What’s New in cartopy 0.19




21st Apr 2021

For a full list of included Pull Requests and closed Issues, please see the 0.19 milestone.


  • Thomas Grainger restored PEP-517 support to improve installations. (PR #1681)

  • @emsterr added the ability to style bounding boxes of labels. (PR #1669)

  • Adrien Berchet added the ability to cache downloaded image tiles. (PR #1533)

  • Greg Lucas changed the vector interpolations to be strictly in the source coordinates, which removed some erroneous extrapolations. (PR #1636)

  • Giacomo Caria added an option to remove the cardinal direction labels from the axes. (PR #1662)

  • Greg Lucas added the ability to update data within a pcolormesh plot using set_array() to enable animations of the fields. (PR #1496) Liam Bindle extended this capability to update the color limits (PR #1655) and Sebastian David Eastham fixed the return values when get_array() was called (PR #1656)

  • @htonchia and Greg Lucas fixed an issue with large cells appearing in pcolormesh plots. (PR #1622)

  • Philippe Miron added an example to demonstrate how to label specific sides of the plot. (PR #1593)

  • Greg Lucas added the option to restrict the limits of gridlines. (PR #1574)

  • Kyle Penner fixed extrapolations using an alpha-channel in imshow(). (PR #1582)

  • Valentin Iovene added pkg-config instructions to help with installations on MacOS. (PR #1596)

  • Luke Davis updated the tight bbox calculations to include the gridliner labels. (PR #1355)

  • Luke Davis fixed the label padding for gridliners to use points which makes the rendered screen image appear the same as the printed image now. (PR #1556)

  • Daryl Herzmann added the ability to make Hexbin plots. (PR #1542)

  • Kyle Penner fixed image plotting when a 2D alpha array is input. (PR #1543)

  • Elliott Sales de Andrade and Hugo van Kemenade removed Python 2 support. (PR #1516, PR #1517, PR #1540, PR #1544, and PR #1547)

What’s New in cartopy 0.18




3rd May 2020

For a full list of included Pull Requests and closed Issues, please see the 0.18 milestone.


  • We are very pleased to announce that Greg Lucas has been added to the cartopy core development team. Greg (@greglucas) added the NightShade feature in the previous release, and has been instrumental in issue and PR triage leading up to 0.18. He has also ensured that CI systems have kept working through various upstream project changes.

  • Kevin Donkers and Phil Elson made the AdaptiveScalar the default for Natural Earth Features. This will make the default features look much nicer when plotting on zoomed in axes. (PR #1105)

  • Elliott Sales de Andrade added support for Matplotlib 3.2 and 3.3 (PR #1425) and Python 3.7 and 3.8 (PR #1428).

  • Alan Snow added the ability to use Proj version 6.x (PR #1289) and Elliott Sales de Andrade updated a lot of the tests and build issues for this upgrade (PR #1417).

  • Andrew Huang added the ability to put the meridian and parallel gridline labels on the gridlines within the plot boundaries rather than only as labels on the boundary. (PR #1089)

  • Stephane Raynaud added longitude and latitude labeling to all projections. It was previously restricted to the Mercator and PlateCarree projections. (PR #1117)

  • Phil Elson added the (long awaited!) ability to label contours on GeoAxes. A Contour labels example has been added to the gallery demonstrating the new capability. (PR #1257)

  • Matthew Bradbury added the ability to query UK Ordnance Survey image tiles. (PR #1214)

  • Phil Elson added the ability to fetch image tiles using multiple threads. (PR #1232)

  • Elliott Sales de Andrade added a cartopy.mpl.geoaxes.GeoAxes.GeoSpine class to replace the cartopy.mpl.geoaxes.GeoAxes.outline_patch that defines the map boundary. (PR #1213)

  • Elliott Sales de Andrade improved appearance of plots with tight layout. (PR #1213 and PR #1422)

  • Ryan May fixed the Geostationary projection boundary so that geometries no longer extend beyond the map domain. (PR #1216)

  • Phil Elson added support for style composition of Features. This means that the styles set on a Feature when it is created, and when it is added to an Axes, will be processed consistently.


