Data and code from: Non-structural carbohydrates drive vegetation productivity: Global insights from a land surface model
Data files
Jun 10, 2026 version files 71.76 MB
-
Codes_and_Data.zip
71.73 MB
-
README.md
29.02 KB
Abstract
Non-structural carbohydrates (NSC), as primary products of photosynthesis, play a crucial role in plant responses to environmental stress and can, in turn, regulate photosynthetic activity under changing climatic conditions. In this study, we evaluate how land surface models represent the bi-directional interactions between NSC and photosynthesis and quantify the emergent causal relationships between NSC and gross primary productivity (GPP) at the global scale. We analyze model outputs from the latest versions of the Community Land Model (CLM5-BGC and CLM5-FATES) for the period 2000–2016, applying the Convergent Cross-Mapping (CCM) approach to assess causal interactions. Our results reveal a widespread bidirectional causal relationship between NSC and GPP, with significant coupling across 88.14% of vegetated land areas globally. Notably, NSC exerts a stronger influence on GPP than the reverse across more than 73.06% of the land surface, with this NSC-dominant control being particularly pronounced in arid and semi-arid ecosystems. These findings highlight the critical importance of NSC dynamics in regulating photosynthesis within land surface models and emphasize the need for improved representation and observational constraints of NSC-related processes, including carbon allocation, phenology, mortality, and canopy structure.
Description of the data and file structure
This dataset (https://doi.org/10.5061/dryad.mcvdnck9m) contains the data and code used to analyze the causal relationship between non-structural carbohydrates (NSC) and gross primary productivity (GPP) at the ecosystem scale. The analyses were conducted globally using outputs from the Community Land Model version 5 (CLM5), including the CLM5-BGC and CLM5-FATES configurations, together with Convergent Cross-Mapping (CCM), Random Forest Regression (RFR), Pearson correlation analysis, global GPP products, FLUXNET2015 observations, and field observations of NSC and Green Chromatic Coordinate (GCC).
The archive contains the code and data necessary to reproduce the main-text figures and supplementary figures of the associated manuscript. The dataset includes:
- Monthly and yearly carbon-related state variables from CLM5-BGC and CLM5-FATES.
- Environmental variables from CLM5-BGC and CLM5-FATES.
- Global GPP products and FLUXNET2015 GPP observations.
- Observed NSC concentrations compiled from published literature.
- Field observations of NSC and GCC from Tiantong.
- CCM results evaluating causal interactions between GPP and NSC.
- CCM results evaluating causal interactions between GCC and NSC at Tiantong.
- Random Forest Regression model outputs for evaluating GPP–NSC causal relationships while accounting for vegetation functional properties and climatic factors.
- Pearson correlation results between GPP and NSC.
- Data linking causal effects between NSC and GPP to GPP prediction errors.
File organization
The dataset is provided as one compressed archive named Codes_and_Data.zip.
File: Code_and_Data.zip
Description: This archive contains all code and data used to reproduce the analyses and figures in the manuscript.
After extraction, the archive contains the following folders:
Code_and_Data/
├── 0-Codes/
│ └── Figures.ipynb
└── 1-Data/
├── BGC_vars.csv
├── CVars-BGC.nc
├── CVars-BGC-yearly.nc
├── CVars-FATES.nc
├── CVars-FATES-yearly.nc
├── Corr-BGC.nc
├── Corr-FATES.nc
├── EnvVars-BGC.nc
├── EnvVars-FATES.nc
├── FATES_vars.csv
├── GCC.csv
├── GPP_Fluxcom.nc
├── GPP_Fluxnet.csv
├── GPP_Fluxnet_Models_Products.csv
├── GPP_GIMMSv2.nc
├── GPP_GOSIF.nc
├── GPP_OCS_exp.nc
├── GPP_P-Model.nc
├── NSC.csv
├── NSC_Globe.csv
├── RFR-GPP2NSC-BGC.csv
├── RFR-GPP2NSC-FATES.csv
├── RFR-NSC2GPP-BGC.csv
├── RFR-NSC2GPP-FATES.csv
├── RFR_GPP2NSC_BGC_bestmodel.pkl
├── RFR_GPP2NSC_FATES_bestmodel.pkl
├── RFR_NSC2GPP_BGC_bestmodel.pkl
├── RFR_NSC2GPP_FATES_bestmodel.pkl
├── ccm_bgc.nc
├── ccm_fates.nc
├── ccm_gcc_nsc.csv
├── ccm_nsc_gcc.csv
├── nsc_mean.csv
├── rho_GPP_RMSE_BGC.csv
└── rho_GPP_RMSE_FATES.csv
Files and variables
Code files
0-Codes/Figures.ipynb
Description: Jupyter Notebook used to reproduce Figures 2–8 and Supplementary Figures 3, 4, 6, 7, and 9 of the manuscript. The notebook reads the data files stored in 1-Data/, performs data processing and visualization, and generates the figures used in the paper.
