Urban Biogenic CO2 fluxes: GPP, Reco and NEE Estimates from SMUrF, 2010-2019

This dataset contains estimates of biogenic CO2 flux components at 0.05 degree resolution from the Solar-Induced Fluorescence (SIF) for Modeling Urban biogenic Fluxes (SMUrF) model. Estimates were produced for the following regions and periods: eastern and western CONUS (2010-2019), western Europe (2010-2014 and 2017-2018), eastern Asia, eastern China, eastern Australia, South America, and Central Africa (2017-2018). Modeled CO2 flux components include gross primary production (GPP), ecosystem respiration (Reco), and net ecosystem exchange (NEE). Four-day means of GPP are estimated from solar-induced fluorescence (SIF) and biome-specific GPP-SIF relationships. Daily estimates of Reco are included. In addition, GPP and Reco were downscaled to hourly estimates and used to generate hourly NEE. Uncertainties for 4-day GPP and daily Reco estimates are provided. The input data streams included 500 m MODIS-based annual land cover classification, 0.05 degree spatiotemporally contiguous SIF, above-ground biomass (AGB) from GlobBiomass, eddy-covariance (EC) flux measurements, and gridded products of air and soil temperatures.

There are 925 data files in NetCDF version 4 (*.nc4) format included in this dataset.

This dataset contains estimates of biogenic CO₂ flux components at 0.05 degree resolution from the Solar-Induced Fluorescence (SIF) for Modeling Urban biogenic Fluxes (SMUrF) model. Estimates were produced for the following regions and periods: eastern and western CONUS (2010–2019), western Europe (2010–2014 and 2017–2018), eastern Asia, eastern China, eastern Australia, South America, and Central Africa (2017–2018). Modeled CO₂ flux components include gross primary production (GPP), ecosystem respiration (Reco), and net ecosystem exchange (NEE). Four-day means of GPP are estimated from solar-induced fluorescence (SIF) and biome-specific GPP-SIF relationships. Daily estimates of Reco are included. In addition, GPP and Reco were downscaled to hourly estimates and used to generate hourly NEE. Uncertainties for 4-day GPP and daily Reco estimates are provided. The input data streams included 500 m MODIS-based annual land cover classification, 0.05 degree spatiotemporally contiguous SIF, above-ground biomass (AGB) from GlobBiomass, eddy-covariance (EC) flux measurements, and gridded products of air and soil temperatures.

Project: North American Carbon Program

The North American Carbon Program (NACP) is a multidisciplinary research program designed to improve understanding of North America's carbon sources, sinks, and stocks. The central objective is to measure and understand the sources and sinks of Carbon Dioxide (CO₂), Methane (CH₄), and Carbon Monoxide (CO) in North America and adjacent oceans. The NACP is supported by a number of different federal agencies.

Related Publication

Wu, D., J.C. Lin, H.F. Duarte, V. Yadav, N.C. Parazoo, T. Oda, and E.A. Kort. 2021. A model for urban biogenic CO₂ fluxes: Solar-Induced Fluorescence for Modeling Urban biogenic Fluxes (SMUrF v1). Geoscientific Model Development 14:3633–3661. https://doi.org/10.5194/gmd-14-3633-2021

Acknowledgments

This work is based upon work supported by the National Aeronautics and Space Administration (grant 80NSSC19K0196).

Spatial Coverage: eastern and western CONUS, western Europe, eastern China, eastern Australia, eastern Asia, South America, and central Africa

Spatial Resolution: 0.05 degree

Temporal Coverage: 2010-01-01 to 2019-12-31

Temporal Resolution: 4-day mean GPP, 4-day mean SIF, daily mean Reco, and hourly mean GPP, Reco, and NEE fluxes

Site Boundaries: Latitude and longitude are given in decimal degrees.

Site	Westernmost Longitude	Easternmost Longitude	Northernmost Latitude	Southernmost Latitude
Entire Extent	-125	155	60	-40

Data File Information

There are 925 data files in NetCDF version 4 (*.nc4) format included in this dataset that provide 4-day means of gross primary production (GPP) and uncertainty with associated SIF measurements, daily estimates of ecosystem respiration (Reco) with uncertainty, and downscaled hourly mean net ecosystem exchange (NEE), GPP, and Reco.

