Landuse mapping with OpenStreetMap in Belgium

What are the (best) practices for land-use mapping adapted to the Belgian landscape?

OpenStreetMap (OSM) is not just a map but rather a big geodatabase with a free tagging model for classification of geographical features. It can be used for land-use mapping, as described in the dedicated wiki page here. For instance, the recent project OSM-Landuse specifically aims at producing land-use maps from OSM data.

Because of its free tagging scheme and its incredible generalist model (OSM is a map of … everything), do not expect a clear classification scheme of land-use/land-cover as in other land-use maps (such as in Corine Land Cover or GlobCover)! A note about the tagging system in OSM: each geographical features (e.g., a house, a road, a city or a forest) is characterised by a couple of keyvalue words, that forms a tag. For instance, a residential road inside an urban area is tagged as “highway=residential”, where “highway” is the key, i.e., a kind of category that spans from small paths to motorways, and “residential is the value of the key. The dedicated tag for mapping land-use in OSM is “landuse=*” but other tags (notably: natural=*) can also be used.

A difficulty in land-use mapping and classification is the definition of common classes that encompasses various landscapes and ecoregions. As a global project, OSM uses tags for mapping land-use worldwide. Therefore, it is important to specify how to use these tags for land-use mapping in a specific ecoregion/country. The purpose of this article is to discuss the best practices in land-use mapping in Belgium with OSM.

Forests

The dedicated tag for forests is landuse=forest. Then, more information about the forest can be added using leaf_cycle=* & leaf_type=*. In Belgium, one can encountered:

  • Broadleaved deciduous forests, tagged as forest=landuse; leaf_cycle=deciduous, leaf_type=broadleaved, mainly formed by beeches, oaks, populus, maples, birches, …
Beechs trees in the Ardenne
  • Needle-leaved evergreen (coniferous) forests, tagged as forest=landuse; leaf_cycle=evergreen, leaf_type=needleleaved, mainly formed by fir trees, Douglas, and pines.
Picea abies trees in the Ardenne
  • Needle-leaved deciduous forests, tagged as forest=landuse; leaf_cycle=deciduous, leaf_type=needleleaved, much more less common than the first two types, formed by larch trees (Larix sp.)
Larix sp. trees

Further notes about forests:

  • IMHO, there are no broadleaved evergreen forests in Belgium, even though some trees are broadleaved evergreen (think about holly trees – Ilex Aquifolium), they do not form large chunks of forests in Belgium.
  • We can also use the tags leaf_cycle=mixed and leaf_type=mixed for mixed forests types. These tags are actually very common in Belgium and used to describe large forests where some parcels are needleleaved-evergreen and other parcels broadleaved-deciduous. In a ideal (OSM) world, we should map separately parcels according to their leaf cycle and leaf type, since “real” mixed forest are not so common in Belgium.
  • No matter the size of the forest, in OSM you can use the above-mentioned combinations of tags for mapping small, isolated patches of forest formed by a dozen of trees up to large forested areas.
  • We can also use the tag taxon=* to further specify the tree species that are grown in the forest. This can be used not only for monospecific forests (taxon=fir trees) but also for multi-species landscapes: in this case, separate the species names by commas (taxon = oak, beech, ash ).
  • The tag wood=* is deprecated and should be replaced by leaf_cycle=* & leaf_type=*.
  • The tag natural=wood is sometimes used for forests, but since natural=wood would refers to unmanaged, natural forests, IMHO it should not be used in Belgium since no more forests are completely natural: every patches of forest in Belgium has experienced human interventions in the recent history.

Agricultural lands

There is no clear consensus about which tags we should use to map agricultural lands in Belgium. Actually, it’s a recent discussion on the Belgian talk mailing list about mapping farmlands that gave me the idea of writing this article.

Agricultural lands are occupied by annual or perennial crops (croplands), or by meadows (grasslands). Meadows can be hay meadows (cut 1, 2, or several times a year for hay producing) or grazed meadows with animals.

Croplands

Just planted potato field near Gembloux

Meadows

A grazed meadow with cows near Nassogne

Meadows also fall under the definition of landuse=farmland. Grazed meadows could be tagged as landuse=farmland + animal=yes and hay meadows using landuse=farmland + crop=grass.

However, there also exists the landuse=meadow tag for specifically mapping meadows. The OSM wiki is not clear about which tag use, here’s an excerpt of the wiki page of landuse=farmland: “Also note that many mappers prefer the more specific tags landuse=meadow for meadows and pastures (…), landuse=orchard for fruit orchards, and use landuse=farmland for cropland only.” Some mappers consider that meadows can be easily ploughed and becomes croplands and so map grasslands using the generic tag landuse=farmland.

However many mappers prefers to use the landuse=meadow over the landuse=farmland. Some Belgian meadows can be considered as permanent (> 10 years according to FAO definition). This is particularly true in Ardennes, Pays de Herve where some meadows are probably several decades old. In addition, some meadows could not easily be ploughed because of the presence of stones, trees or because they are too wet.

Note also the meadow=agricultural tag to specify that the meadow is used for agricultural purpose.