The notebook includes code for:
- LoadingCLM5-BGC and CLM5-FATES outputs.
- Loading observational and remote-sensing-based GPP datasets.
- Loading observed NSC and GCC datasets.
- ReadingCCM, Random Forest Regression, and Pearson correlation analysis outputs.
- Producing the main text and supplementary figures.
Software requirements: The notebook was written in Python. Required packages include, but may not be limited to:
numpypandasxarraynetCDF4matplotlibscipyscikit-learnpickle
Please refer to the notebook for any additional packages required to reproduce the figures.
Data files
1. CLM5 carbon-related state variables
These files contain the main carbon-related outputs from CLM5-BGC and CLM5-FATES and form the basis of the analysis.
1-Data/CVars-BGC.nc
Description: Monthly carbon-related state variables simulated by CLM5-BGC.
Dimensions include:
time: Monthly time step.lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
GPP: Gross primary productivity simulated by CLM5-BGC. Units: gC m-2 month-1.cStore: Non-structural carbohydrate storage simulated by CLM5-BGC. Units: kgC m-2.NPP: Net primary productivity simulated by CLM5-BGC. Units: gC m-2 month-1.cVeg: Carbon storage in biomass pools simulated by CLM5-BGC. Units: kgC m-2.cStruc: Carbon storage in structural pools simulated by CLM5-BGC. Units: kgC m-2.cLeaf: Carbon storage in the leaf pool simulated by CLM5-BGC. Units: kgC m-2.LAI: Leaf area index simulated by CLM5-BGC. Units: m2 m-2.
1-Data/CVars-FATES.nc
Description: Monthly carbon-related state variables simulated by CLM5-FATES.
Dimensions include:
time: Monthly time step.lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
GPP: Gross primary productivity simulated by CLM5-FATES. Units: gC m-2 month-1.cStore: Non-structural carbohydrate storage simulated by CLM5-FATES. Units: kgC m-2.NPP: Net primary productivity simulated by CLM5-FATES. Units: gC m-2 month-1.cVeg: Carbon storage in biomass pools simulated by CLM5-FATES. Units: kgC m-2.cStruc: Carbon storage in structural pools simulated by CLM5-FATES. Units: kgC m-2.cLeaf: Carbon storage in the leaf pool simulated by CLM5-FATES. Units: kgC m-2.LAI: Leaf area index simulated by CLM5-FATES. Units: m2 m-2.
1-Data/CVars-BGC-yearly.nc
Description: Yearly aggregation of carbon-related state variables from CLM5-BGC. This file was derived from CVars-BGC.nc by aggregating monthly data to annual values.
Dimensions include:
time: Yearly time step.lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
GPP: Annual gross primary productivity from CLM5-BGC. Units: gC m-2 year-1.cStore: Annual carbon storage in the NSC pool from CLM5-BGC. Units: kgC m-2.NPP: Annual net primary productivity from CLM5-BGC. Units: gC m-2 year-1.cVeg: Annual carbon storage in biomass pools from CLM5-BGC. Units: kgC m-2.cStruc: Annual carbon storage in structural pools from CLM5-BGC. Units: kgC m-2.cLeaf: Annual carbon storage in the leaf pool from CLM5-BGC. Units: kgC m-2.LAI: Annual leaf area index from CLM5-BGC. Units: m2 m-2.