Table 1. File names and descriptions. <yyyy> and <yyyymm> represent the year and year and month of the data file, respectively. See Table 2 for a description of each <region>.

File Name	Description
fourday_mean_SIF_GPP_uncert_<region>_<yyyy>.nc4 (e.g., fourday_mean_SIF_GPP_uncert_easternChina_2018.nc4)	4-day mean SIF and GPP uncertainties for each region, for each available year
daily_mean_Reco_uncert_<region>_<yyyymm>.nc4 (e.g., daily_mean_Reco_uncert_centralAfrica_201701.nc4)	Daily mean Reco uncertainties for each region, for each month of each available year
hrly_mean_GPP_Reco_NEE_<region>_<yyyymm>.nc4 (e.g., hrly_mean_GPP_Reco_NEE_westernCONUS_201602.nc4)	Hourly mean GPP, Reco, and NEE for each region, for each month for each available year

Data File Details

The no-data value is -999.

Table 2. Years of data available for each region.

Region	Year
central Africa, eastern Asia, eastern Australia, eastern China, and South America	2017, 2018
eastern CONUS, western CONUS	2010–2019
western Europe	2010–2014, 2017, 2018

Table 3. Variables included in all data files.

Variables	Dimension and Unit	Description
lon	degree_east	Longitude at cell center
lat	degree_north	Latitude at cell center
time	seconds since 1970-01-01 00:00:00Z	UTC time

Table 4. Variables in files named fourday_mean_SIF_GPP_uncert_<region>_<yyyy>.nc4.

Variable	Units/format	Description
SIF_mean	mW m−2 nm−1 sr−1	4-day mean clear-sky CSIF from Zhang et al. (2018)
GPP_mean	µmol m-2 s-1	4-day mean Gross Primary Production (best estimates) based on clear-sky CSIF and GPP- SIF slopes aggregated from 500 m
GPP_sd	µmol m-2 s-1	1-sigma uncertainty of the 4-day mean Gross Primary Productions (based on model-FLUXNET comparisons)

 Table 5. Variables in files named daily_mean_Reco_uncert_<region>_<yyyymm>.nc4.

Variable	Units/format	Description
Reco_mean	µmol m-2 s-1	Daily mean Ecosystem Respiration (best estimates) based on pretrained biome-specific neural network models and ERA5-based temperature fields
Reco_sd	µmol m-2 s-1	1-sigma uncertainty of Daily Mean Ecosystem Respiration (based on model-FLUXNET comparisons)

Table 6. Variables in files named hrly_mean_GPP_Reco_NEE_<region>_<yyyymm>.nc4.

Variable	Units/format	Description
GPP_mean	µmol m-2 s-1	Hourly mean Gross Primary Production (best estimates) using hourly downscaling factors based on hourly fields of surface solar radiation downwards from ERA5 reanalysis
Reco_mean	µmol m-2 s-1	Hourly mean Ecosystem Respiration (best estimates) using hourly downscaling factors based on hourly air temperature fields from ERA5 reanalysis
NEE_mean	µmol m-2 s-1	Hourly mean Net Ecosystem Exchanges (best estimates) based on hourly GPP and Reco

When estimating fossil fuel carbon dioxide emissions from observed CO₂ concentrations, the accuracy can be hampered by biogenic carbon exchanges during the growing season even for urban areas where strong fossil fuel emissions are found. While biogenic carbon fluxes have been studied extensively across natural vegetation types, biogenic carbon fluxes within an urban area have been challenging to quantify due to limited observations and differences between urban versus rural regions.

The grid-level hourly mean NEE fluxes were evaluated against (1) non-gap-filled measurements at 67 eddy-covariance (EC) sites in North America and Europe from FLUXNET during 2010-2014 (r>0.7 for most data-rich biomes); (2) independent observations at two urban vegetation and two crop EC sites over Indianapolis from Aug 2017 to Dec 2018 (r=0.75); and (3) an urban biospheric model based on fine-grained land cover classification within Los Angeles (r=0.83). In addition, error statistics are derived for the modeled 4-day mean GPP and predicted daily mean Reco (testing set), when FLUXNET observations are compared. Those error statistics include correlation coefficient (r), root-mean-square error (RMSE), absolute mean bias, and coefficient of variation (CV). Refer to Wu et al. (2021) for details.

