scholarly journals A Mass- and Energy-Conserving Framework for Using Machine Learning to Speed Computations

2020 ◽  
Author(s):  
Patrick Obin Sturm ◽  
Anthony S. Wexler
Keyword(s):  
Machine Learning ◽  
Land Surface ◽  
Climate Models ◽  
Property Values ◽  
Model Performance ◽  
Chemical Properties ◽  
Atmospheric Composition ◽  
Mathematical Framework ◽  
Environmental Models ◽  
Computational Performance

Abstract. Large air quality models and large climate models simulate the physical and chemical properties of the ocean, land surface and/or atmosphere to predict atmospheric composition, energy balance, and the future of our planet. All of these models employ some form of operator splitting, also called the method of fractional steps, in their structure, which enables each physical or chemical process to be simulated in a separate operator or module within the overall model. In this structure, each of the modules calculates property changes for a fixed period of time; that is, property values are passed into the module which calculates how they change for a period of time and then returns the new property values, all in round robin between the various modules of the model. Some of these modules require the vast majority of the computer resources consumed by the entire model so increasing their computational efficiency can either improve the model's computational performance or enable more realistic physical or chemical representations in the module, or a combination of these two. Recent efforts have attempted to replace these modules with ones that use machine learning tools to memorize the input-output relationships of the most time-consuming modules. One shortcoming of some of the original modules and their machine learned replacements is lack of adherence to conservation principles that are essential to model performance. In this work, we derive a mathematical framework for machine learned replacements that conserves properties, say mass, atoms, or energy, to machine precision. This framework can be used to develop machine learned operator replacements in environmental models.

scholarly journals A mass- and energy-conserving framework for using machine learning to speed computations: a photochemistry example

2020 ◽  
Vol 13 (9) ◽  
pp. 4435-4442
Author(s):  
Patrick Obin Sturm ◽  
Anthony S. Wexler
Keyword(s):  
Machine Learning ◽  
Land Surface ◽  
Climate Models ◽  
Property Values ◽  
Model Performance ◽  
Chemical Properties ◽  
Atmospheric Composition ◽  
Mathematical Framework ◽  
Environmental Models ◽  
Computational Performance

Abstract. Large air quality models and large climate models simulate the physical and chemical properties of the ocean, land surface, and/or atmosphere to predict atmospheric composition, energy balance and the future of our planet. All of these models employ some form of operator splitting, also called the method of fractional steps, in their structure, which enables each physical or chemical process to be simulated in a separate operator or module within the overall model. In this structure, each of the modules calculates property changes for a fixed period of time; that is, property values are passed into the module, which calculates how they change for a period of time and then returns the new property values, all in round-robin between the various modules of the model. Some of these modules require the vast majority of the computer resources consumed by the entire model, so increasing their computational efficiency can either improve the model's computational performance, enable more realistic physical or chemical representations in the module, or a combination of these two. Recent efforts have attempted to replace these modules with ones that use machine learning tools to memorize the input–output relationships of the most time-consuming modules. One shortcoming of some of the original modules and their machine-learned replacements is lack of adherence to conservation principles that are essential to model performance. In this work, we derive a mathematical framework for machine-learned replacements that conserves properties – say mass, atoms, or energy – to machine precision. This framework can be used to develop machine-learned operator replacements in environmental models.

scholarly journals A Review of Machine Learning Applications in Land Surface Modeling

Earth ◽  
2021 ◽  
Vol 2 (1) ◽  
pp. 174-190
Author(s):  
Sujan Pal ◽  
Prateek Sharma
Keyword(s):  
Machine Learning ◽  
Neural Networks ◽  
Land Surface ◽  
Climate Models ◽  
Short Term Memory ◽  
Lower Boundary ◽  
Model Performance ◽  
Heat Fluxes ◽  
Momentum Exchange ◽  
High Resolution Data