1-Data/CVars-FATES-yearly.nc
Description: Yearly aggregation of carbon-related state variables from CLM5-FATES. This file was derived from CVars-FATES.nc by aggregating monthly data to annual values.
Dimensions include:
time: Yearly time step.lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
GPP: Annual gross primary productivity from CLM5-FATES. Units: gC m-2 year-1.cStore: Annual carbon storage in the NSC pool from CLM5-FATES. Units: kgC m-2.NPP: Annual net primary productivity from CLM5-FATES. Units: gC m-2 year-1.cVeg: Annual carbon storage in biomass pools from CLM5-FATES. Units: kgC m-2.cStruc: Annual carbon storage in structural pools from CLM5-FATES. Units: kgC m-2.cLeaf: Annual carbon storage in the leaf pool from CLM5-FATES. Units: kgC m-2.LAI: Annual leaf area index from CLM5-FATES. Units: m2 m-2.
1-Data/BGC_vars.csv
Description: Aggregated environmental variables, carbon-related variables, vegetation functional properties, and causal-effect metrics for all CLM5-BGC pixels. This file was used for Figure 8, correlation matrix analysis, and Random Forest Regression analysis.
Variables include:
rho_diff: Difference between the causal effect of NSC on GPP (cStore2GPP) and the causal effect of GPP on NSC (GPP2cStore). Unitless.cStore2GPP: Causal effect of NSC on GPP. Unitless.GPP2cStore: Causal effect of GPP on NSC. Unitless.Fleaf: Fraction of leaf carbon in structural biomass. Units: %.SLA: Specific leaf area. Units: m2 gC-1.PLAI: LAI-level photosynthesis. Units: gC m-2 year-1.LAI: Leaf area index. Units: m2 m-2.CUE: Carbon-use efficiency. Units: %.tauveg: Vegetation carbon turnover time. Units: year.FNSC: Fraction of NSC in total biomass. Units: %.Tair: Multi-year averaged air temperature. Units: °C.Tsoil: Multi-year averaged soil temperature in the top 10 cm of soil. Units: °C.Prec: Multi-year averaged precipitation. Units: mm.SWC: Multi-year averaged soil moisture content. Units: kg m-2.VPD: Multi-year averaged canopy vapor pressure deficit. Units: kPa.SWdown: Multi-year averaged atmospheric incident solar radiation. Units: W m-2.PARVEGLN: Multi-year averaged photosynthetically active radiation absorbed by vegetation at local noon. Units: W m-2.GPP: Multi-year averaged gross primary productivity. Units: gC m-2 year-1.cStore: Multi-year averaged NSC storage. Units: kgC m-2.
1-Data/FATES_vars.csv
Description: Aggregated environmental variables, carbon-related variables, vegetation functional properties, and causal-effect metrics for all CLM5-FATES pixels. This file was used for Figure 8, correlation matrix analysis, and Random Forest Regression analysis.
Variables include:
rho_diff: Difference between the causal effect of NSC on GPP (cStore2GPP) and the causal effect of GPP on NSC (GPP2cStore). Unitless.cStore2GPP: Causal effect of NSC on GPP. Unitless.GPP2cStore: Causal effect of GPP on NSC. Unitless.Fleaf: Fraction of leaf carbon in structural biomass. Units: %.SLA: Specific leaf area. Units: m2 gC-1.PLAI: LAI-level photosynthesis. Units: gC m-2 year-1.LAI: Leaf area index. Units: m2 m-2.CUE: Carbon-use efficiency. Units: %.tauveg: Vegetation carbon turnover time. Units: year.FNSC: Fraction of NSC in total biomass. Units: %.Tair: Multi-year averaged air temperature. Units: °C.Tsoil: Multi-year averaged soil temperature in the top 10 cm of soil. Units: °C.Prec: Multi-year averaged precipitation. Units: mm.SWC: Multi-year averaged soil moisture content. Units: kg m-2.VPD: Multi-year averaged canopy vapor pressure deficit. Units: kPa.SWdown: Multi-year averaged atmospheric incident solar radiation. Units: W m-2.PARVEGLN: Multi-year averaged photosynthetically active radiation absorbed by vegetation at local noon. Units: W m-2.GPP: Multi-year averaged gross primary productivity. Units: gC m-2 year-1.cStore: Multi-year averaged NSC storage. Units: kgC m-2.