Similar to many terrestrial biospheric models, SMUrF estimates gridded GPP, Reco, and NEE (Reco - GPP) fluxes based on biomes. The main required data streams include (1) 500 m MODIS-based annual land cover classification; (2) 0.05 degree spatiotemporally contiguous SIF product; (3) above-ground biomass (AGB) from GlobBiomass; (4) eddy-covariance (EC) flux measurements; and (5) gridded products of air and soil temperatures. Figure 2 provides a summary of the data products used.

Figure 2. Description of the data products and observations used. Modified from Figure 1 of Wu et al. (2021).

Biomes

The land cover types defined by the International Geosphere-Biosphere Programme (IGBP) from the MODIS product MCD12Q1 were adopted to inform biomes over global lands. The preprocessed version of MCD12Q1 v006 was accessed from Friedl and Sulla-Menashe (2019) with latitude and longitude grids from Application for Extracting and Exploring Analysis Ready Samples (AppEEARS). Twelve biomes include croplands (CRO), closed and open shrublands (CSHR, OSHR), deciduous/evergreen broadleaf/needleleaf forests (DBF, DNF, EBF, ENF), grasslands (GRA), mixed forests (MF), savannas (SAV), woody savannas (WSAV), and permanent wetlands (WET). Unfortunately, MCD12Q1 simply treats the entire urban area as one category (URB), while in reality both grass and trees exist within the MODIS-based URB category. Two special treatments are carried out over urban areas and croplands and are further explained below.

Fluxes

SMUrF first estimates the 4-day mean GPP and daily mean Reco and then performs temporal downscaling to arrive at the hourly NEE fluxes at 0.05 degree (NEE=Reco-GPP). Hourly scaling factors were derived from surface downward shortwave radiation and air temperature from the ECMWF ReAnalysis-5 (ERA5; Copernicus Climate Change Service Information, 2017) using methods similar to Fisher et al. (2016) and Olsen and Randerson (2004).

In general, 4-day mean GPP was calculated using 4-day mean clear-sky SIF from CSIF (Zhang et al., 2018) and estimated biome-specific GPP-SIF slopes (α). The daily mean Reco was predicted based on ERA5-based air and soil temperatures (Tair and Tsoil) as well as the modeled GPP fluxes via neural network technique.

4-Day Mean GPP Estimates

A spatiotemporally Contiguous SIF (CSIF; Zhang et al., 2018) product together with GPP fluxes from FLUXNET 2015 (Pastorello et al., 2017) were used to calculate biome-specific GPP-CSIF slopes. Over 80 global EC tower sites with screened data points (quality flag<3) from 2010 to 2014 were chosen to represent various biomes. CSIF offers a global 4-day mean SIF at the grid spacing of 0.05 degrees during 2000–2018 using the NN approach (Zhang et al., 2018). α values are then assigned to each 500 m grid cell with various biome types and aggregated to 0.05 degrees to match the spatial resolution of the CSIF product. The 4-day mean GPP fluxes are calculated as the product of clear-sky CSIF and biome-weighted mean α values.

Two special treatments were performed for urban areas and croplands

1) The areal estimates of 42 crop species from the Spatial Production Allocation Model (MapSPAM 2010V1.1; You et al., 2014) were adopted to estimate the relative C3:C4 fractions. Four of the selected 13 cropland FLUXNET sites are associated with higher C4 fractions (>50%) than the rest (<10%). Assuming a linear relationship between GPP and SIF, two different α values were obtained for C3 and C4 crops based on the aforementioned EC sites with higher or lower C4 fractions. Eventually, the α values over croplands are calculated using the MapSPAM-based C3:C4 fractions and the individual α values.

