemistry, Год журнала: 2019, Номер 143(2), С. 171 - 190

Опубликована: Фев. 28, 2019

In the last 40 years, large areas of Mau forest, largest contiguous tropical montane forest in East Africa, have been cleared for agriculture. To date, there are no empirical data on how this land use change affects carbon dioxide (CO2) fluxes from soil respiration and methane (CH4) fluxes. This study reports measured annual CO2 CH4 native previously forested lands converted to smallholder grazing land, commercial tea plantations eucalyptus plantations. Fluxes were weekly August 2015 2016 using static chamber method. Grazing had highest (p = 0.028) cumulative respiratory (25.6 ± 2.9 Mg CO2–C ha−1 year−1), whereas lowest observed (5.6 0.5 year−1). Soil positively correlated with pH, but negatively C:N ratio. Annual explained by bulk density interaction between pH Most soils sinks atmospheric across all types. Methane uptake was sites (− 3.08 0.35 − 5.84 0.61 kg CH4–C year−1) 3.43 0.19 Uptake decreased significantly increasing intensity (smallholder plantations: 1.42 0.09 year−1, 1.44 0.29 Soils rates 0.36 0.25 mean water-filled pore space < 0.01) 0.003) inorganic NH4+ concentrations 0.03). can be mainly water content these factors related gas diffusion. Our shows that converting forests managed types Specifically, sink strength reduced less than half forest. also altered emitting 3–4 times more other

Язык: Английский

Процитировано

63

Global Research Alliance N2O chamber methodology guidelines: Recommendations for deployment and accounting for sources of variability DOI Creative Commons
Alice F. Charteris, David R. Chadwick, R. E. Thorman

и другие.

Journal of Environmental Quality, Год журнала: 2020, Номер 49(5), С. 1092 - 1109

Опубликована: Июль 16, 2020

Adequately estimating soil nitrous oxide (N2 O) emissions using static chambers is challenging due to the high spatial variability and episodic nature of these fluxes. We discuss how design experiments better account for this reduce uncertainty N2 O emission estimates. This paper part a series, each discussing different facets chamber methodology. Aspects experimental sampling affected by include site selection layout, size, areal coverage. Where used, treatment application adds further level variability. Time day, frequency, duration (both individual closure overall experiment duration) affect temporal captured. also present best practice recommendations installation protocols uncertainty. To obtain estimates, resources should be allocated minimize in line with objectives. Sometimes will mean prioritizing flux measurements increasing their accuracy precision by, example, collecting four or more headspace samples during closure. However, where fluxes are exceptionally spatially variable (e.g., heterogeneous agricultural landscapes, such as uneven woody grazed pastures), available deploy fewer per may beneficial. Similarly, particularly fluxes, generated example irrigation freeze-thaw cycles, frequency improve temporally interpolated

Язык: Английский

Процитировано

61

Global Research Alliance N2O chamber methodology guidelines: Considerations for automated flux measurement DOI Creative Commons
Peter Grace, Tony J. van der Weerden, David Rowlings

и другие.

Journal of Environmental Quality, Год журнала: 2020, Номер 49(5), С. 1126 - 1140

Опубликована: Июль 17, 2020

Nitrous oxide (N2 O) emissions are highly episodic in response to nitrogen additions and changes soil moisture. Automated gas sampling provides the necessary high temporal frequency capture these emission events real time, ensuring development of accurate N2 O inventories effective mitigation strategies reduce global warming. This paper outlines design operational considerations automated chamber systems including deployment, sampling, options terms analysis samples. The basic hardware software requirements for chambers described, major challenges obstacles their implementation operation a wide range environments. Detailed descriptions provided that have been deployed assess impacts agronomy on other significant greenhouse gases. information will assist researchers across world successful deployment systems.

Язык: Английский

Процитировано

53

A roadmap for sampling and scaling biological nitrogen fixation in terrestrial ecosystems DOI Creative Commons
Fiona M. Soper, Benton N. Taylor, Joy B. Winbourne

и другие.