2. CLM5 environmental variables
These environmental variables were used as explanatory variables or covariates in the analysis of GPP–NSC relationships.
1-Data/EnvVars-BGC.nc
Description: Multi-year averaged environmental variables from CLM5-BGC used as explanatory variables or covariates in the analysis of GPP–NSC relationships.
Dimensions include:
lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
Tair: Multi-year averaged air temperature. Units: °C.Tsoil: Multi-year averaged soil temperature in the top 10 cm of soil. Units: °C.Prec: Multi-year averaged precipitation. Units: mm.SWC: Multi-year averaged soil moisture content. Units: kg m-2.VPD: Multi-year averaged canopy vapor pressure deficit. Units: kPa.SWdown: Multi-year averaged atmospheric incident solar radiation. Units: W m-2.PARVEGLN: Multi-year averaged photosynthetically active radiation absorbed by vegetation at local noon. Units: W m-2.
1-Data/EnvVars-FATES.nc
Description: Multi-year averaged environmental variables from CLM5-FATES used as explanatory variables or covariates in the analysis of GPP–NSC relationships.
Dimensions include:
lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
Tair: Multi-year averaged air temperature. Units: °C.Tsoil: Multi-year averaged soil temperature in the top 10 cm of soil. Units: °C.Prec: Multi-year averaged precipitation. Units: mm.SWC: Multi-year averaged soil moisture content. Units: kg m-2.VPD: Multi-year averaged canopy vapor pressure deficit. Units: kPa.SWdown: Multi-year averaged atmospheric incident solar radiation. Units: W m-2.PARVEGLN: Multi-year averaged photosynthetically active radiation absorbed by vegetation at local noon. Units: W m-2.
3. Global GPP products and FLUXNET observations
These files contain annual or multi-year mean GPP estimates from global GPP products and FLUXNET2015 observations. They were used to compare model-simulated GPP with observation-based or remote-sensing-based GPP estimates.
1-Data/GPP_Fluxcom.nc
Description: Multi-year mean GPP from the FLUXCOM product.
Dimensions include:
lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
GPP: Gross primary productivity from FLUXCOM. Units: gC m-2 year-1.
1-Data/GPP_GIMMSv2.nc
Description: Multi-year mean GPP from the GIMMSv2 GPP product.
Dimensions include:
lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
GPP: Gross primary productivity from GIMMSv2. Units: gC m-2 year-1.
1-Data/GPP_GOSIF.nc
Description: Multi-year mean GPP derived from the GOSIF-GPP dataset.
Dimensions include:
lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
GPP: Gross primary productivity from GOSIF. Units: gC m-2 year-1.
1-Data/GPP_OCS_exp.nc
Description: Multi-year mean GPP derived from a carbonyl sulfide-based GPP product.
Dimensions include:
lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
GPP: Gross primary productivity from the OCS-based GPP product. Units: gC m-2 year-1.
1-Data/GPP_P-Model.nc
Description: Multi-year mean GPP from the P-model.
Dimensions include:
lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
GPP: Gross primary productivity from the P-model. Units: gC m-2 year-1.
1-Data/GPP_Fluxnet.csv
Description: Site-level GPP observations from the FLUXNET2015 dataset.
Variables include:
SITE_ID: FLUXNET site identifier.YEAR: Year of observation.LOCATION_LAT: Latitude of the FLUXNET site in degrees north.LOCATION_LONG: Longitude of the FLUXNET site in degrees east.GPP: Gross primary productivity at the FLUXNET site. Units: gC m-2 year-1.
1-Data/GPP_Fluxnet_Models_Products.csv
Description: Compiled table comparing GPP from FLUXNET observations, CLM5 model simulations, and global GPP products. This file was used to evaluate modelled and product-based GPP against site-level FLUXNET observations.
