Posts by Anita Graser

It’s been a couple of busy weeks, with the QGIS 4.2 release and meetings and conferences all over the place before a few, hopefully quieter, weeks of summer break.

QGIS

First, the Austrian QGIS user group met online on 25 June. The topic was webmapping, with multiple users presenting their webmapping solutions, ranging from Lizmap to QGIS Cloud.

A few days later, QGIS 4.2 was released on 3 July 2026. This release is named Belém do Pará, after the Brazilian city that hosted both FOSS4G and a QGIS user meeting back in 2024. MundoGEO has the details on the naming, if you’re curious about the backstory. Worth noting: 4.2 “Belém do Pará” will be the next LTR, so if you’re using the long-term release, this is the one to watch for.

On a side note, while designing the Belém splash screen, it was interesting to see that historic maps of Belém don’t have north at the top. Instead, they’re rotated with north pointing either left or right. A nice little reminder that “north-up” is a convention, not a law of cartography.

AGIT 2026

This week, I made my way to Salzburg for AGIT, my favorite Austrian GIS conference.

On the first day, I took part in the AGEO Podium discussion, together with Andreas Hocevar (the father of OpenLayers), on the OGC API standards, since many users aren’t even aware of these new standards yet.

Thursday was talk day for me: I presented MobiML, a new Python library designed to streamline the development of machine learning workflows for trajectory data. I hope this library can help make Mobility Data Science more approachable and results more reproducible.

Right after my talk, Michael Szell presented “Assessing the Danish Bicycle Node Network”, building the data and algorithm foundation for active mobility planning and research. If you want to dig into the tools behind it, check out bikenetwork.dk and bikenetkit.org.

Michael is giving a full talk on this on Tuesday at the Complexity Science Hub in Vienna, if you want to hear more.

FOSS4G Europe, from the sidelines

Unfortunately, I missed FOSS4G Europe in Timișoara the week before, so I followed along via the #foss4ge2026 hashtag instead. Iván Sánchez’s talk on the BOSCO ruling, arguing that all government software must be explainable, is just one example of the talks I would have loved to see in person. There was also the already traditional QGIS Feature Frenzy by Kurt Menke and a QGIS hydrological analysis workshop by Hans who also has a full FOSS4G Europe 2026 summary worth reading.

Also relevant: the Birds of a Feather session on AI in OSGeo projects has spilled over onto the OSGeo discuss mailing list. Definitely a thread worth following or getting involved in if you maintain or contribute to open source geospatial projects.

What’s next

Besides MobiML, work also continues on the MovingPandas front. There are a few open pull requests I want to work through ahead of the next release.

After the summer break, the conference season picks back up quickly: FOSS4G 2026 in Hiroshima (30 August–5 September), Spatial Data Science across Languages (SDSL) 2026 in Jena (16–17/18 September), and the QGIS conference 2026 in Switzerland (5–6 October), where I’ll be speaking about AI in the QGIS ecosystem.

For a more complete picture of what is going on in geospatial worldwide, check out (and don’t forget to bookmark) Jakub‘s comprehensive list of geospatial conferences at github.com/Nowosad/geospatial-conferences.

Finally it’s here: Jupyter notebooks inside QGIS. I don’t know about you but I’ve been hoping for someone to get around to doing this for quite a while.

Qiusheng Wu published the first version of the Notebook plugin on 26 Dec 2025. Late Christmas present?!

For the setup, there’s a handy tutorial by Hans van der Kwast and, additionally, Qiusheng published an intro video:

Development is going fast (version 0.3.0 at the time of writing) so there will be new features when you install / update the plugin compared to both the tutorial and the video.

The user interface is pretty stripped down with just a few buttons to add new code or markdown cells and to run them. And there is a neat drop-down menu with all kinds of ready-made code snippets to get you started:

For other functionalities, for example, to delete cells, you need to right-click on the cell to access the function through the context menu. And, as far as I can tell, there is currently no way to rearrange cells (moving them up or down).

I also haven’t quite understood yet what kinds of outputs are displayed and which are not because – quite often – the cell output just stays empty, even though the same code generates output on the console:

Some of the plugin settings I would have liked to experiment with, such as adjusting the font size or enabling line numbers, don’t seem to work yet. So a little more patience seems to be necessary.