Methods in Ecology and Evolution, Год журнала: 2021, Номер 12(6), С. 1122 - 1137

Опубликована: Март 2, 2021

Abstract Accurately quantifying rates and patterns of biological nitrogen fixation (BNF) in terrestrial ecosystems is essential to characterize ecological biogeochemical interactions, identify mechanistic controls, improve BNF representation conceptual numerical modelling, forecast limitation constraints on future carbon (C) cycling. While many resources address the technical advantages limitations different methods for measuring BNF, less systematic consideration has been given broader decisions involved planning studies, interpreting data, extrapolating results. Here, we present a practical road map study design, execution, data analysis scaling, outlining key considerations at each step. We issues including defining N‐fixing niches interest, identifying important sources temporal spatial heterogeneity, designing sampling scheme (including method selection, measurement conditions, replication, hotspots hot moments), approaches analysing, scaling reporting BNF. also review comparability estimates derived using literature, provide sample R code simulating symbiotic frames upscaling. Improving standardizing design these stages will accuracy interpretability define limits extrapolation, facilitate use downstream applications. highlight aspects—such as scales statistical dealing with non‐normality, versus significance—that are ripe further development.

Язык: Английский

Процитировано

47

Hot moments drive extreme nitrous oxide and methane emissions from agricultural peatlands DOI
Tyler L. Anthony, Whendee L. Silver

Global Change Biology, Год журнала: 2021, Номер 27(20), С. 5141 - 5153

Опубликована: Июль 14, 2021

Agricultural peatlands are estimated to emit approximately one third of global greenhouse gas (GHG) emissions from croplands, but the temporal dynamics and controls these poorly understood, particularly for nitrous oxide (N2 O). We used cavity ring-down spectroscopy automated chambers in a drained agricultural peatland measure over 70,000 individual N2 O, methane (CH4 ), carbon dioxide (CO2 ) fluxes 3 years. Our results showed that O were high, contributing 26% (annual range: 16%-35%) annual CO2 e emissions. Total averaged 26 ± 0.5 kg O-N ha-1 y-1 exhibited significant inter- intra-annual variability with maximum flux 42 1.8 . Hot moments CH4 represented 1.1 0.2 1.3 0.2% measurements, respectively, contributed 45 1% mean 140 9% fluxes. Soil moisture, soil temperature, bulk oxygen (O2 concentrations strongly correlated emissions; nitrate ( NO3- also significantly These suggest IPCC benchmarks underestimate high emitting by up 70%. Scaling regional similar management suggests ecosystems could 1.86 Tg (range: 1.58-2.21 ). Data large sources GHGs, short-term hot fraction total budgets.

Язык: Английский

Процитировано

47

Evaluating nitrogen cycling in terrestrial biosphere models: a disconnect between the carbon and nitrogen cycles DOI Creative Commons
Sian Kou‐Giesbrecht, Vivek K. Arora, Christian Seiler

и другие.

Earth System Dynamics, Год журнала: 2023, Номер 14(4), С. 767 - 795

Опубликована: Авг. 14, 2023

Abstract. Terrestrial carbon (C) sequestration is limited by nitrogen (N), an empirically established constraint that could intensify under CO2 fertilization and future global change. The terrestrial C sink estimated to currently sequester approximately a third of annual anthropogenic emissions based on ensemble biosphere models, which have been evaluated in their ability reproduce observations the C, water, energy cycles. However, N cycling thus regulation largely unexplored. Here, we evaluate models with coupled C–N performance at simulating cycling, outlining framework for evaluating can be applied across models. We find exhibit significant variability pools fluxes, different magnitudes trends over historical period, despite generally sink. Furthermore, there are no correlations between model nor differences representations fundamental processes. This suggests underlying processes regulate operate differently appear disconnected from cycling. Models tend overestimate tropical biological fixation, vegetation : ratio, soil ratio but underestimate temperate fixation relative observations. uncertainty associated measurements given scarcity (especially those processes) high spatiotemporal variability. Overall, our results suggest represent overestimating storage per unit N, lead biases projections change (let alone without representation cycling). More extensive comparisons against experimental manipulations crucial its impact guide development

Язык: Английский

Процитировано

22