Deep Learning Optimizes Data-Driven Representation of Soil Organic Carbon in Earth System Model Over the Conterminous United States
Soil organic carbon (SOC) is a key component of the global carbon cycle, yet it is not well-represented in Earth system models to accurately predict global carbon dynamics in response to climate change. This novel study integrated deep learning, data assimilation, 25,444 vertical soil profiles, and...
Main Authors: | , , , , , , , , , , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Frontiers Media S.A.
2020-06-01
|
Series: | Frontiers in Big Data |
Subjects: | |
Online Access: | https://www.frontiersin.org/article/10.3389/fdata.2020.00017/full |
id |
doaj-ba7e853ff0ed4a08a846d414e43b4be2 |
---|---|
record_format |
Article |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Feng Tao Feng Tao Zhenghu Zhou Yuanyuan Huang Qianyu Li Qianyu Li Xingjie Lu Xingjie Lu Shuang Ma Xiaomeng Huang Xiaomeng Huang Yishuang Liang Yishuang Liang Gustaf Hugelius Lifen Jiang Russell Doughty Zhehao Ren Yiqi Luo |
spellingShingle |
Feng Tao Feng Tao Zhenghu Zhou Yuanyuan Huang Qianyu Li Qianyu Li Xingjie Lu Xingjie Lu Shuang Ma Xiaomeng Huang Xiaomeng Huang Yishuang Liang Yishuang Liang Gustaf Hugelius Lifen Jiang Russell Doughty Zhehao Ren Yiqi Luo Deep Learning Optimizes Data-Driven Representation of Soil Organic Carbon in Earth System Model Over the Conterminous United States Frontiers in Big Data soil organic carbon representation Earth system model data assimilation deep learning Community Land Model version 5 (CLM5) soil carbon dynamics |
author_facet |
Feng Tao Feng Tao Zhenghu Zhou Yuanyuan Huang Qianyu Li Qianyu Li Xingjie Lu Xingjie Lu Shuang Ma Xiaomeng Huang Xiaomeng Huang Yishuang Liang Yishuang Liang Gustaf Hugelius Lifen Jiang Russell Doughty Zhehao Ren Yiqi Luo |
author_sort |
Feng Tao |
title |
Deep Learning Optimizes Data-Driven Representation of Soil Organic Carbon in Earth System Model Over the Conterminous United States |
title_short |
Deep Learning Optimizes Data-Driven Representation of Soil Organic Carbon in Earth System Model Over the Conterminous United States |
title_full |
Deep Learning Optimizes Data-Driven Representation of Soil Organic Carbon in Earth System Model Over the Conterminous United States |
title_fullStr |
Deep Learning Optimizes Data-Driven Representation of Soil Organic Carbon in Earth System Model Over the Conterminous United States |
title_full_unstemmed |
Deep Learning Optimizes Data-Driven Representation of Soil Organic Carbon in Earth System Model Over the Conterminous United States |
title_sort |
deep learning optimizes data-driven representation of soil organic carbon in earth system model over the conterminous united states |
publisher |
Frontiers Media S.A. |
series |
Frontiers in Big Data |
issn |
2624-909X |
publishDate |
2020-06-01 |
description |
Soil organic carbon (SOC) is a key component of the global carbon cycle, yet it is not well-represented in Earth system models to accurately predict global carbon dynamics in response to climate change. This novel study integrated deep learning, data assimilation, 25,444 vertical soil profiles, and the Community Land Model version 5 (CLM5) to optimize the model representation of SOC over the conterminous United States. We firstly constrained parameters in CLM5 using observations of vertical profiles of SOC in both a batch mode (using all individual soil layers in one batch) and at individual sites (site-by-site). The estimated parameter values from the site-by-site data assimilation were then either randomly sampled (random-sampling) to generate continentally homogeneous (constant) parameter values or maximally preserved for their spatially heterogeneous distributions (varying parameter values to match the spatial patterns from the site-by-site data assimilation) so as to optimize spatial representation of SOC in CLM5 through a deep learning technique (neural networking) over the conterminous United States. Comparing modeled spatial distributions of SOC by CLM5 to observations yielded increasing predictive accuracy from default CLM5 settings (R2 = 0.32) to randomly sampled (0.36), one-batch estimated (0.43), and deep learning optimized (0.62) parameter values. While CLM5 with parameter values derived from random-sampling and one-batch methods substantially corrected the overestimated SOC storage by that with default model parameters, there were still considerable geographical biases. CLM5 with the spatially heterogeneous parameter values optimized from the neural networking method had the least estimation error and less geographical biases across the conterminous United States. Our study indicated that deep learning in combination with data assimilation can significantly improve the representation of SOC by complex land biogeochemical models. |
topic |
soil organic carbon representation Earth system model data assimilation deep learning Community Land Model version 5 (CLM5) soil carbon dynamics |
url |
https://www.frontiersin.org/article/10.3389/fdata.2020.00017/full |
work_keys_str_mv |