I would propose the following tagging for agricultural grasslands in Belgium:

And if you know more about the way it is managed:

More thought about agricultural lands

The issue is that agricultural land affectation often change from one year to another! Most croplands do actually change of crops every year because of the agricultural benefits to make such crop rotations. Furthermore, most croplands are also covered by intermediate cover crops between two crops, since it is imposed by agro-environmental regulations. And sometimes, agricultural lands do change from croplands to meadows and vice-versa.

Further notes:

  • Farms buildings and around should be tagged as landuse=farmyard.
  • Orchards, a kind of perennial crop, should be tagged as landuse=orchard. They are particularly common in some places of Belgium (Sint-Truidden, Gembloux)
  • Hedges around fields can be mapped as barrier=hedge, fences as barrier=fence.

 

Other vegetated lands

Hautes-Fagnes in Belgium

Compared to wilder countries, Belgium is a very intensively managed country, with a few lands that is really made of spontaneous vegetation. Anyway, we can encounter some lands that are not forest or agricultural lands: marsh, wetlands, shrubs… For these lands, you can consider the following tags:

Note that often, on aerial imagery, some lands appears as scrub while there are actually forest lands in a process of regeneration after a clear cut. So should we mapped them as landuse=forest or natural=scrub? To me it depends on the case. If I make a ground survey and see newly planted trees or fast regenerating trees, then I’d map it as a forest.

Other landuse

I won’t cover the tagging of urban areas in this article. Yet, many other landuse values apply to residential, industrial, commercial areas, but also the leisure=*, amenity=* and tourism=* keys. See the landuse page for further information.

About land-use geometries & way-of-mapping

  • How to draw land-use polygons in Belgium? Usually, large forested areas are mapped as one large polygon, which often has to be a multipolygon (a polygon with holes, or more complex polygons). Some large portions of agricultural lands in Belgium are represented as multipolygons, but more often, they are represented as small patches of lands, as they actually appear in reality, corresponding to a single management unit. I do prefer this way of tagging over the large multipolygon approach, because it allows to quickly change attributes of farmland in the future, when, e.g., a meadow is changed to a cropland.
  • There are some discussions about the way to connect a land-use polygon to the other elements of the map. Should the land-use polygon be stuck to nearby roads (or track, or river) or should we separate the land-use polygon from the road by a small space? I prefer the last option, as there is often something between a road and an agricultural land: a fence, a hedge, a ditch that can also be mapped as a separate way in OSM.

Conclusion: land-use mapping in Belgium (with OSM)

There is actually still a lot of land-use elements to map in Belgium, especially in Wallonia. In other well-mapped areas, land-use would need a serious “refresh”, since land can change of affectation from time to time. Pure mapping applications benefit from the land-use information from OSM, for instance the maps of OpenTopoMap. With a larger completness, OSM could be also increasingly used in environmental research and operational applications, such land-use change monitoring, in relation with vegetation and climate modelling. Land-use/land-cover is a hot topic in environmental research and the deployment of new generations of earth observation satellite such as Sentinel 2 offers increasing opportunities for land-use mapping. Nevertheless, automatic classification of satellite imagery still needs on-ground validation and usually could not compete with grounded, crowdsourced information from a dense and active local community. Combining crowdsourced land-use information from OSM with satellite imagery classification is certainly the way to go.

 

Julien Minet

PS: An interesting discussion followed this blog post at the Belgian OSM mailing list: here are the e-mails. I’d to summarize this discussion whenever I’ll have some time.

 

Photos credits (Flickr): Marc Dufrene (cropland), kaskitewatim (picea abies forest), helenetchandjiabo (grazed meadow), rire&vivre (Fagus sylvatica trees), peupleloup (Larix trees), bushman_k (Hautes Fagnes)

 

 

Crowdsourcing, citizen science, and (many) other buzz words

As I’m currently working on a article about the development of crowdsourcing projects, I found dozen of buzz words about similar concepts in the field of participatory/citizen science. Here’s my list of these buzz words: crowdsourcing, crowd science, citizen science, citizen sensing, community-based monitoring, community-based management, open science, participatory science, participatory sensing, user-generated content, volunteered geographic information, and how they are related to each other…

EDIT SEPT. 2017:

The paper is out now: Minet, J., Curnel, Y., Gobin, A., Goffart, J. P., Mélard, F., Tychon, B., Wellens, J. & Defourny, P. (2017). Crowdsourcing for agricultural applications: A review of uses and opportunities for a farmsourcing approach. Computers and Electronics in Agriculture, 142, 126-138.

–> https://doi.org/10.1016/j.compag.2017.08.026

Let’s start by crowdsourcing

The first definition ever of crowdsourcing was made in Wired by J. Howe (2006), who first coined crowdsourcing as “the act of a company or institution taking a function once performed by employees and outsourcing it to an undefined (and generally large) network of people in the form of an open call”.