Machine learning (ML), as an artificial intelligence tool, has acquired significant progress in data-driven research in Earth sciences. Land Surface Models (LSMs) are important components of the climate models, which help to capture the water, energy, and momentum exchange between the land surface and the atmosphere, providing lower boundary conditions to the atmospheric models. The objectives of this review paper are to highlight the areas of improvement in land modeling using ML and discuss the crucial ML techniques in detail. Literature searches were conducted using the relevant key words to obtain an extensive list of articles. The bibliographic lists of these articles were also considered. To date, ML-based techniques have been able to upgrade the performance of LSMs and reduce uncertainties by improving evapotranspiration and heat fluxes estimation, parameter optimization, better crop yield prediction, and model benchmarking. Widely used ML techniques used for these purposes include Artificial Neural Networks and Random Forests. We conclude that further improvements in land modeling are possible in terms of high-resolution data preparation, parameter calibration, uncertainty reduction, efficient model performance, and data assimilation using ML. In addition to the traditional techniques, convolutional neural networks, long short-term memory, and other deep learning methods can be implemented.

scholarly journals Integration of nitrogen dynamics into the Noah-MP land model v1.1 for climate and environmental predictions

2015 ◽  
Vol 8 (5) ◽  
pp. 4113-4153 ◽  
Author(s):  
X. Cai ◽  
Z.-L. Yang ◽  
J. B. Fisher ◽  
X. Zhang ◽  
M. Barlage ◽  
...  
Keyword(s):  
Land Surface ◽  
Climate Models ◽  
Net Primary Productivity ◽  
Assessment Tool ◽  
Land Surface Model ◽  
Coupled Model ◽  
Nitrogen Dynamics ◽  
Surface Model ◽  
Meteorological Model ◽  
Environmental Models

Abstract. Climate and terrestrial biosphere models consider nitrogen an important factor in limiting plant carbon uptake, while operational environmental models view nitrogen as the leading pollutant causing eutrophication in water bodies. The community Noah land surface model with multi-parameterization options (Noah-MP) is unique in that it is the next generation land surface model for the Weather Research and Forecasting meteorological model and for the operational weather/climate models in the National Centers for Environmental Prediction. In this study, we add capability to Noah-MP to simulate nitrogen dynamics by coupling the Fixation and Uptake of Nitrogen (FUN) plant model and the Soil and Water Assessment Tool (SWAT) soil nitrogen dynamics. This incorporates FUN's state-of-the-art concept of carbon cost theory and SWAT's strength in representing the impacts of agricultural management on the nitrogen cycle. Parameterizations for direct root and mycorrhizal-associated nitrogen uptake, leaf retranslocation, and symbiotic biological nitrogen fixation are employed from FUN, while parameterizations for nitrogen mineralization, nitrification, immobilization, volatilization, atmospheric deposition, and leaching are based on SWAT. The coupled model is then evaluated at the Kellogg Biological Station – a Long-term Ecological Research site within the U.S. Corn Belt. Results show that the model performs well in capturing the major nitrogen state/flux variables (e.g., soil nitrate and nitrate leaching). Furthermore, the addition of nitrogen dynamics improves the modeling of the carbon and water cycles (e.g., net primary productivity and evapotranspiration). The model improvement is expected to advance the capability of Noah-MP to simultaneously predict weather and water quality in fully coupled Earth system models.

scholarly journals Development of a Unified Land Model for Prediction of Surface Hydrology and Land–Atmosphere Interactions

2011 ◽  
Vol 12 (6) ◽  
pp. 1299-1320 ◽  
Author(s):  
Ben Livneh ◽  
Pedro J. Restrepo ◽  
Dennis P. Lettenmaier
Keyword(s):  
Soil Moisture ◽  
Land Surface ◽  
Climate Models ◽  
A Priori ◽  
Model Performance ◽  
Surface Flux ◽  
Surface Fluxes ◽  
Frozen Soil ◽  
Surface Model ◽  
Soil Moisture Accounting

Abstract A unified land model (ULM) is described that combines the surface flux parameterizations in the Noah land surface model (used in most of NOAA’s coupled weather and climate models) with the Sacramento Soil Moisture Accounting model (Sac; used for hydrologic prediction within the National Weather Service). The motivation was to develop a model that has a history of strong hydrologic performance while having the ability to be run in the coupled land–atmosphere environment. ULM takes the vegetation, snow model, frozen soil, and evapotranspiration schemes from Noah and merges them with the soil moisture accounting scheme from Sac. ULM surface fluxes, soil moisture, and streamflow simulations were evaluated through comparisons with observations from the Ameriflux (surface flux), Illinois Climate Network (soil moisture), and Model Parameter Estimation Experiment (MOPEX; streamflow) datasets. Initially, a priori parameters from Sac and Noah were used, which resulted in ULM surface flux simulations that were comparable to those produced by Noah (Sac does not predict surface energy fluxes). ULM with the a priori parameters had streamflow simulation skill that was generally similar to Sac’s, although it was slightly better (worse) for wetter (more arid) basins. ULM model performance using a set of parameters identified via a Monte Carlo search procedure lead to substantial improvements relative to the a priori parameters. A scheme for transfer of parameters from streamflow simulations to nearby flux and soil moisture measurement points was also evaluated; this approach did not yield conclusive improvements relative to the a priori parameters.

