A differentiable, physics-informed ecosystem modeling and learning framework for large-scale inverse problems: demonstration with photosynthesis simulations
<p>Photosynthesis plays an important role in carbon, nitrogen, and water cycles. Ecosystem models for photosynthesis are characterized by many parameters that are obtained from limited in situ measurements and applied to the same plant types. Previous site-by-site calibration approaches could...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Copernicus Publications
2023-07-01
|
| Series: | Biogeosciences |
| Online Access: | https://bg.copernicus.org/articles/20/2671/2023/bg-20-2671-2023.pdf |
| _version_ | 1797786104505565184 |
|---|---|
| author | D. Aboelyazeed C. Xu F. M. Hoffman F. M. Hoffman J. Liu A. W. Jones C. Rackauckas K. Lawson C. Shen |
| author_facet | D. Aboelyazeed C. Xu F. M. Hoffman F. M. Hoffman J. Liu A. W. Jones C. Rackauckas K. Lawson C. Shen |
| author_sort | D. Aboelyazeed |
| collection | DOAJ |
| description | <p>Photosynthesis plays an important role in carbon,
nitrogen, and water cycles. Ecosystem models for photosynthesis are
characterized by many parameters that are obtained from limited in situ
measurements and applied to the same plant types. Previous site-by-site
calibration approaches could not leverage big data and faced issues like
overfitting or parameter non-uniqueness. Here we developed an end-to-end
programmatically differentiable (meaning gradients of outputs to variables
used in the model can be obtained efficiently and accurately) version of the
photosynthesis process representation within the Functionally Assembled
Terrestrial Ecosystem Simulator (FATES) model. As a genre of
physics-informed machine learning (ML), differentiable models couple
physics-based formulations to neural networks (NNs) that learn parameterizations
(and potentially processes) from observations, here photosynthesis rates. We
first demonstrated that the framework was able to correctly recover multiple assumed
parameter values concurrently using synthetic training data. Then, using a
real-world dataset consisting of many different plant functional types (PFTs), we
learned parameters that performed substantially better and greatly reduced
biases compared to literature values. Further, the framework allowed us to
gain insights at a large scale. Our results showed that the carboxylation
rate at 25 <span class="inline-formula"><sup>∘</sup></span>C (<span class="inline-formula"><i>V</i><sub>c,max25</sub></span>) was more impactful than a factor
representing water limitation, although tuning both was helpful in
addressing biases with the default values. This framework could potentially
enable substantial improvement in our capability to learn parameters and
reduce biases for ecosystem modeling at large scales.</p> |
| first_indexed | 2024-03-13T01:03:22Z |
| format | Article |
| id | doaj.art-caaf71af3f2a43cb88285b74250201eb |
| institution | Directory Open Access Journal |
| issn | 1726-4170 1726-4189 |
| language | English |
| last_indexed | 2024-03-13T01:03:22Z |
| publishDate | 2023-07-01 |
| publisher | Copernicus Publications |
| record_format | Article |
| series | Biogeosciences |
| spelling | doaj.art-caaf71af3f2a43cb88285b74250201eb2023-07-06T09:54:37ZengCopernicus PublicationsBiogeosciences1726-41701726-41892023-07-01202671269210.5194/bg-20-2671-2023A differentiable, physics-informed ecosystem modeling and learning framework for large-scale inverse problems: demonstration with photosynthesis simulationsD. Aboelyazeed0C. Xu1F. M. Hoffman2F. M. Hoffman3J. Liu4A. W. Jones5C. Rackauckas6K. Lawson7C. Shen8Civil and Environmental Engineering, The Pennsylvania State University, University Park, PA 16802, USAEarth and Environmental Sciences Division, Los Alamos National Laboratory, Los Alamos, NM 87544, USAComputational Sciences & Engineering Division and the Climate Change Science Institute, Oak Ridge National Laboratory, Oak Ridge, TN 37830, USADepartment of Civil and Environmental Engineering, University of Tennessee, Knoxville, TN 37996, USACivil and Environmental Engineering, The Pennsylvania State University, University Park, PA 16802, USASciML, Open Source Software Organization, Cambridge, MA, USAComputer Science and Artificial Intelligence Laboratory (CSAIL), Massachusetts Institute of Technology, Cambridge, MA 02139, USACivil and Environmental Engineering, The Pennsylvania State University, University Park, PA 16802, USACivil and Environmental Engineering, The Pennsylvania State University, University Park, PA 16802, USA<p>Photosynthesis plays an important role in carbon, nitrogen, and water