I’ll definitely keep an eye on this one :)

The journey continues: QgsArrowIterator is now merged! This makes it possible to iterate over QgsFeatures as Arrow batches.

This is where we are now, quoting Dewey Dunnington:

import geopandas
from nanoarrow.c_array import allocate_c_array
import qgis
from qgis.core import QgsVectorLayer

# Create a vector layer
layer = QgsVectorLayer("tests/testdata/zonalstatistics/polys.shp", "layer_name", "ogr")
schema = qgis.core.QgsArrowIterator.inferSchema(layer)

it = qgis.core.QgsArrowIterator(layer.getFeatures())
it.setSchema(schema, 1)

c_array = allocate_c_array()
schema.exportToAddress(c_array.schema._addr())
it.nextFeatures(5, c_array._addr())

print(geopandas.GeoDataFrame.from_arrow(c_array))
#> lev3_name                                           geometry
#> 0    poly_1  MULTIPOLYGON (((100.37934 -0.96049, 100.37934 ...
#> 1    poly_2  MULTIPOLYGON (((100.37944 -0.96044, 100.37955 ...
#> 2    poly_3  MULTIPOLYGON (((100.37938 -0.96049, 100.37949 ...

print(geopandas.read_file("tests/testdata/zonalstatistics/polys.shp"))
#> lev3_name                                           geometry
#> 0    poly_1  POLYGON ((100.37934 -0.96049, 100.37934 -0.960...
#> 1    poly_2  POLYGON ((100.37944 -0.96044, 100.37955 -0.960...
#> 2    poly_3  POLYGON ((100.37938 -0.96049, 100.37949 -0.960...

Further improvements are already being planned. To quote from the ticket:

“The final state after this improvement would be a compact way for Arrow Python consumers like GeoPandas to ergonomically consume a layer. Maybe:

geopandas.GeoDataFrame.from_arrow(qgis_layer_object)

Or maybe:

geopandas.GeoDataFrame.from_arrow(qgis_layer_object.getArrowStream())

Looking forward to seeing this develop further.

The conversation around Looking for better ways to convert between QGIS VectorLayer and (Geo)DataFrame is continuing over at https://fosstodon.org/@underdarkGIS/115442614331293320

What I’ve learned so far:

• QgsVectorLayer.as_geopandas() has landed in QGIS master on 13 Oct 2025.
• There’s also QgsVectorLayer.field_to_numpy() which will be useful for many applications and has landed on 29 Oct 2025.
• QgsArrowIterator is in the works right now.

Exciting times for spatial data science tooling

Plugin developers who want to use (Geo)Pandas-based functionality in their plugins regularly face the challenge of converting QGIS vector layers to (Geo)DataFrames. There is currently no built-in convenience function.

In Trajectools, so far, I have been performing the conversion manually, looping through all features and taking care of tricky column types, such as datetimes and geometries:

def df_from_layer_trajectools(layer,time_field_name="t"):
    # Original Trajectools 2.7 version
    names = [field.name() for field in layer.fields()]
    data = []
    for feature in layer.getFeatures():
        my_dict = {}
        for i, a in enumerate(feature.attributes()):
            if names[i] == time_field_name and isinstance(a, QDateTime):
                a = a.toPyDateTime()
            my_dict[names[i]] = a
        pt = feature.geometry().asPoint()
        my_dict["geom_x"] = pt.x()
        my_dict["geom_y"] = pt.y()
        data.append(my_dict)
    df = pd.DataFrame(data)
    return df

It works (mostly), but it’s far from fast. For the 25 million Geolife points, it takes 4 minutes:

In an attempt to speed-up (and make the conversion more robust, e.g. regarding datetime/timezone conversion and null values), I’ve spent some time at SDSL2025 with Joris Van den Bossche trying a workaround that writes the QGIS layer to an Arrow file and then reads that file with pyogrio:

def gdf_from_layer_arrow(layer):
    # SDSL2025 version
    with tempfile.TemporaryDirectory() as tmpdirname:
        path = os.path.join(tmpdirname, "data.arrow")

        options = QgsVectorFileWriter.SaveVectorOptions()
        options.actionOnExistingFile = QgsVectorFileWriter.CreateOrOverwriteFile 
        options.layerName = 'data'
        options.driverName = "arrow"
        
