Overview#

climepi is a Python package for combining ensemble climate projections with models of climate suitability for vector-borne diseases (VBDs) to analyze the impacts of climate change and its uncertainties on VBD risks. climepi is designed with a modular architecture, allowing users to apply and extend its functionality to a wide range of climate datasets and climate-sensitive epidemiological models.

Core dependencies#

climepi relies heavily on the xarray package for working with labelled multi-dimensional arrays and datasets. We highly recommend familiarizing yourself with the xarray documentation before using climepi.

Other key dependencies include:

  • Dask: used under the hood by xarray for parallel computing and handling large datasets (see https://docs.xarray.dev/en/stable/user-guide/dask.html).

  • xCDAT: a package for climate data analysis with xarray, providing (among other features) methods for temporal averaging and bounds handling that are wrapped by climepi.

  • pymc: used for parameterizing climate-VBD suitability models by fitting temperature response curves of vector and pathogen traits to laboratory data.

  • hvPlot: a high-level plotting API which is used within climepi’s plotting methods.

  • Panel: used to build the browser-based front-end application.

Conventions#

climepi’s features assume that climate datasets are represented as xarray Dataset objects and follow certain naming conventions:

  • Data variables describing temperature and precipitation values (where present) should be named ‘temperature’ and ‘precipitation’, respectively. Inbuilt example epidemiological models assume temperature values in °C and precipitation values in mm/day, respectively.

  • The time dimension should be named ‘time’.

  • Longitude and latitude coordinates should be named ‘lon’ and ‘lat’, respectively. Note that some climepi methods return datasets with a dimension ‘location’ indexing named locations.

  • For ensemble datasets, multiple realizations of a single climate model under a single scenario should be indexed by a dimension named ‘realization’ (typically provided as integer values starting from 0). If multiple models and/or scenarios are included, they should be indexed by dimensions named ‘model’ and ‘scenario’, respectively.

  • If temporal bounds are present, they should be included as a data variable named ‘time_bnds’, with dimensions ‘time’ and ‘bnds’.

We welcome requests or contributions for relaxing these requirements (for example, using cf_xarray) or converting other data structures (such as climate time series stored as pandas DataFrame objects) to compliant xarray Dataset objects (see Contributing).

Functionality#

Climate data subpackage#

The climdata subpackage contains methods for retrieving climate projection data. It provides a single interface for obtaining projections of key climate variables used in climate-VBD models (currently, temperature and precipitation) from various sources. Currently supported data sources are:

  • LENS2 (Community Earth System Model version 2 Large Ensemble), which provides 100 ensemble members for analyzing internal climate variability.

  • ISIMIP (Inter-Sectoral Impact Model Intercomparison Project Phase 3b), which provides downscaled and bias-adjusted data from multiple CMIP6 climate models and scenarios.

  • ARISE-SAI (Assessing Responses and Impacts of Solar intervention on the Earth system with Stratospheric Aerosol Injection), which supports analysis of the impacts of solar climate intervention.

  • GLENS (Geoengineering Large Ensemble), which also supports analysis of the impacts of solar climate intervention.

For example, LENS2 data for London and Paris from 2030 to 2100 for the first two ensemble members can be retrieved as follows:

from climepi import climdata

ds_clim = climdata.get_climate_data(
    data_source="lens2",
    frequency="daily",
    subset={
        "years": list(range(2030, 2101)),
        "locations": ["London", "Paris"],
        "realizations": [0, 1],
    },
)

See the documentation for climdata.get_climate_data() for details of data sources and subsetting options. Example datasets are also provided (see the documentation for climdata.get_example_dataset()).

Epidemiological model subpackage#

The epimod subpackage contains classes and methods for climate-sensitive epidemiological models, particularly models describing climate suitability for VBDs. For example, a simple model in which transmission is possible within a certain temperature range (in the example below, in the range 15-30 degrees celsius) can be defined as follows:

from climepi import epimod

suitability_model = epimod.SuitabilityModel(temperature_range=(15, 30))

Running the model on a climate dataset (either using suitability_model.run(), or via the climepi accessor as described below) will then yield a dataset with a Boolean data variable ‘suitability’ indicating whether or not each temperature value falls within the specified range).

Methods are also provided for inferring temperature responses of vector and pathogen traits in order to construct suitability models in the ParameterizedSuitabilityModel class. See Gallery for detailed usage examples.

Accessor class for xarray datasets#

The climepi accessor class for Dataset objects provides methods for running epidemiological models on climate datasets, and for analyzing and visualizing the impact of climate uncertainty sources, as well as other utility methods.

The accessor class can be used by chaining the climepi attribute to a Dataset. For example, the run_epi_model() method can be used to run an epidemiological model on a climate dataset as follows:

import climepi # noqa

ds_epi = ds_clim.climepi.run_epi_model(suitability_model)

Acknowledgement#

This package has been developed as part of a project funded by a Digital Technology Development Award (Climate-Sensitive Infectious Disease Modelling) from Wellcome (grant number 226057/Z/22/Z).

Wellcome logo