Variables include:
SITE_ID: FLUXNET site identifier.YEAR: Year of observation.lat: Latitude of the 0.5° grid cell containing the FLUXNET site. Units: degrees north.lon: Longitude of the 0.5° grid cell containing the FLUXNET site. Units: degrees east.GPP_BGC: GPP simulated by CLM5-BGC. Units: gC m-2 year-1.GPP_FATES: GPP simulated by CLM5-FATES. Units: gC m-2 year-1.GPP_Fluxnet: GPP from FLUXNET observations assigned to the corresponding 0.5° grid cell. Units: gC m-2 year-1.- Variables for global GPP products, including FLUXCOM, GIMMSv2, GOSIF, OCS, and P-model, if present. Units: gC m-2 year-1.
4. Observed NSC and GCC data
These NSC observation datasets were used to validate the NSC simulated by CLM5-BGC and CLM5-FATES. The NSC and GCC observations at Tiantong were used to evaluate the site-scale causal relationship between vegetation greenness and NSC.
1-Data/NSC_Globe.csv
Description: Compiled dataset of observed NSC concentrations from published literature across the Northern Hemisphere. This dataset was used to evaluate or compare model-simulated NSC patterns with field observations.
Variables include:
Tissue type: Plant tissue sampled for NSC measurement, e.g., leaf, stem, or root.latitude: Latitude of the observation site in degrees north.longitude: Longitude of the observation site in degrees east.Xc_%: Observed non-structural carbohydrate concentration. Units: %.reference: Literature source of the observation.
1-Data/NSC.csv
Description: Field observations of non-structural carbohydrates at Tiantong. These data were used together with GCC observations to evaluate the causal relationship between vegetation greenness and NSC at the site scale.
Variables include:
Date: Date of observation.Plot: Sample plot identifier. Each plot contains three replicate samples.NSC_mean: Mean NSC concentration. Units: mg g-1.NSC_std: Standard deviation of NSC concentration. Units: mg g-1.
1-Data/nsc_mean.csv
Description: Aggregated NSC observations under ambient treatment at Tiantong, derived from NSC.csv.
Variables include:
Date: Date of observation.NSC_mean: Mean NSC concentration. Units: mg g-1.NSC_std: Standard deviation of NSC concentration. Units: mg g-1.
1-Data/GCC.csv
Description: Green Chromatic Coordinate observations at Tiantong. GCC is a camera-based vegetation greenness index and was used as an observational proxy for canopy phenology or canopy greenness. Two species were observed in each plot.
Variables include:
Date: Date of observation.Plot: Observation plot identifier.GCC_mean: Mean Green Chromatic Coordinate. Unitless.GCC_std: Standard deviation of Green Chromatic Coordinate. Unitless.
5. Convergent Cross-Mapping results
These files contain the main CCM results of the study.
1-Data/ccm_bgc.nc
Description: Convergent Cross-Mapping results between GPP and NSC using CLM5-BGC outputs. These results were used to evaluate the causal relationship between NSC and GPP across global grid cells.
Dimensions include:
lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
cStore2GPP_rho: Causal effect of NSC on GPP. Unitless.cStore2GPP_lag: Time delay for the causal effect of NSC on GPP. Units: month.GPP2cStore_rho: Causal effect of GPP on NSC. Unitless.GPP2cStore_lag: Time delay for the causal effect of GPP on NSC. Units: month.
1-Data/ccm_fates.nc
Description: Convergent Cross-Mapping results between GPP and NSC using CLM5-FATESS outputs. This file has the same general structure as ccm_bgc.nc, but the results are based on CLM5-FATES.
Dimensions include:
lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
cStore2GPP_rho: Causal effect of NSC on GPP. Unitless.cStore2GPP_lag: Time delay for the causal effect of NSC on GPP. Units: month.GPP2cStore_rho: Causal effect of GPP on NSC. Unitless.GPP2cStore_lag: Time delay for the causal effect of GPP on NSC. Units: month.
1-Data/ccm_gcc_nsc.csv
Description: Convergent Cross-Mapping results evaluating the causal influence of GCC on NSC at Tiantong station.
Variables include:
CCM_boot_A.Lobs: Library size used in CCM. Units: number of observations.CCM_boot_A.rho: Cross-map skill or causal-effect estimate from CCM. Unitless.CCM_boot_A.sdevrho: Standard deviation of the CCM causal-effect estimate. Unitless.