        QgsVectorFileWriter.writeAsVectorFormatV3(
            layer, path, QgsProject.instance().transformContext(), options
        )
       
        meta, table = pyogrio.read_arrow(path)
        gdf = gpd.GeoDataFrame.from_arrow(table)

    return gdf

Not only do we get a GeoDataFrame in return, this also runs in half the time, i.e. in 2 minutes instead of 4:

Switching to this approach will require adding pyogrio to the plugin dependencies. Looks like it could be worth it.

We also discussed another alternative: It would be faster to read the vector layer data source directly, in case it is a supported file format. However, this means we’d need separate handling for other input layers.

There’s also the issue of supporting the Processing feature that allows users to run the algorithm only on the selected features because selected features are only exposed through QgsProcessingParameterFeatureSource (and not through QgsProcessingParameterVectorLayer). Maybe the Export Selected Features algorithm can cover this case but it will export an empty layer if there is no selection.

Are you aware of any other / better ways to approach this issue? Any pointers are appreciated.

The last time I preprocessed the whole GeoLife dataset, I loaded it into PostGIS. Today, I want to share a new workflow that creates a (Geo)Parquet file and that is much faster.

The dataset (GeoLife)

“This GPS trajectory dataset was collected in (Microsoft Research Asia) Geolife project by 182 users in a period of over three years (from April 2007 to August 2012). A GPS trajectory of this dataset is represented by a sequence of time-stamped points, each of which contains the information of latitude, longitude and altitude. This dataset contains 17,621 trajectories with a total distance of about 1.2 million kilometers and a total duration of 48,000+ hours. These trajectories were recorded by different GPS loggers and GPS-phones, and have a variety of sampling rates. 91 percent of the trajectories are logged in a dense representation, e.g. every 1~5 seconds or every 5~10 meters per point.”

The GeoLife GPS Trajectories download contains 182 directories full of .plt files:

Basically, CSV files with a custom header:

Creating the (Geo)Parquet using DuckDB

DuckDB installation

Following the official instructions, installation is straightforward:

curl https://install.duckdb.org | sh

From there, I’ve been using the GUI which we can launch using:

duckdb -ui

The spatial extension is a DuckDB core extension, so it’s readily available. We can create a spatial db with:

ATTACH IF NOT EXISTS ':memory:' AS memory;
INSTALL spatial;
LOAD spatial;

Reading a spatial file is as simple as:

SELECT * 
FROM '/home/anita/Documents/Codeberg/trajectools/sample_data/geolife.gpkg'

thanks to the GDAL integration.

But today, we want to do to get a bit more involved …

DuckDB SQL magic

The issues we need to solve are:

1. Read all CSV files from all subdirectories
2. Parse the CSV, ignoring the first couple of lines, while assigning proper column names
3. Assign the CSV file name as the trajectory ID (because there is no ID in the original files)
4. Create point geometries that will work with our GeoParquet file
5. Create proper datetimes from the separate date and time fields

Luckily, DuckDB’s read_csv function comes with the necessary features built-in. Putting it all together:

CREATE OR REPLACE TABLE geolife AS 
SELECT 
  parse_filename(filename, true) as vehicle_id, 
  strptime(date||' '||time, '%c') as t, 
  ST_Point(lon, lat) as geometry -- do NOT use ST_MakePoint
FROM read_csv('/home/anita/Documents/Geodata/Geolife/Geolife Trajectories 1.3/Data/*/*/*.plt',
    skip=6,
    filename = true, 
    columns = {
        'lat': 'DOUBLE', 
        'lon': 'DOUBLE', 
        'ignore': 'INT', 
        'alt': 'DOUBLE', 
        'epoch': 'DOUBLE', 
        'date': 'VARCHAR',
        'time': 'VARCHAR'
    });

It’s blazingly fast:

I haven’t tested reading directly from ZIP archives yet, but there seems to be a community extension (zipfs) for this exact purpose.