What’s New in cartopy 0.17




16th Nov 2018

For a full list of included Pull Requests and closed Issues, please see the 0.17 milestone.


  • The cartopy.feature.NaturalEarthFeature class now allows a cartopy.feature.AdaptiveScaler object to be passed as the scale argument. This will automatically choose the appropriate feature scale from the GeoAxes extent. This can also be used interactively while panning and zooming in a figure. cartopy.feature.NaturalEarthFeature.scale is now read-only. (PR #1102, PR #983)

  • Proj version 5.x is now supported in Cartopy, thanks to hard work by Elliott Sales de Andrade. As part of making this version work, the inner workings and boundaries of many projections were improved. (PR #1124, PR #1148) Elliott also improved support for warped rectangular projections (PR #1180) as well as added support for the Eckert family of projections (PR #1168) and Equal Earth projection. (PR #1182)

  • Greg Lucas contributed functionality to plot day/night across the globe, which was turned into a map feature by Phil Elson. The shading can be added to a map with cartopy.feature.nightshade.Nightshade(datetime). For more information, see the Nightshade feature example. (PR #1135, PR #1181)

  • Elliott Sales de Andrade added optional support for the use of pykdtree when performing image transformations. This module has been demonstrated to be twice as fast as the old code for most of the Cartopy examples, with one example (geostationary) having a 95% reduction in run time. (PR #1150)

  • Greg Lucas added a Fiona-based shapefile reader. If Fiona is installed on a user’s system, this will now be the default shapefile reader, adding significant speed improvements. (PR #1000)

  • Phil Elson added the ability to control the appearance of Shapely geometries using a function. cartopy.mpl.geoaxes.GeoAxes.add_geometries() gained a styler argument that takes a function that given a geometry, returns a dictionary of style keyword arguments. The Hurricane Katrina example has been updated to use this. (PR #1019)

  • Kevin Donkers, with help from Phil Elson and Peter Killick, improved the interactivity of panning and zooming images by adding a raster image cache. (PR #1192, PR #1195, PR #1197)

  • Peter Killick and Phil Elson improved the use of Cartopy in Jupyter notebook environments by adding an HTML representation for projections. These render vector images of the coastlines using a given projection to enable a quick preview. (PR #951, PR #1196)

  • Fixes were added by Elliott Sales de Andrade to support the Matplotlib 3.x series. (PR #1130)

  • Ryan May fixed up the Geostationary and NearsidePerspective projections as well as added additional options to the Mercator projection. (PR #1189, PR #1043)

  • Andrey Kiselev contributed support for the Equidistant Conic projection. (PR #1022)

  • Peter Killick updated and improved the interface to Mapbox image tiles. (PR #1170)

  • Manuel Garrido and Phil Elson collaborated to add support for more themes for the Stamen map tile set. (PR #1013, PR #1188)

  • Support for WMTS sources was made more robust by Alex Crosby. (PR #1052, PR #1053)

  • Passing a color argument to cartopy.mpl.geoaxes.GeoAxes.add_feature() now overrides default feature edgecolor and facecolor thanks to a change by Elliott Sales de Andrade. (PR #1029)

  • Phil Elson added cartopy.geodesic.Geodesic.geometry_length() to calculated the length in physical meters of any Shapely geometry. (PR #1096)

  • Elliott Sales de Andrade improved the interpolation code by normalizing values, reducing issues due to precision. (PR #1042)

  • Ryan May fixed a few corner cases in the plotting and transform code. (PR #1062, PR #1090)

  • A pyproject.toml file has been added to Cartopy by Elliott Sales de Andrade to make it easier to build Cartopy. Newer versions of pip should now automatically install Cython and NumPy before trying to build Cartopy. (PR #1132)