One stomatal model to rule them all? Evaluating competing hypotheses to regulate the exchange of carbon and water against experimental data

2020 ◽  
Author(s):  
Manon Sabot ◽  
Martin De Kauwe ◽  
Belinda Medlyn ◽  
Andy Pitman
Keyword(s):  
Stomatal Conductance ◽  
Land Surface ◽  
Model Comparison ◽  
Climate Models ◽  
Global Climate ◽  
Model Performance ◽  
Vapour Pressure Deficit ◽  
Global Climate Models ◽  
Model Complexity ◽  
Coupled Climate Models

<p>Nearly 2/3 of the annual global evapotranspiration (ET) over land arises from the vegetation. Yet, coupled-climate models only attribute between 22% – 58% of the annual terrestrial ET to plants. In coupled-climate models, the exchange of carbon and water between the terrestrial biosphere and the atmosphere is simulated by land-surface models (LSMs). Within those LSMs, stomatal conductance (g<sub>s</sub>) models allow plants to regulate their transpiration and carbon uptake, but most are empirically linked to climate, soil moisture availabilty, and CO<sub>2</sub>. Therefore, how and which g<sub>s</sub> schemes are implemented within LSMs is a key source of model uncertainty. This uncertainty has led to considerable investment in theory development in the recent years; multiple alternative hypotheses of optimal leaf-level regulation of gas exchange have been proposed as solutions to reduce existing model biases. However, a systematic inter-model evaluation is lacking (i.e. inter-model comparison within a single framework is needed to understand how different mechanistic assumptions across these new g<sub>s</sub> models affect plant behaviour). Here, we asked how, and under what conditions, nine novel optimal g<sub>s</sub> models differ from one another. The models were trained to match under average conditions before being subjected to: (i) a dry-down, (ii) high vapour pressure deficit, and (iii) elevated CO<sub>2</sub>. These experiments allowed us to identify the models’ specific responses and sensitivities. To further assess whether the models’ responses were realistic, we tested them against photosynthetic and hydraulic field data measured along mesic-xeric gradients in Europe and Australia. Finally, we evaluated model performance versus model complexity and the amount of information taken in by each model, which enables us to make recommendations regarding the use of stomatal conductance schemes in global climate models.</p>

scholarly journals Integrated approaches to climate–crop modelling: needs and challenges

2005 ◽  
Vol 360 (1463) ◽  
pp. 2049-2065 ◽  
Author(s):  
Richard A. Betts
Keyword(s):  
Climate Change ◽  
Atmospheric Chemistry ◽  
Land Surface ◽  
Radiative Forcing ◽  
Crop Production ◽  
Large Scale ◽  
Climate Models ◽  
Atmospheric Composition ◽  
Short Time Scale ◽  
Crop Models