Still, “crowdsourcing is an ill-defined but common term” (Wiggins & Crowston, 2011) and dozens of other definitions were made after the one of J. Howe. The outsourcing of tasks by crowdsourcing can be envisaged as an alternative to the traditional labour executed by professional workers, as defined by Schenk & Guittard (2011): “Crowd sourcing represents the act of taking a job traditionally performed by a designated agent (usually an employee) and outsourcing it to an undefined, and generally large group of people in the form of an open call”, or for tackling tasks that computer intelligence struggles to complete, as in the definition of Rahman et al. (2015): “a network of humans can be used to solve challenging and computationally expensive problems that machine intelligence cannot accurately solve”.

From crowdsourcing to citizen sensing and community-based monitoring

The concept now expands from the outsourcing of tasks to the collection of data by volunteers and/or mobile devices, as stated by Muller et al. (2015): “Crowdsourcing is traditionally defined as obtaining data or information by enlisting the services of a (potentially large) number of people. However, due to recent innovations, this definition can now be expanded to include ‘and/or from a range of public sensors, typically connected via the Internet.’” … that is related to the concept of citizen sensing, as developed in Boulos et al. (2011), which is about the collection of data or information by volunteers. A term largely related to the one of participatory sensing, coined by Burke et al. (2006), focused on the use of mobile devices for collecting data: “Participatory sensing will task deployed mobile devices to form interactive, participatory sensor networks that enable public and professional users to gather, analyze and share local knowledge.”

Yet, the action of collaboratively collecting data did exist well before the invention of the crowdsourcing term (and of mobile phones), as portrayed in the paper done by Conrad & Hichley (2011), who reviewed projects based on community-based monitoring or community-based management of the environment involving citizen volunteers in environmental conservation projects.

Internet-based platforms

These projects are now mostly based on internet platforms, which is related to the concept of user-generated content platforms, from which Wikipedia is the outstanding example. In the field of (neo)geography, a specific term was also coined to describe the development of online collaborative mapping platforms (e.g., OpenStreetMap, Wikimapia): volunteered geographic information (Goodchild, 2007).

The global picture: citizen science

Finally, the more embracing concept of citizen science can be linked to crowdsourcing, as stated by Wiggins & Crowston (2011): “Citizen science projects that are entirely mediated by information and communication technologies (ICTs) are often considered a form of crowdsourcing applied to science.” Franzoni & Sauermann remarkably discussed these concepts in a broader perspective for the scientific research in their paper entitled “Crowd science: The organization of scientific research in open collaborative projects” (2014) and further distinguished between crowdsourcing and crowd science, the latter term describing research projects which are not only open to participation (as in crowdsourcing projects) but also imply the openness of intermediate inputs and results (as in open science projects). Yet, these “new” ways of doing science are also related to the longer-standing concept of participatory science. 

References

Boulos, M. N. K.; Resch, B.; Crowley, D. N.; Breslin, J. G.; Sohn, G.; Burtner, R.; Pike, W. A.; Jezierski, E. & Chuang, K.-Y. S. (2011), ‘Crowdsourcing, citizen sensing and sensor web technologies for public and environmental health surveillance and crisis management: trends, OGC standards and application examples’, International journal of health geographics 10(1), 1.

Burke, J. A.; Estrin, D.; Hansen, M.; Parker, A.; Ramanathan, N.; Reddy, S. & Srivastava, M. B. (2006), ‘Participatory sensing’, Center for Embedded Network Sensing.

Conrad, C. C. & Hilchey, K. G. (2011), ‘A review of citizen science and community-based environmental monitoring: issues and opportunities’, Environmental monitoring and assessment 176(1-4), 273–291.

Franzoni, C. & Sauermann, H. (2014), ‘Crowd science: The organization of scientific research in open collaborative projects’, Research Policy 43(1), 1–20.

Goodchild, M. F. (2007), ‘Citizens as sensors: the world of volunteered geography’, GeoJournal 69(4), 211–221.

Howe, J. (2006), ‘The rise of crowdsourcing’, Wired magazine 14(6), 1–4.

Muller, C.; Chapman, L.; Johnston, S.; Kidd, C.; Illingworth, S.; Foody, G.; Overeem, A. & Leigh, R. (2015), ‘Crowdsourcing for climate and atmospheric sciences: current status and future potential’, International Journal of Climatology 35(11), 3185–3203.

Rahman, M.; Blackwell, B.; Banerjee, N. & Saraswat, D. (2015), ‘Smartphone-based hierarchical crowdsourcing for weed identification’, Computers and Electronics in Agriculture 113, 14–23.

Schenk, E. & Guittard, C. (2011), ‘Towards a characterization of crowdsourcing practices’, Journal of Innovation Economics & Management(1), 93–107.

Wiggins, A. & Crowston, K. (2011), From conservation to crowdsourcing: A typology of citizen science, in ‘System Sciences (HICSS), 2011 44th Hawaii international conference on’, pp. 1–10.

 

***

Note: You may use this text for your own work, but please respect the terms of the Creative Commons license By Sa:

Creative Commons License
Crowdsourcing, citizen science, and (many) other buzz words, by Julien Minet, nobohan.be, May 2016, is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Save

Save

Save