  • Andrew Dawson fixed a crash when calculating the boundary for the Lambert Azimuthal Equal Area projection. (PR #1100)

  • Elliott Sales de Andrade and Andrew Dawson removed the use of deprecated functionality in NumPy. (PR #1101, PR #1122)

  • Kevin Donkers added all 60 UTM zones to the images in the supported projection documentation. (PR #1103)

  • Broken URLs to the SRTM imagery were corrected by Elliott Sales de Andrade. (PR #1143)


  • cartopy.mpl.clip_path.clip_path() has been deprecated. It is a simple wrapper for Matplotlib’s path clipping, so use that instead. You can replace clip_path(subject, clip_bbox) by subject.clip_to_bbox(clip_bbox).

  • cartopy.io.img_tiles.StamenTerrain has been deprecated. Use Stamen('terrain-background') instead.

  • In CartoPy 0.18, the default value for the origin argument to cartopy.mpl.geoaxes.GeoAxes.imshow() will change from 'lower' to 'upper' to match the default in Matplotlib.

Incompatible Changes

  • Support for Matplotlib < 1.5.1 and NumPy < 1.10 has been removed.

What’s New in cartopy 0.16




21st Feb 2018


  • We are very pleased to announce that Ryan May has been added to the cartopy core development team. Ryan (@dopplershift) brings a wealth of experience, and has already made significant contributions to the Matplotlib interface, extended projections, and helped modernise the development infrastructure.

  • The Gnomonic projection was brought up-to-date to include the central_longitude argument. (PR #855)

  • Ryan May improved the formulation of the boundary ellipse for the Geostationary projection and added the sweep_angle_axis keyword argument. (PR #890, PR #897)

  • Elliott Sales de Andrade made a number of micro-optimisations to the Matplotlib interface, fixed a number of documentation issues with Python 3 and added Matplotlib 2.0 & 2.1 compatibility. (PR #886, PR #901, PR #780, PR #773, PR #977)

  • Tick padding was added to the gridliner. cartopy.mpl.gridliner.Gridliner.xpadding and ypadding relate. (PR #783)

  • Ryan May added the with_scale() method to the NaturalEarthFeature class. For example, it is now possible to access higher resolution land features with cartopy.feature.LAND.with_scale('50m'). In addition to this, cartopy.feature.STATES was added to easily access administrative area boundaries that were previously only accessible by manually constructing NaturalEarthFeature instances (as is done in the Feature Creation example). (PR #898)

  • Daryl Herzmann and Robert Redl improved cartopy’s internal conversion between Shapely objects and Matplotlib Paths. (PR #885 & PR #1021)

  • Åsmund Steen Skjæveland fixed cartopy.mpl.geoaxes.GeoAxes.tissot() to use the documented units of kilometres, where before it had been using metres. (PR #904)

  • Andrew Dawson wrote a new tutorial for the user guide: Understanding the transform and projection keywords. (PR #914)

  • Daniel Kirkham and Daryl Herzmann made significant improvements to the stability of polygon transformation. The changes reduce the frequency of messages such as Unidentified problem with geometry, linestring being re-added and Self-intersection at or near point <X> <Y> occurring. (PR #974 and PR #903)

  • Chris Holdgraf and Corinne Bosley worked collaboratively to bring sphinx-gallery to the cartopy docs. (PR #969)

  • Ray Bell neatened up many of the examples to explicitly pass the coordinate system when calling set_extent(). (PR #975)

  • Ryan May changed the default zorder of LAND and OCEAN to -1, thus fixing an issue with LAND/OCEAN appearing above some data elements such as vectors. (PR #916)

  • Kevin Donkers added the 60 UTM projections example to the gallery in PR #954:

  • Andrey Kiselev added support for reading shapes with a third (Z) dimension. (PR #958)

  • Corinne Bosley standardised the docstring format for improved readability and visual consistency. (PR #987)

  • Cartopy now no longer enables shapely.speedups() at cartopy import time. (PR #990)

  • Mahé Perrette and Ryan May collaborated to improve the Stereographic projection. (PR #929)

What’s New in cartopy 0.15




1st February 2017


  • The cartopy.crs.Mercator class now allows a latitude_true_scale to be specified.