This paper discusses the need for a more integrated approach to modelling changes in climate and crops, and some of the challenges posed by this. While changes in atmospheric composition are expected to exert an increasing radiative forcing of climate change leading to further warming of global mean temperatures and shifts in precipitation patterns, these are not the only climatic processes which may influence crop production. Changes in the physical characteristics of the land cover may also affect climate; these may arise directly from land use activities and may also result from the large-scale responses of crops to seasonal, interannual and decadal changes in the atmospheric state. Climate models used to drive crop models may, therefore, need to consider changes in the land surface, either as imposed boundary conditions or as feedbacks from an interactive climate–vegetation model. Crops may also respond directly to changes in atmospheric composition, such as the concentrations of carbon dioxide (CO 2 ), ozone (O 3 ) and compounds of sulphur and nitrogen, so crop models should consider these processes as well as climate change. Changes in these, and the responses of the crops, may be intimately linked with meteorological processes so crop and climate models should consider synergies between climate and atmospheric chemistry. Some crop responses may occur at scales too small to significantly influence meteorology, so may not need to be included as feedbacks within climate models. However, the volume of data required to drive the appropriate crop models may be very large, especially if short-time-scale variability is important. Implementation of crop models within climate models would minimize the need to transfer large quantities of data between separate modelling systems. It should also be noted that crop responses to climate change may interact with other impacts of climate change, such as hydrological changes. For example, the availability of water for irrigation may be affected by changes in runoff as a direct consequence of climate change, and may also be affected by climate-related changes in demand for water for other uses. It is, therefore, necessary to consider the interactions between the responses of several impacts sectors to climate change. Overall, there is a strong case for a much closer coupling between models of climate, crops and hydrology, but this in itself poses challenges arising from issues of scale and errors in the models. A strategy is proposed whereby the pursuit of a fully coupled climate–chemistry–crop–hydrology model is paralleled by continued use of separate climate and land surface models but with a focus on consistency between the models.

scholarly journals Analysis and Reduction of Systematic Errors through a Seamless Approach to Modeling Weather and Climate

2010 ◽  
Vol 23 (22) ◽  
pp. 5933-5957 ◽  
Author(s):  
G. M. Martin ◽  
S. F. Milton ◽  
C. A. Senior ◽  
M. E. Brooks ◽  
S. Ineson ◽  
...  
Keyword(s):  
Time Scales ◽  
Land Surface ◽  
Large Scale ◽  
Climate Models ◽  
Spatial Scales ◽  
Full Range ◽  
Model Development ◽  
Model Performance ◽  
Systematic Errors ◽  
Wide Range

Abstract The reduction of systematic errors is a continuing challenge for model development. Feedbacks and compensating errors in climate models often make finding the source of a systematic error difficult. In this paper, it is shown how model development can benefit from the use of the same model across a range of temporal and spatial scales. Two particular systematic errors are examined: tropical circulation and precipitation distribution, and summer land surface temperature and moisture biases over Northern Hemisphere continental regions. Each of these errors affects the model performance on time scales ranging from a few days to several decades. In both cases, the characteristics of the long-time-scale errors are found to develop during the first few days of simulation, before any large-scale feedbacks have taken place. The ability to compare the model diagnostics from the first few days of a forecast, initialized from a realistic atmospheric state, directly with observations has allowed physical deficiencies in the physical parameterizations to be identified that, when corrected, lead to improvements across the full range of time scales. This study highlights the benefits of a seamless prediction system across a wide range of time scales.

scholarly journals Estimating Potential Evaporation from Vegetated Surfaces for Water Management Impact Assessments Using Climate Model Output

2011 ◽  
Vol 12 (5) ◽  
pp. 1127-1136 ◽  
Author(s):  
Victoria A. Bell ◽  
Nicola Gedney ◽  
Alison L. Kay ◽  
Roderick N. B. Smith ◽  
Richard G. Jones ◽  
...  
Keyword(s):  
Climate Change ◽  
Water Management ◽  
Land Surface ◽  
Climate Models ◽  
Climate Model ◽  
Atmospheric Composition ◽  
Hydrological Models ◽  
Potential Evaporation ◽  
Land Surface Scheme ◽  
Surface Scheme

Abstract River basin managers concerned with maintaining water supplies and mitigating flood risk in the face of climate change are taking outputs from climate models and using them in hydrological models for assessment purposes. While precipitation is the main output used, evaporation is attracting increasing attention because of its significance to the water balance of river basins. Climate models provide estimates of actual evaporation that are consistent with their simplified land surface schemes but do not naturally provide the estimates of potential evaporation (PE) commonly required as input to hydrological models. There are clear advantages in using PE estimates controlled by atmospheric forcings when using stand-alone hydrological models with integral soil-moisture accounting schemes. The atmosphere–land decoupling approximation that PE provides can prove to be of further benefit if it is possible to account for the effect of different, or changing, land cover on PE outside of the climate model. The methods explored here estimate Penman–Monteith PE from vegetated surfaces using outputs from climate models that have an embedded land surface scheme. The land surface scheme enables an examination of the dependence of canopy stomatal resistance on atmospheric composition, and the sensitivity of PE estimates to the choice of canopy resistance values under current and changing climates is demonstrated. The conclusions have practical value for climate change impact studies relating to flood, drought, and water management applications.

