Published July 25, 2024 | Version v1
Dataset Open

dataset for: Fire derived ozone intensifies carbon loss in Amazon forests under extreme droughts.

Description

Abstract

This dataset relates to work undertaken for publication 'Fire‑derived ozone intensifies carbon loss in Amazon forests under extreme droughts .'

Datasets and jupyter notebooks allow the figures and data to be reproduced.

 

ABSTRACT: The Amazon rainforest is a vital contributor to the removal of anthropogenic carbon dioxide from the atmosphere. Ozone (O3) is an atmospheric pollutant that reduces carbon dioxide uptake by vegetation. Yet the degree to which variability in ozone, elevated by anthropogenic-derived fire activity, and climate contribute to changes in Amazon forest productivity remains poorly understood. Here, we employ a land surface model to quantify the interannual variability of fire-derived ozone damage across the Amazon and explore its coupling with extreme drought and stomatal responses. We estimate that fire-derived ozone damage increases carbon losses by approximately one quarter of direct fire emissions, revealing a potentially important indirect pathway by which may fires amplify carbon loss. Notably, our simulations suggest high fire activity during extreme droughts, such as in 2024, likely result in ozone damage despite extensive stomatal limitation, highlighting the urgency to reduce deforestation and forest fire as the climate changes.

Methods (English)

Model description

This dataset relates to simulated data produced using a land surface model. The study used the Joint UK Land Environment Simulator (JULES) vn. 5.6 at a spatial resolution of 1.25° latitude by 1.875° longitude.

Simulations

To isolate the effect of O3-plant damage, we compared two simulations for the period 1997–2014: one with and one without O3-plant damage. The simulation that included O3-plant damage used an average O3 susceptibility for tropical trees based on an average response observed in tropical trees. The simulation without O3-plant damage was identical, except with the damage parameter set to 0. Both simulations underwent separate 1000-year spin ups (cycling 50 times over a 20-year period from 1975 − 1995) in which O3 was fixed at 1975 concentrations. Only changes during the period 1997–2014 were evaluated due to data availability of fire emissions (beginning in 1997) and high temporal-resolution O3 concentrations (available until 2014).

Other datasets

For comparison to observations, we also use fire data (The Global Fire Emissions Database 4 including small fires (GFED4s) was used for fire carbon emissions); NOx column data (Satellite-derived atmospheric NOx columns were taken from the TROPOMI instrument on the Sentinel-5 satellite); and GPP data (GOSIF, the Global OCO-2 based SIF product, used sun-induced fluorescence (SIF) to infer GPP, and Fluxsat, a satellite product used vegetation index (EVI) from MODIS and calibrated with FLUXNET 2015 GPP).

Analysis

All datasets were regridded to the resolution of the JULES simulation (1.25° latitude by 1.875° longitude), if they were not already available at that resolution, and the Amazon region was selected. This was defined by the tropical broadleaf tree fraction from the 2015 land cover map produced by Harper et al. 2016 . Analysis was limited to the period 1997–2014, constrained by GFED fire data beginning in 1997 and modelled O3 data ending in 2014.

To obtain correlations between variables (O3, fire activity, total land carbon), annual mean gridded data were averaged over the Amazon region, weighted by fraction of tropical broadleaf tree plant functional type and gridcell area. These values were presented as anomalies relative to the 1997–2014 mean.

To establish drought effects, monthly means were calculated for the northern and southern Amazon by dividing datasets at the equator, weighted by fraction of tropical broadleaf tree and gridcell area. Data were then divided into drought years (1998, 2005, 2007, 2010) and non-drought years. The average of drought years was compared to the average of non-drought years.

For comparison to GPP satellite products, GPP from simulations and satellite data was masked over all gridcells with less than 90% tropical tree cover. Annual means were detrended to remove impacts of CO2 fertilisation that were included in the JULES simulations but cannot be detected by the EVI product. For evaluation of seasonality, JULES GPP was compared to satellite-derived GPP in the central, less seasonal part of the Amazon (4°S–4°N) and the outer, more seasonal part of the Amazon (> 4°S, > 4°N) as well as separating into drought and non-drought years.

Table of contents

This dataset provides netCDF files and jupyter notebooks that can be used to recreate figures from the publication of the same name. In the study, we identify drivers of interannual variability in O3-plant damage in the Amazon using a land surface model validated against satellite products.

Data is presented in netCDF format for 4 main figures and 8 supplementary figures. Metadata attached to the netCDF file describes its contents. Below gives an overview of the data and scripts needed to recreate the figures.

Description of the data and file structure

Fig.1: 

  • Fig.1 (Jupyter notebook)
  • Annual mean detrended GPP for simulations with and without O3 damage (2 netCDFs)
  • Annual mean detrended GPP derived from GOSIF and FluxSat products (2 netCDFs)

Fig.2, Fig. S4, Fig. S8:

  • These figures are similar and use some shared datasets
  • Fig.2, Fig. S4, Fig. S8 (Jupyter notebooks)
  • Annual mean simulated O3 concentration from UKESM1 (3 netCDFs)
  • Annual mean simulated O3 concentration from 5 other ESMs (1 netCDF)
  • Annual mean simulated fire carbon emissions from GFED4s (1 netCDF)
  • Annual mean simulated O3 damage under present day and elevated CO2 (2 netCDFs)

Fig. 3, Fig.4, Fig.S5, Fig.S6:

  • These figures are similar and use some shared datasets
  • Fig.3, Fig. 4, Fig. S5 , Fig. S6 (Jupyter notebooks)
  • Fraction of tropical tree in each grid cell (1 netCDF)
  • Spatial monthly mean O3 distribution (1 netCDF)
  • Spatial monthly mean O3 flux into stomata (1 netCDF)
  • Spatial monthly mean NPP with and without O3 damage (2 netCDFs)

Fig.S1:

  • Fig.S1 (Jupyter notebook)
  • Annual mean NO2 column from UKESM1 (simulated) and OMI (satellite) (2 netCDFs)

Fig.S2:

  • Fig.S2 (Jupyter notebook)
  • Spatially explicit NO2 column from UKESM1 (simulated) and OMI (satellite) (2 netCDFs)

Fig.S3:

  • Fig.S3  (Jupyter notebook)
  • GPP from outer amazon region (greater than 4N-4S) from simulations, GOSIF and FluxSat (8 .npy files)
  • GPP from central amazon region (within 4N-4S) from simulations, GOSIF and FluxSat (8 .npy files)

Sharing/Access information

Links to other publicly accessible locations of the data:

Code/Software

To analyse the data, Python 3 with Jupyter notebooks on the JASMIN sever was used: https://help.jasmin.ac.uk/docs/interactive-computing/jasmin-notebooks-service/. The scripts are provided to recreate the data using the following packages:

  • numpy vn 1.26.4
  • matplotlib vn 3.8.4
  • iris vn 3.9.0
  • cartopy vn 0.23.0
  • scipy vn 1.13.0

Files

Brown_et_al_data.zip

Files (27.2 MB)

Name Size Download all
md5:cbda0d362167c659b95d91cca37601d5
27.2 MB Preview Download

Additional details

Funding

Natural Environment Research Council
NERC GW4+ DTP NE/R001812/1
Met Office
CSSP Brazil Climate Science for Service Partnership

Dates

Submitted
2024-07-25

Software

Programming language
Python , Jupyter Notebook , Jasmin