1-Data/ccm_nsc_gcc.csv
Description: Convergent Cross-Mapping results evaluating the causal influence of NSC on GCC at Tiantong station.
Variables include:
CCM_boot_B.Lobs: Library size used in CCM. Units: number of observations.CCM_boot_B.rho: Cross-map skill or causal-effect estimate from CCM. Unitless.CCM_boot_B.sdevrho: Standard deviation of the CCM causal-effect estimate. Unitless.
6. Relationship between causal effect and GPP prediction error
1-Data/rho_GPP_RMSE_BGC.csv
Description: Relationship between the causal effect of NSC–GPP interactions and GPP prediction error in CLM5-BGC. This file was used to evaluate whether stronger NSC–GPP causal coupling is associated with lower or higher GPP prediction error.
Variables include:
lat: Latitude in degrees north.lon: Longitude in degrees east.rho: Causal-effect estimate from CCM. Unitless.rho_type: Direction of the causal effect.cStore2GPPmeans NSC -> GPP, andGPP2cStoremeans GPP -> NSC.GPP_BGC: Multi-year mean GPP simulated by CLM5-BGC at the grid cell. Units: gC m-2 year-1.RMSE: Root Mean Square Error between simulated GPP and other GPP products. Units: gC m-2 year-1.
1-Data/rho_GPP_RMSE_FATES.csv
Description: Relationship between the causal effect of NSC–GPP interactions and GPP prediction error in CLM5-FATES. This file was used to evaluate whether stronger NSC–GPP causal coupling is associated with lower or higher GPP prediction error.
Variables include:
lat: Latitude in degrees north.lon: Longitude in degrees east.rho: Causal-effect estimate from CCM. Unitless.rho_type: Direction of the causal effect.cStore2GPPmeans NSC -> GPP, andGPP2cStoremeans GPP -> NSC.GPP_FATES: Multi-year mean GPP simulated by CLM5-FATES at the grid cell. Units: gC m-2 year-1.RMSE: Root Mean Square Error between simulated GPP and other GPP products. Units: gC m-2 year-1.
7. Random Forest Regression outputs
The following files contain Random Forest Regression outputs used to evaluate causal relationships between GPP and NSC while accounting for vegetation functional properties and climatic factors.
1-Data/RFR-GPP2NSC-BGC.csv
Description: Tabular output from Random Forest Regression for predicting the causal effect of GPP on NSC in CLM5-BGC using explanatory variables.
Variables include:
FeatureName: Predictor variable name.FeatureImportance: Variable importance estimated by the Random Forest Regression model. Units: %.FeatureLabel: Label used for plotting.BarColor: Bar color used for plotting.
1-Data/RFR-GPP2NSC-FATES.csv
Description: Tabular output from Random Forest Regression for predicting the causal effect of GPP on NSC in CLM5-FATES using explanatory variables.
Variables include:
FeatureName: Predictor variable name.FeatureImportance: Variable importance estimated by the Random Forest Regression model. Units: %.FeatureLabel: Label used for plotting.BarColor: Bar color used for plotting.
1-Data/RFR-NSC2GPP-BGC.csv
Description: Tabular output from Random Forest Regression for predicting the causal effect of NSC on GPP in CLM5-BGC using explanatory variables.
Variables include:
FeatureName: Predictor variable name.FeatureImportance: Variable importance estimated by the Random Forest Regression model. Units: %.FeatureLabel: Label used for plotting.BarColor: Bar color used for plotting.
1-Data/RFR-NSC2GPP-FATES.csv
Description: Tabular output from Random Forest Regression for predicting the causal effect of NSC on GPP in CLM5-FATES using explanatory variables.
Variables include:
FeatureName: Predictor variable name.FeatureImportance: Variable importance estimated by the Random Forest Regression model. Units: %.FeatureLabel: Label used for plotting.BarColor: Bar color used for plotting.
1-Data/RFR_GPP2NSC_BGC_bestmodel.pkl
Description: Save the best Random Forest Regression model for predicting the causal effect of GPP on NSC in CLM5-BGC using explanatory variables.
1-Data/RFR_GPP2NSC_FATES_bestmodel.pkl
Description: Saved the best Random Forest Regression model for predicting the causal effect of GPP on NSC in CLM5-FATES using explanatory variables.