scholarly journals Integration of nitrogen dynamics into the Noah-MP land surface model v1.1 for climate and environmental predictions

2016 ◽  
Vol 9 (1) ◽  
pp. 1-15 ◽  
Author(s):  
X. Cai ◽  
Z.-L. Yang ◽  
J. B. Fisher ◽  
X. Zhang ◽  
M. Barlage ◽  
...  
Keyword(s):  
Land Surface ◽  
Climate Models ◽  
Net Primary Productivity ◽  
Assessment Tool ◽  
Land Surface Model ◽  
Coupled Model ◽  
Nitrogen Dynamics ◽  
Surface Model ◽  
Meteorological Model ◽  
Environmental Models

Abstract. Climate and terrestrial biosphere models consider nitrogen an important factor in limiting plant carbon uptake, while operational environmental models view nitrogen as the leading pollutant causing eutrophication in water bodies. The community Noah land surface model with multi-parameterization options (Noah-MP) is unique in that it is the next-generation land surface model for the Weather Research and Forecasting meteorological model and for the operational weather/climate models in the National Centers for Environmental Prediction. In this study, we add a capability to Noah-MP to simulate nitrogen dynamics by coupling the Fixation and Uptake of Nitrogen (FUN) plant model and the Soil and Water Assessment Tool (SWAT) soil nitrogen dynamics. This model development incorporates FUN's state-of-the-art concept of carbon cost theory and SWAT's strength in representing the impacts of agricultural management on the nitrogen cycle. Parameterizations for direct root and mycorrhizal-associated nitrogen uptake, leaf retranslocation, and symbiotic biological nitrogen fixation are employed from FUN, while parameterizations for nitrogen mineralization, nitrification, immobilization, volatilization, atmospheric deposition, and leaching are based on SWAT. The coupled model is then evaluated at the Kellogg Biological Station – a Long Term Ecological Research site within the US Corn Belt. Results show that the model performs well in capturing the major nitrogen state/flux variables (e.g., soil nitrate and nitrate leaching). Furthermore, the addition of nitrogen dynamics improves the modeling of net primary productivity and evapotranspiration. The model improvement is expected to advance the capability of Noah-MP to simultaneously predict weather and water quality in fully coupled Earth system models.

scholarly journals Automatically evaluating balance using machine learning and data from a single inertial measurement unit

2021 ◽  
Vol 18 (1) ◽  
Author(s):  
Fahad Kamran ◽  
Kathryn Harrold ◽  
Jonathan Zwier ◽  
Wendy Carender ◽  
Tian Bao ◽  
...  
Keyword(s):  
Machine Learning ◽  
Inertial Measurement Unit ◽  
Physical Therapists ◽  
Physical Therapist ◽  
Model Performance ◽  
Machine Learning Techniques ◽  
Measurement Unit ◽  
Balance Performance ◽  
Inertial Measurement ◽  
Self Assessments

Abstract Background Recently, machine learning techniques have been applied to data collected from inertial measurement units to automatically assess balance, but rely on hand-engineered features. We explore the utility of machine learning to automatically extract important features from inertial measurement unit data for balance assessment. Findings Ten participants with balance concerns performed multiple balance exercises in a laboratory setting while wearing an inertial measurement unit on their lower back. Physical therapists watched video recordings of participants performing the exercises and rated balance on a 5-point scale. We trained machine learning models using different representations of the unprocessed inertial measurement unit data to estimate physical therapist ratings. On a held-out test set, we compared these learned models to one another, to participants’ self-assessments of balance, and to models trained using hand-engineered features. Utilizing the unprocessed kinematic data from the inertial measurement unit provided significant improvements over both self-assessments and models using hand-engineered features (AUROC of 0.806 vs. 0.768, 0.665). Conclusions Unprocessed data from an inertial measurement unit used as input to a machine learning model produced accurate estimates of balance performance. The ability to learn from unprocessed data presents a potentially generalizable approach for assessing balance without the need for labor-intensive feature engineering, while maintaining comparable model performance.

Sign in / Sign up

Export Citation Format

Share Document