  • A tiles url can now be passed directly to the cartopy.io.img_tiles.GoogleTiles class.

  • The background_img() method has been added. This allows users to add a background image to the map, from a selection of pre-prepared images held in a directory specified by the CARTOPY_USER_BACKGROUNDS environment variable.

  • The Web Map Tile Service (WMTS) interface has been extended so that WMTS layers can be added to GeoAxes in different projections.

  • The NearsidePerspective projection has been added.

  • Optional keyword arguments can now be supplied to the add_wmts() method, which will be passed to the OGC WMTS gettile method.

  • New additions to the gallery:


What’s New in cartopy 0.14




24th March 2016


  • Zachary Tessler and Raj Kesavan added the cartopy.crs.Sinusoidal projection, allowing MODIS data to be visualised in its native projection. Additionally, a prepared cartopy.crs.Sinusoidal.MODIS projection has been made available for convenience.

  • Joseph Hogg and Daniel Atton Beckmann added the cartopy.geodesic.Geodesic class which wraps the proj.4 geodesic library. This allows users to solve the direct and inverse geodesic problems (calculating distances between points etc). It also contains a convenience function that returns geodetic circles. This is used by cartopy.mpl.geoaxes.GeoAxes.tissot() which draws Tissot’s indicatrices on the axes.

  • The SRTM3 data source has been changed to the LP DAAC Data Pool. The Data Pool is more consistent, fixing several missing tiles, and the data is void-filled. Consequently, the cartopy.srtm.fill_gaps() function has been deprecated as it has no purpose within the STRM context. The SRTM example has also been updated to skip the void-filling step. Additionally, this data source provides SRTM at a higher resolution of 1 arc-second, which may be accessed via cartopy.io.srtm.SRTM1Source.

  • All downloaders will use secure connections where available. Not every service supports this method, and so those will use non-secured HTTP connections instead. (See PR #736 for full details.)

  • Cartopy now supports, and is tested against, Matplotlib 1.3 and 1.5 as well as NumPy 1.7, 1.8 and 1.10.

  • Daniel Eriksson added a new example to the gallery:


Incompatible changes


  • cartopy.crs.GOOGLE_MERCATOR has been moved to cartopy.crs.Mercator.GOOGLE.

What’s new in cartopy 0.13




30th June 2015


What’s new in cartopy 0.12




14th April 2015


  • We are very pleased to announce that Elliott Sales de Andrade was added to the cartopy core development team. Elliott has added several new projections in this release, as well as setting up cartopy’s Python 3 testing on TravisCI and generally improving the cartopy codebase.

  • Installing cartopy became much easier for conda users. A scitools channel has been added which makes getting cartopy and all of its dependencies on Linux, OSX and Windows possible with:

    conda install -c scitools cartopy
  • Support for Python 3, specifically 3.3 and 3.4, has been added. Some features that depend on OWSLib will not be available as it does not support Python 3.

  • Two new projections, AzimuthalEquidistant and AlbersEqualArea have been added. See the Cartopy projection list for the full list of projections now available in cartopy.

