projectresilience/ELUC-committed

Project Resilience Emissions from Land-Use Change Dataset

Project Resilience

To contribute to this project see Project Resilience (Github Repo).

The goal of Project Resilience is "to build a public AI utility where a global community of innovators and thought leaders can enhance and utilize a collection of data and AI approaches to help with better preparedness, intervention, and response to environmental, health, information, or economic threats to our communities, and contribute the general efforts towards meeting the Sustainable Development Goals (SDGs)."

This dataset was used in the paper: Discovering Effective Policies for Land-Use Planning at the NeurIPS 2023 Workshop: Tackling Climate Change with Machine Learning. A recorded presentation on the paper can be found here.

A preliminary demo utilizing this dataset was presented at the United Nations AI For Good Summit 2023 in Geneva, Switzerland on the main stage as well as in a workshop

Sources

This dataset contains estimates for CO2 Emissions from Land-Use Change (ELUC) provided by the Bookkeeping of Land-Use Emissions (BLUE) model. BLUE contributes ELUC estimates to the annually updated Global Carbon Budget (eg. to Global Carbon Budget 2023)

Note: this dataset contains committed land-use change emissions from BLUE, which is different than the ELUC data used in the Global Carbon Budget. Details are found below.

BLUE uses land-use change data from the Land-Use Harmonization Project 2 (LUH2), which provides gridded data on land-use changes from 850-2100.

Please contact Clemens Schwingshackl or Julia Pongratz from Ludwig-Maximilians-Universität Munich, Germany with any questions about the ELUC data from BLUE.

Land-Use Types in LUH2

• Primary: Vegetation that is untouched by humans

  • primf: Primary forest
  • primn: Primary nonforest vegetation

• Secondary: Vegetation that has been touched by humans

  • secdf: Secondary forest
  • secdn: Secondary nonforest vegetation

• Urban

• Crop

  • c3ann: Annual C3 crops (e.g. wheat)
  • c4ann: Annual C4 crops (e.g. maize)
  • c3per: Perennial C3 crops (e.g. banana)
  • c4per: Perennial C4 crops (e.g. sugarcane)
  • c3nfx: Nitrogen fixing C3 crops (e.g. soybean)

• Pasture

  • pastr: Managed pasture land
  • range: Grazed natural grassland, savannah, etc.

Dataset

The processed dataset is indexed by latitude, longitude, and time, with each row consisting of the land use of a given year, the land-use change from year to year+1, and the committed ELUC at the end of year in tons of carbon per hectare (tC/ha).

Committed ELUC means that the CO2 fluxes in the year of the land-use change event and all subsequent CO2 fluxes (e.g. due to decay of biomass after clearing or due to regrowth of forest after wood harvest or re/afforestation) are summed and attributed to the year of the event.

In addition, the cell area of the cell in hectares and the name of the country the cell is located in are provided.

A crop and crop_diff column consisting of the sums of all the crop types and crop type diffs is provided as well as the BLUE model treats all crop types the same.

The processed dataset was created by:

• Joining the 2 raw data files by index
• Shifting all diff columns back 1 year so they align with their corresponding ELUC
• Aggregating the crop columns into a single column
• Adding country names to each cell

Raw

Raw data files are provided as: merged_aggregated_dataset_1850_2022.zarr.zip and BLUE_LUH2-GCB2022_ELUC-committed_gridded_net_1850-2021.nc, which are the land-use changes and the committed emissions respectively.

The file BLUE_LUH2-GCB2022_ELUC-committed_gridded_net_1850-2021.nc contains committed emissions from BLUE (variable ELUC). It is indexed by latitude, longitude, and time with a spatial resolution of 0.25°, covering the years 1850-2021. The file also contains the cell area of the cell in hectares (variable cell_area) and the name of the country the cell is located in are provided.

---

dataset_info: features:

• name: ELUC_diff dtype: float32
• name: c3ann dtype: float32
• name: c3ann_diff dtype: float32
• name: c3nfx dtype: float32
• name: c3nfx_diff dtype: float32
• name: c3per dtype: float32
• name: c3per_diff dtype: float32
• name: c4ann dtype: float32
• name: c4ann_diff dtype: float32
• name: c4per dtype: float32
• name: c4per_diff dtype: float32
• name: cell_area_diff dtype: float32
• name: pastr dtype: float32
• name: pastr_diff dtype: float32
• name: primf dtype: float32
• name: primf_diff dtype: float32
• name: primn dtype: float32
• name: primn_diff dtype: float32
• name: range dtype: float32
• name: range_diff dtype: float32
• name: secdf dtype: float32
• name: secdf_diff dtype: float32
• name: secdn dtype: float32
• name: secdn_diff dtype: float32
• name: urban dtype: float32
• name: urban_diff dtype: float32
• name: ELUC dtype: float32
• name: cell_area dtype: float32
• name: country dtype: float64
• name: crop dtype: float32
• name: crop_diff dtype: float32
• name: country_name dtype: string
• name: time dtype: int64
• name: lat dtype: float64
• name: lon dtype: float64 splits:
• name: train num_bytes: 6837499488 num_examples: 41630020 download_size: 3195082319 dataset_size: 6837499488 configs:
• config_name: default data_files:
  • split: train path: data/train-*

---