AT fengtao deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT fengtao deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT zhenghuzhou deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT yuanyuanhuang deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT qianyuli deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT qianyuli deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT xingjielu deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT xingjielu deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT shuangma deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT xiaomenghuang deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT xiaomenghuang deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT yishuangliang deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT yishuangliang deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT gustafhugelius deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT lifenjiang deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT russelldoughty deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT zhehaoren deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates AT yiqiluo deeplearningoptimizesdatadrivenrepresentationofsoilorganiccarboninearthsystemmodelovertheconterminousunitedstates |
_version_ |
1724444772041490432 |
spelling |
doaj-ba7e853ff0ed4a08a846d414e43b4be22020-11-25T04:02:00ZengFrontiers Media S.A.Frontiers in Big Data2624-909X2020-06-01310.3389/fdata.2020.00017509746Deep Learning Optimizes Data-Driven Representation of Soil Organic Carbon in Earth System Model Over the Conterminous United StatesFeng Tao0Feng Tao1Zhenghu Zhou2Yuanyuan Huang3Qianyu Li4Qianyu Li5Xingjie Lu6Xingjie Lu7Shuang Ma8Xiaomeng Huang9Xiaomeng Huang10Yishuang Liang11Yishuang Liang12Gustaf Hugelius13Lifen Jiang14Russell Doughty15Zhehao Ren16Yiqi Luo17Ministry of Education Key Laboratory for Earth System Modeling, Department of Earth System Science, Tsinghua University, Beijing, ChinaNational Supercomputing Center in Wuxi, Wuxi, ChinaCenter for Ecological Research, Northeast Forestry University, Harbin, ChinaLe Laboratoire des Sciences du Climat et de l'Environnement, IPSL-LSCECEA/CNRS/UVSQ Saclay, Gif-sur-Yvette, FranceMinistry of Education Key Laboratory for Earth System Modeling, Department of Earth System Science, Tsinghua University, Beijing, ChinaNational Supercomputing Center in Wuxi, Wuxi, ChinaSchool of Atmospheric Sciences, Sun Yat-sen University, Guangzhou, ChinaDepartment of Biological Sciences, Center for Ecosystem Science and Society, Northern Arizona University, Flagstaff, AZ, United StatesDepartment of Biological Sciences, Center for Ecosystem Science and Society, Northern Arizona University, Flagstaff, AZ, United StatesMinistry of Education Key Laboratory for Earth System Modeling, Department of Earth System Science, Tsinghua University, Beijing, ChinaNational Supercomputing Center in Wuxi, Wuxi, ChinaMinistry of Education Key Laboratory for Earth System Modeling, Department of Earth System Science, Tsinghua University, Beijing, ChinaNational Supercomputing Center in Wuxi, Wuxi, ChinaDepartment of Physical Geography and Bolin Centre for Climate Research, Stockholm University, Stockholm, SwedenDepartment of Biological Sciences, Center for Ecosystem Science and Society, Northern Arizona University, Flagstaff, AZ, United StatesDepartment of Microbiology and Plant Biology, University of Oklahoma, Norman, OK, United StatesMinistry of Education Key Laboratory for Earth System Modeling, Department of Earth System Science, Tsinghua University, Beijing, ChinaDepartment of Biological Sciences, Center for Ecosystem Science and Society, Northern Arizona University, Flagstaff, AZ, United StatesSoil organic carbon (SOC) is a key component of the global carbon cycle, yet it is not well-represented in Earth system models to accurately predict global carbon dynamics in response to climate change. This novel study integrated deep learning, data assimilation, 25,444 vertical soil profiles, and the Community Land Model version 5 (CLM5) to optimize the model representation of SOC over the conterminous United States. We firstly constrained parameters in CLM5 using observations of vertical profiles of SOC in both a batch mode (using all individual soil layers in one batch) and at individual sites (site-by-site). The estimated parameter values from the site-by-site data assimilation were then either randomly sampled (random-sampling) to generate continentally homogeneous (constant) parameter values or maximally preserved for their spatially heterogeneous distributions (varying parameter values to match the spatial patterns from the site-by-site data assimilation) so as to optimize spatial representation of SOC in CLM5 through a deep learning technique (neural networking) over the conterminous United States. Comparing modeled spatial distributions of SOC by CLM5 to observations yielded increasing predictive accuracy from default CLM5 settings (R2 = 0.32) to randomly sampled (0.36), one-batch estimated (0.43), and deep learning optimized (0.62) parameter values. While CLM5 with parameter values derived from random-sampling and one-batch methods substantially corrected the overestimated SOC storage by that with default model parameters, there were still considerable geographical biases. CLM5 with the spatially heterogeneous parameter values optimized from the neural networking method had the least estimation error and less geographical biases across the conterminous United States. Our study indicated that deep learning in combination with data assimilation can significantly improve the representation of SOC by complex land biogeochemical models.https://www.frontiersin.org/article/10.3389/fdata.2020.00017/fullsoil organic carbon representationEarth system modeldata assimilationdeep learningCommunity Land Model version 5 (CLM5)soil carbon dynamics |