  • The Web Map Service (WMS) interface has been extended to support on-the-fly reprojection of imagery if the service does not support the projection of the map being drawn. The following example demonstrates the process by adding WMS imagery to an Interrupted Goode Homolosine map - unsurprisingly this WMS service does not provide IGH imagery, so cartopy has had to reproject them from a projection the WMS does support:

  • Peter Killick added an interface for accessing MapBox tiles using the MapBox Developer API. A MapBox client can be created with, MapboxTiles and as with the other imagery from a simple URL based imagery service, it can be added to a GeoAxes with the add_image() method. The following example demonstrates the interface for another source of imagery:

  • Some improvements were made to the geometry transformation algorithm to improve the stability of geometry winding. Several cases of geometries being incorrectly inverted when transformed have now been resolved. (PR #545)

  • Mark Hedley added the central_rotated_longitude keyword to cartopy.crs.RotatedPole, which is particularly useful for limited area rotated pole models in areas such as New Zealand:

  • A new method has been added to the GeoAxes to allow control of the neatline of a map drawn with the Matplotlib interface. The method, set_boundary(), takes a matplotlib Path object, which means that arbitrary shaped edges can be achieved:

  • A new SRTM3 RasterSource has been implemented allowing interactive pan/zoom of 3 arc-second elevation data from the Shuttle Radar Topography Mission. The SRTM example has also been updated to use the new interface.

  • New additions to the gallery:



  • The SRTM module has been re-factored for simplicity and to take advantage of the new raster source interface. Some methods have therefore been deprecated and will be removed in future releases. The function cartopy.io.srtm.srtm() has been replaced with the cartopy.io.srtm.SRTM3Source.single_tile() method. Similarly, cartopy.io.srtm.srtm_composite() and cartopy.io.srtm.SRTM3_retrieve() have been replaced with the cartopy.io.srtm.SRTM3Source.combined() and cartopy.io.srtm.SRTM3Source.srtm_fname() methods respectively.

  • The cartopy.io.RasterSource.fetch_raster interface has been changed such that a sequence of cartopy.io.LocatedImage must be returned, rather than a single image and its associated extent.

  • The secant_latitudes keyword in cartopy.crs.LambertConformal has been deprecated in favour of standard_parallels.

What’s new in cartopy 0.11




19 June 2014

  • Richard Hattersley added epsg() support for generating a Cartopy projection at run-time based on the EPSG code of a projected coordinate system. This mechanism utilises https://epsg.io/ as a coordinate system resource and employs EPSG request caching using pyepsg

  • Phil Elson added WMSRasterSource which provides interactive pan and zoom OGC web services support for a Web Map Service (WMS) aware axes. This capability may be added to an axes via the add_wms() method. Generic interactive slippy map panning and zooming capability is managed through the new SlippyImageArtist and use of the add_raster() method.

  • WMTSRasterSource was added by Richard Hattersley to provide interactive pan and zoom OGC web services support for a Web Map Tile Service (WMTS) aware axes, which is available through the add_wmts() method. This includes support for the Google Mercator projection and efficient WMTS tile caching. This new capability determines how to match up the available tiles projections with the target projection and chooses the zoom level to best match the pixel density in the rendered image.

  • Thomas Lecocq added functionality to cartopy.io.srtm allowing intelligent filling of missing elevation data, as well as a function to compute elevation shading for relief style mapping. An example has been added which uses both of these functions to produce a grayscale shaded relief map

  • Lion Krischer extended the capability of GoogleTiles to allow support for street, satellite, terrain and street_only style Google Map tiles.

  • Nat Wilson’s contribution brought us a major step closer to Python 3 compatibility.

  • Support for the UTM projection was added by Mark Hedley.

  • Andrew Dawson has added a new convenience utility function add_cyclic_point() to add a cyclic point to an array and optionally to a corresponding 1D coordinate.

  • Andrew Dawson added formatters for producing longitude/latitude tick labels for rectangular projections. The formatters are customizable and can be used to produce nice tick labels in a variety of styles:


What’s new in cartopy 0.10




17 January 2014

We are very pleased to announce that Andrew Dawson was added to the cartopy core development team. In this release Andrew has single-handedly implemented comprehensive vector transformation and visualisation capabilities, including:

  • The ability to transform vector fields between different coordinate reference systems via the transform_vectors() CRS method.

  • GeoAxes.quiver and GeoAxes.barbs for arrow and barb plotting. More information is available at Vector plotting.