cycles. Ecosystem models for photosynthesis are characterized by many parameters that are obtained from limited in situ measurements and applied to the same plant types. Previous site-by-site calibration approaches could not leverage big data and faced issues like overfitting or parameter non-uniqueness. Here we developed an end-to-end programmatically differentiable (meaning gradients of outputs to variables used in the model can be obtained efficiently and accurately) version of the photosynthesis process representation within the Functionally Assembled Terrestrial Ecosystem Simulator (FATES) model. As a genre of physics-informed machine learning (ML), differentiable models couple physics-based formulations to neural networks (NNs) that learn parameterizations (and potentially processes) from observations, here photosynthesis rates. We first demonstrated that the framework was able to correctly recover multiple assumed parameter values concurrently using synthetic training data. Then, using a real-world dataset consisting of many different plant functional types (PFTs), we learned parameters that performed substantially better and greatly reduced biases compared to literature values. Further, the framework allowed us to gain insights at a large scale. Our results showed that the carboxylation rate at 25 <span class="inline-formula"><sup>∘</sup></span>C (<span class="inline-formula"><i>V</i><sub>c,max25</sub></span>) was more impactful than a factor representing water limitation, although tuning both was helpful in addressing biases with the default values. This framework could potentially enable substantial improvement in our capability to learn parameters and reduce biases for ecosystem modeling at large scales.</p>https://bg.copernicus.org/articles/20/2671/2023/bg-20-2671-2023.pdf |
| spellingShingle | D. Aboelyazeed C. Xu F. M. Hoffman F. M. Hoffman J. Liu A. W. Jones C. Rackauckas K. Lawson C. Shen A differentiable, physics-informed ecosystem modeling and learning framework for large-scale inverse problems: demonstration with photosynthesis simulations Biogeosciences |
| title | A differentiable, physics-informed ecosystem modeling and learning framework for large-scale inverse problems: demonstration with photosynthesis simulations |
| title_full | A differentiable, physics-informed ecosystem modeling and learning framework for large-scale inverse problems: demonstration with photosynthesis simulations |
| title_fullStr | A differentiable, physics-informed ecosystem modeling and learning framework for large-scale inverse problems: demonstration with photosynthesis simulations |
| title_full_unstemmed | A differentiable, physics-informed ecosystem modeling and learning framework for large-scale inverse problems: demonstration with photosynthesis simulations |
| title_short | A differentiable, physics-informed ecosystem modeling and learning framework for large-scale inverse problems: demonstration with photosynthesis simulations |
| title_sort | differentiable physics informed ecosystem modeling and learning framework for large scale inverse problems demonstration with photosynthesis simulations |
| url | https://bg.copernicus.org/articles/20/2671/2023/bg-20-2671-2023.pdf |
| work_keys_str_mv | AT daboelyazeed adifferentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT cxu adifferentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT fmhoffman adifferentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT fmhoffman adifferentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT jliu adifferentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT awjones adifferentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT crackauckas adifferentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT klawson adifferentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT cshen adifferentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT daboelyazeed differentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT cxu differentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT fmhoffman differentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT fmhoffman differentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT jliu differentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT awjones differentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT crackauckas differentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT klawson differentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations AT cshen differentiablephysicsinformedecosystemmodelingandlearningframeworkforlargescaleinverseproblemsdemonstrationwithphotosynthesissimulations |