2) Since MCD12Q1 treats the entire urban areas as one category, an algorithm was developed to approximate vegetation types and fractions within cities. An empirical relationship between the 100 m AGB from GlobBiomass (Santoro et al., 2018) and the tree cover fraction (ftree=tree/vegetated) from a high-resolution NAIP-based land cover classification over Los Angeles (Coleman et al., 2020) was derived. The nonlinear AGB-ftree relation is applied to approximate the tree fractions over global cities. The AGB-based tree fractions yield more trees than the tree fraction derived from the MOD44B product over Los Angeles.

The relative non-tree vegetated fractions (fnon-tree=1-ftree) is simply split into half grass and half shrub. The relative tree fractions are further divided into five specific tree types (i.e., DBF, DNF, EBF, ENF, MF). Due to insufficient reported data on urban tree species, the share of each tree type in cities is approximated as a function of latitude based on climatology (e.g., high fractions of ENF over high-latitudes, EBF over tropical lands, and DBF plus MF over the mid-latitudes).

Finally, each 500 m urban grid cell is linked to a weighted mean α value according to the predicted tree/grass fractions and their associated α values. To account for the potential negative bias of ~14.5% in CSIF over cities (Zhang et al., 2018), we scaled up the urban α values by 1.145.

Refer to Wu et al. (2021) for additional information.

Coleman, R.W., N. Stavros, V. Yadav, and N. Parazoo. 2020. A simplified framework for high-resolution urban vegetation classification with optical imagery in the Los Angeles Megacity. Remote Sensing 12:2399. https://doi.org/10.3390/rs12152399

Copernicus Climate Change Service (C3S). 2017. ERA5: Fifth generation of ECMWF atmospheric reanalyses of the global climate. Copernicus Climate Change Service Climate Data Store (CDS). https://cds.climate.copernicus.eu/cdsapp, https://doi.org/10.24381/cds.bd0915c6. Accessed 2020-04-14.

Fisher, J.B., M. Sikka, D.N. Huntzinger, C. Schwalm, and J. Liu. 2016. Technical note: 3-hourly temporal downscaling of monthly global terrestrial biosphere model net ecosystem exchange. Biogeosciences 13:4271–4277. https://doi.org/10.5194/bg-13-4271-2016

Friedl, M., and D. Sulla-Menashe. 2019. MCD12Q1 MODIS/Terra+Aqua Land Cover Type Yearly L3 Global 500m SIN Grid V006. NASA EOSDIS Land Processes DAAC. https://doi.org/10.5067/MODIS/MCD12Q1.006

Olsen, S.C. and J.T. Randerson. 2004. Differences between surface and column atmospheric CO₂ and implications for carbon cycle research. Journal of Geophysical Research 109. https://doi.org/10.1029/2003JD003968

Pastorello, G., D. Papale, H. Chu, C. Trotta, D. Agarwal, E. Canfora, D. Baldocchi, and M. Torn. 2017. A New Data Set to Keep a Sharper Eye on Land-Air Exchanges. Eos. https://doi.org/10.1029/2017eo071597

Santoro, M. 2018. GlobBiomass - global datasets of forest biomass. PANGAEA - Data Publisher for Earth &amp; Environmental Science. https://doi.org/10.1594/PANGAEA.894711

Wu, D., J.C. Lin, H.F. Duarte, V. Yadav, N.C. Parazoo, T. Oda, and E.A. Kort. 2021. A model for urban biogenic CO₂ fluxes: Solar-Induced Fluorescence for Modeling Urban biogenic Fluxes (SMUrF v1). Geoscientific Model Development 14:3633–3661. https://doi.org/10.5194/gmd-14-3633-2021

You, L., S. Wood, U. Wood-Sichra, and W. Wu. 2014. Generating global crop distribution maps: From census to grid. Agricultural Systems 127:53–60. https://doi.org/10.1016/j.agsy.2014.01.002

Zhang, Y., J. Joiner, S.H. Alemohammad, S. Zhou, and P. Gentine. 2018. A global spatially Continuous Solar Induced Fluorescence (CSIF) dataset using neural networks. https://doi.org/10.5194/bg-2018-255