Overview

This module produces monthly land use/cover fraction maps for 15 SSP-RCP combinations to be used as input to the LISFLOOD hydrological model with the TransientLandUseChange option enabled. The maps account for yearly changes in the forest, sealed, irrigation (no rice), and irrigated rice fractions. The water fraction represents a monthly climatology and thus does not change yearly. The maps are resampled and subsetted to match the resolution and area of the template map (located at the templatemap_path specified in the configuration file). The data are produced for the period spanning year_start to year_end and saved to output_folder (all specified in the configuration file). The raw data sources start in 2015 and 2020 and are provided at 5- and 10-year intervals. The module linearly interpolates between subsequent maps to ensure gradual changes from year to year. For years prior to 2015, the module performs nearest-neighbour extrapolation and thus simply uses the earliest available data from 2015 and 2020. The fractions are output to netCDF-4 format (one file for each RCP-SSP combination and fraction).

Data

The module requires four datasets (Chen et al., 2020, GCAM-Demeter, GSWE, and HILDA+) as input:

1. Chen et al. (2020) urban land expansion projections under SSP1–5 (1-km resolution; 2020–2100). Each map was reprojected to 0.01° using gdalwarp -t_srs EPSG:4326 -tr 0.01 0.01 -r near -te -180.0 -90.0 180.0 90.0 -of GTiff <ssp>/global_<ssp>_<year>.tif <ssp>/global_<ssp>_<year>_reprojected.tif, where <ssp> is SSP1, SSP2, ..., SSP5 and <year> is 2020, 2030, ..., 2100. The files should be put in chen_2020_urban_folder.
2. GCAM-Demeter land use change projections for 15 SSP-RCP combinations (0.05° resolution; 2015–2100). Downloading all files at once did not work as the download kept getting interrupted. Downloading each SSP separately did work but required a lot of clicking. The files should be put in gcam_demeter_folder.
3. JRC GSWE open water extent data (30-m resolution; 1985–2019). We used a version of the GSWE resampled to 1-km resolution by Susann Guenther available on the JRC network and on request from Hylke Beck. The files should be put in gswe_folder.
4. HILDA+ historic land use change dataset (1-km resolution; 1900–2019). The files should be put in hildaplus_folder.

The locations of the datasets are specified in the configuration file. The module also requires a template map which defines the output resolution and area. The template map should be in netCDF-4 format and contain lat and lon variables and a data variable (any name). The location of the template map is specified using templatemap_path in the configuration file.

Methods

The module consists of four scripts:

1. step1_Chen_2020_urban.py: Resamples the reprojected Chen et al. (2020) sealed fraction projections under SSP1–5 to the template resolution.
2. step2_GCAM_Demeter.py: Resamples the GCAM-Demeter forest, irrigation, and rice fraction projections for 15 SSP and RCP combinations to the template resolution.
3. step3_JRC_GSWE.py: Resamples the JRC GSWE water fraction to the template resolution and computes a monthly climatology. HILDA+ data are used at high latitudes.
4. step4_harmonization.py: Loops through the 15 SSP-RCP combinations, years, and months and harmonizes the different data sources such that the sum of all fractions equals one. For each loop, the script:
   1. loads the Chen et al. (2020) fracsealed data (interpolating between years);
   2. loads the GSWE fracwater data (monthly climatology);
   3. reduces fracwater if fracsealed+fracwater >1 (fracsealed overrides fracwater);
   4. computes fracother as residual of fracsealed and fracwater;
   5. loads GCAM-Demeter fracforest, fracirrigation, and fracrice data (interpolating between years);
   6. reduces fracforest, fracirrigation, and fracrice if sum exceeds fracother;
   7. recalculates fracother as the residual; and
   8. subsets all data to the template map area.

System requirements

The script can be run on a normal desktop PC (Windows and Linux) with 16 GB or more of physical memory.

Instructions

Clone the repository:

git clone https://github.com/hylken/lisflood-land-cover-future
cd lisflood-land-cover-future

Produce a configuration file with the correct paths and folders based on the provided template.

Create and activate a Conda environment and run the script as follows:

conda create --name <env> --file requirements.txt
conda activate <env>
python main.py <config file>

If the environment creation step fails, we recommend creating the environment and installing the packages as follows:

conda create -n <env> -c conda-forge geopandas h5py pandas numpy netcdf4 matplotlib rasterio scikit-image