  • A regridding function for “regularising” a vector field in the target coordinate system. See also cartopy.vector_transform.vector_scalar_to_grid(). Both quiver() and barbs() accept the regrid_shape keyword to trigger this behaviour automatically.

  • GeoAxes.streamplot adds the ability to draw streamlines in any projection from a vector field in any other projection.


What’s new in cartopy 0.9




12 September 2013

  • We are very pleased to announce that Bill Little was added to the cartopy core development team. Bill has made some excellent contributions to cartopy, and his presentation at EuroScipy’13 on “Iris & Cartopy” was voted best talk of the conference.

  • Other talks and tutorials during this release cycle include Phil Elson’s talk at SciPy’13 (with video), Thomas Lecocq’s tutorial at EuroSciPy and a forthcoming talk at FOSS4G.

  • Christoph Gohlke updated cartopy to support Windows 7.

  • The Plate Carree projection was updated to fully handle arbitrary globe definitions.

  • Peter Killick updated the Mercator class’ default globe to WGS84. His refactor paved the way for some follow on work to fully implement the Google Spherical Mercator (EPSG:3857) projection.

  • The TransverseMercator class saw a tidy up to include several common arguments (pull request)

  • Bill Little added the Geostationary projection to allow geolocation of satellite imagery.

  • Byron Blay added the Lambert conformal conic projection.

What’s new in cartopy 0.8




3 June 2013

  • Bill Little added support for the OSNI projection and enhanced the image nest capability. (PR #263)

  • cartopy.io.img_nest.Img has been extended to include a cartopy.io.img_nest.Img.from_world_file() static method for easier loading of georeferenced images.

  • Phil Elson added a major performance improvement when plotting data from PlateCarree onto a PlateCarree map. (PR #260)

  • Byron Blay and Richard Hattersley added a cartopy.crs.Globe class to encapsulate ellipsoid and optionally datum information for CRSs. Globe handling in many projections, including Stereographic, has been added.

What’s new in cartopy 0.7




21 Mar 2013

This is a quick release which targets two very specific requirements. The goals outlined in the development plan at v0.6 still remain the primary target for v0.8 and beyond.

What’s new in cartopy 0.6




19 Feb 2013

  • Patrick Peglar added the ability to draw ticks for some limited projections when using the gridlines() method on an Axes.

  • Phil Elson and Carwyn Pelley extended the cartopy documentation to include new tutorials such as Using the cartopy shapereader.

  • Ian Edwards added a new example to create a favicon for cartopy.

  • Phil Elson added a new example to show polygon analysis and visualisation with Shapely and cartopy.

  • Edward Campbell added a new cartopy.crs.EuroPP projection for UTM zone 32.

  • Andrew Dawson added a central_longitude keyword for the Stereographic family of projections.

  • Phil Elson added a Downloader class which allows automatic downloading of shapefiles (currently from Natural Earth and GSHHS). The extension requires no user action and can be configured via the cartopy.config dictionary.

Development plans for cartopy 0.7 and beyond

  • Improve the projection definitions to support better control over datum definitions and consider adding WKT support (ticket).

  • Begin work on vector field support (barbs, quiver, streamlines etc.).

  • Continue identifying and implementing performance enhancements (particularly in contour drawing).

  • Extend the number of projections for which it is possible to draw tick marks.

What’s new in cartopy 0.5




7 Dec 2012

This document explains the new/changed features of cartopy in version 0.5.

Release 0.5 of cartopy continues the work to expand the feature-set of cartopy to encompass common operations, and provide performance improvements.

Cartopy 0.5 features

A summary of the main features added with version 0.5:

  • An improved feature API to support future expansion and sophistication, and a wider range of pre-defined Natural Earth datasets.

Incompatible changes



  • The method Axes.natural_earth_shp() has been replaced by the method Axes.add_feature() and the cartopy.feature module.

Feature API

A new features API is now available, see Using the cartopy shapereader.