Ready to QGIS

GeoParquet files can be drag-n-dropped into QGIS:

I’m running QGIS 3.42.1-Münster from conda-forge on Linux Mint.

Yes, it takes a while to render all 25 million points … But you know what? It get’s really snappy once we zoom in closer, e.g. to the situation in Germany:

Let’s have a closer look at what’s going on here.

Trajectools time

Selecting the 9,438 points in this extent, let’s compute movement metrics (speed & direction) and create trajectory lines:

Looks like we have some high-speed sections in there (with those red > 100 km/h streaks):

When we zoom in to Darmstadt and enable the trajectories layer, we can see each individual trip. Looks like car trips on the highway and walks through the city:

That looks like quite the long round trip:

Let’s see where they might have stopped to have a break:

If I had to guess, I’d say they stayed at the Best Western:

Conclusion

DuckDB has been great for this ETL workflow. I didn’t use much of its geospatial capabilities here but I was pleasantly surprised how smooth the GeoParquet creation process has been. Geometries are handled without any special magic and are recognized by QGIS. Same with the timestamps. All ready for more heavy spatiotemporal analysis with Trajectools.

If you haven’t tried DuckDB or GeoParquet yet, give it a try, particularly if you’re collaborating with data scientists from other domains and want to exchange data.

The QGISUC2025 team has done an awesome job recording and editing the conference presentations. All “presentation” type talks where the presenter has accepted to be published are now available in a dedicated list on the QGIS Youtube channel.

I also had the pleasure of presenting our Trajectools plugin and you can see this talk here:

Thank you to all the organizers, speakers, and participants for the great time!

The latest releases of MovingPandas and Trajectools come with many “under the hood” changes that aim to make your movement analytics faster:

1. Instead of immediately creating a GeoPandas GeoDataFrame and populating the geometry column with Point objects, MovingPandas now has “lazy geometry column creation” that holds off on this operation until / if the geometries are actually needed. This way, for many operations, no geometry objects have to be generated at all.
2. MovingPandas TrajectorySplitters now support parallel processing and Trajectools uses parallel processing whenever available (e.g. for adding speed & direction metrics, detecting stops, splitting trajectories).
3. When a minimum length is specified for trajectories, MovingPandas now avoids computing the total trajectory length and, instead, immediately stops once the threshold value has been reached (“early skip”).
4. Trajectools now offers the option to skip computation of movement metrics (speed & direction). This way, we can skip unnecessary computations and leverage the lazy geometry column creation, wherever applicable.

Let’s have a look at some example performance measurements!

Example 1: MovingPandas ValueChangeSplitter

The ValueChangeSplitter splits trajectories when it detects a value change in the specified column. This is useful, for example, to split up public trajectories that contain a “next_stop” column.

The following graph shows ValueChangeSplitter runtimes for different minimum trajectory length settings (from 0 to 1km, 100km, and 10,000km):

We see that the new, lazy geometry column initialization outperforms the old original code in all cases (e.g. 57% runtime reduction for 1km), except for the worst-case scenario, when the original implementation discards all trajectories as too short right from the start. (For most use cases, min_length will be set to rather small values to avoid creation of undesired short trajectory fragments, similar to sliver polygons in classic geometry operations.)

Additionally, we can engage multiprocessing by setting the n_processes parameter, e.g. to the number of CPUs to achieve further speedup:

Example 2: Trajectools

By applying all above-mentioned speedup techniques, Trajectools is now considerably faster. For example, the following runtime reductions can be achieved by deactivating the “Add movement metrics (speed, direction)” option in the algorithm dialog:

• Create trajectories: 62%
• Spatiotemporal generalization (TDTR): 78%
• Temporal generalization: 81%
• Split trajectories at stops: 53%

I have also updated the default trajectory points output style. It now uses a graduated renderer to visualize the speed values (if they have been calculated) instead of the previously used data-defined override. This makes the style faster to customize and provides a user-friendly legend:

For more infos, have a look at:

• MovingPandas 0.22 release notes
• Trajectools 2.6 release notes

Enjoy the latest performance increases!