1-Data/RFR_NSC2GPP_BGC_bestmodel.pkl
Description: Saved the best Random Forest Regression model for predicting the causal effect of NSC on GPP in CLM5-BGC using explanatory variables.
1-Data/RFR_NSC2GPP_FATES_bestmodel.pkl
Description: Saved the best Random Forest Regression model for predicting the causal effect of NSC on GPP in CLM5-FATES using explanatory variables.
Note: The .pkl Files are binary Python objects containing saved Random Forest and Regression models. They are not plain-text tabular data files. They can be loaded in Python using pickle or joblib, depending on how they were generated.
8. Pearson correlation analysis results
1-Data/Corr-BGC.nc
Description: Pearson correlation analysis results between monthly GPP and NSC using CLM5-BGC outputs. The file contains spatially explicit correlation statistics across global grid cells.
Dimensions include:
lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
rvalue: Pearson correlation coefficient between monthly GPP and NSC. Unitless.pvalue: Statistical significance of the correlation. Unitless.
1-Data/Corr-FATES.nc
Description: Pearson correlation analysis results between monthly GPP and NSC using CLM5-FATESS outputs. The file has the same general purpose as Corr-BGC.nc, but the results are based on CLM5-FATES.
Dimensions include:
lat: Latitude in degrees north.lon: Longitude in degrees east.
Variables include:
rvalue: Pearson correlation coefficient between monthly GPP and NSC. Unitless.pvalue: Statistical significance of the correlation. Unitless.
Repeated variables and abbreviations
The following variables or abbreviations are used across multiple files.
NSC: Non-structural carbohydrates. NSC generally refers to soluble sugars and starch that can be stored and remobilized by plants.cStore: Modelled non-structural carbohydrate storage. Units are usually kgC m-2 unless otherwise stated.GPP: Gross primary productivity, representing the total carbon fixed by photosynthesis. Units depend on the file but are usually expressed as gC m-2 month-1 or gC m-2 year-1.GCC: Green Chromatic Coordinate, a unitless vegetation greenness index calculated from digital images.CCM: Convergent Cross-Mapping is a method used to infer causal relationships from nonlinear dynamical systems.RFR: Random Forest Regression.CLM5: Community Land Model version 5.CLM5-BGC: Biogeochemistry module of the Community Land Model version 5.CLM5-FATES: Functionally Assembled Terrestrial Ecosystem Simulator configuration coupled with CLM5.RMSE: Root Mean Square Error. Units are the same as the predicted variable.rho: Correlation coefficient, causal effect strength, depending on the file. Unitless.rvalue: Pearson correlation coefficient. Unitless.pvalue: Statistical significance value. Unitless.lat: Latitude in degrees north.lon: Longitude in degrees east.time: Time coordinate of the data. The temporal resolution depends on the file, e.g., monthly or yearly.year: Calendar year.Date: Date of observation. Format: year-month-day, unless otherwise specified in the file.
Notes on data reuse
The NetCDF files can be opened using common software packages such as xarray, netCDF4, Panoply, or other NetCDF-compatible software. The .csv files can be opened using spreadsheet software or data-analysis tools such as R or Python. The .pkl files are Python binary files and should be loaded in Python usingCompatible versions of the packages used to generate them.
Example Python code for opening NetCDF files:
import xarray as xr
ds = xr.open_dataset("1-Data/CVars-BGC.nc")
print(ds)
Example Python code for openinusing compatiblemple Python code for opening pickle files:
import pickle
with open("1-Data/RFR_NSC2GPP_BGC_bestmodel.pkl", "rb") as f:
model = pickle.load(f)
Reproducibility
To reproduce the figures, extract Code_and_Data.zip, open 0-Codes/Figures.ipynb, and run the notebook after installing the required Python packages. The notebook assumes that the folder structure remains unchanged, with all input data stored in 1-Data/.
The code was used to reproduce Figures 2–8 and Supplementary Figures 3, 4, 6, 7, and 9 in the associated manuscript.
Contact
For questions about the dataset, please contact Zhiqin Tu (zhqtu@stu.ecnu.edu.cn).
