QGIS: maps for development data

Layers, the map canvas and coordinate systems

A QGIS project (a .qgz file) holds a list of layers and how each is drawn. The project file stores the path to your data, so keep the project and its data in one folder and move them together.

• The Layers panel lists the layers. A layer higher in the list is drawn on top. Tick a layer to show or hide it.
• The Browser panel shows folders and data sources; you can drag a file from it onto the map.
• The map canvas draws the visible layers.
• A vector layer holds features (points, lines or polygons) with an attribute table: one row per feature. District boundaries are polygons, one row per district.

Coordinate reference systems

Every layer has a coordinate reference system (CRS) that says how its numbers map to places on the earth. DataMeet's boundaries are in WGS84, EPSG:4326 (latitude and longitude in degrees), as their project page states. QGIS shows the project's CRS at the bottom right of the window. For a district choropleth of India, WGS84 is fine to start with; if you measure areas or distances, reproject to a projected CRS first, because degrees are not a unit of length.

Opening the attribute table

Select a layer in the Layers panel and choose Layer ► Open Attribute Table, or right-click it and choose Open Attribute Table, or press F6. Do this every time you load or join something: the table is where mistakes show first.

Load district boundaries

Where to get boundaries, and their licences

• DataMeet, Community Created Maps of India (projects.datameet.org/maps). The district boundaries page says the dataset is "shared under Creative Commons Attribution 2.5 India license" (CC BY 2.5 IN), and gives an attribution line to use. The repository README says that, unless stated otherwise, its datasets are under CC BY 4.0, so read the licence stated for the folder you download: the files differ. The district file sits at Districts/Census_2011/ in the repository, and its README says the names and extents come from the Census of India 2011 Administrative Atlas. The project page warns that the data "is not perfect".
• GADM (gadm.org, version 4.1 on the download page). Its licence page says the data "are freely available for academic use and other non-commercial use. Redistribution or commercial use is not allowed without prior permission." It allows maps made with GADM data in published academic articles. For an NGO report or a consultancy deliverable, check with GADM first.

Load the layer

1. Download the district shapefile from DataMeet. A shapefile is several files with the same name and different extensions (.shp, .shx, .dbf, .prj, sometimes more). Keep them together in one folder; the .shp alone is useless.
2. In QGIS choose Layer ► Add Layer ► Add Vector Layer… (Ctrl + Shift + V), or open the Data Source Manager with Ctrl + L.
3. Select the .shp file (or a .geojson file, which is a single file) and click Add.
4. Open the attribute table (F6). In the DataMeet 2011 district file the fields are DISTRICT, ST_NM (state name), ST_CEN_CD, DT_CEN_CD and censuscode. When we read the file on 6 October 2026 it held 641 districts.

Join a CSV to the boundaries

A choropleth needs one value per district in the boundary layer's attribute table. Your survey figures are in a separate table, so you join them, matching rows by a field the two tables share.

Make the district table

The course files are households.csv (240 households) and districts.csv (10 districts). Illustrative data, invented for teaching The district names are real places; every number is made up, so a map of them says nothing about the real districts. The cell summarises the households to one row per district and prints it as CSV for you to save.

Each district has 24 households. Mean monthly per capita expenditure runs from Rs 2,395.0 in Purnia to Rs 6,757.5 in Kozhikode, and the share of households with a toilet from 50.0% in Rewa to 83.3% in Indore, Kozhikode and Patna. Copy the CSV lines into a text file called district_summary.csv.

Load the CSV and join it

1. Open the Data Source Manager (Ctrl + L) and choose the Delimited Text tab. Browse to district_summary.csv, choose CSV as the format, and under geometry choose No geometry (attribute only table). Add it. It appears in the Layers panel as a table.
2. Open the boundary layer's properties: double-click the layer, or right-click it and choose Properties…. Go to the Joins tab and click the Add new join button.
3. Set Join layer to district_summary, Join field to district and Target field to DISTRICT.
4. Under Joined fields, tick only the columns you need, and set a short Custom field name prefix such as s_ so the new columns are easy to find. Click OK, then OK again.
5. Open the attribute table. The ten districts now carry values; the other 631 rows hold NULL in the joined columns. That is correct here, because the survey covers ten districts.

The QGIS manual describes this join as one-to-one, based on an attribute that both layers share, so the values must match exactly, letter for letter.

The name-mismatch problem

District names in survey and MIS files rarely match a boundary file exactly. The cell below joins a hand-typed MIS table to the ten names as they appear in the DataMeet file.

Only 5 of 10 districts match. Betul, Gaya, Kozhikode, Patna and Purnia get no value, because the MIS typed Baitul, Gaya with a trailing space, Calicut, PATNA and Purnea. QGIS would give no warning: those districts would simply be blank on the map, which reads as "no data" or, with some colour ramps, as "lowest".

After trimming spaces, fixing case and a three-line lookup, all 10 match. The second part shows the other trap: Aurangabad is a district in both Bihar and Maharashtra (the DataMeet 2011 file has both, along with other repeated names such as Bilaspur, Hamirpur and Pratapgarh), so joining on the name alone turns one row into two. Join on district and state, or better, on a code.

Choropleth styling and classification

1. Open the boundary layer's properties and go to Symbology.
2. Change the renderer at the top from Single Symbol to Graduated.
3. Set Value to the joined field, for example s_mean_pc_exp.
4. Pick a Color ramp: one hue from light to dark for a quantity where more is more.
5. On the Classes tab, set the number of classes and the Mode, then click Classify. The Histogram tab shows where the class breaks fall on the distribution of values.

The classification modes

The QGIS manual lists these modes for a graduated renderer:

• Equal Interval: every class covers the same range of values.
• Equal Count (Quantile): every class holds the same number of features.
• Natural Breaks (Jenks): "the variance within each class is minimized while the variance between classes is maximized".
• Also: Fixed Interval, Logarithmic scale, Pretty Breaks (round numbers) and Standard Deviation.

Same data, different modes, different maps. The cell below puts the ten district means into four classes three ways. The breaks come from code written for this course (numpy quantiles, and an exact search for natural breaks); QGIS computes its own, which may differ slightly, so compare them with the Histogram tab.

The three methods disagree. Equal interval (breaks at about Rs 3,486, 4,576 and 5,667) puts six districts in the lowest class, three in the second, none in the third and only Kozhikode in the top, because Kozhikode's Rs 6,757.5 stretches the range. Quantile puts Indore, Wayanad and Kozhikode together in the top class, although Kozhikode's mean is Rs 2,800 above Wayanad's, and it separates Udaipur (class 3) from Indore (class 4), which are Rs 147 apart. Natural breaks (about Rs 2,569, 2,962 and 3,957) keeps Udaipur, Indore and Wayanad together and leaves Kozhikode alone at the top. Three honest-looking maps of the same ten numbers would tell three different stories about Indore.

The chart draws each district as a dot on one line and the class breaks as dashed lines. In the equal-interval panel two of the three breaks fall in the empty stretch between Wayanad and Kozhikode; in the quantile panel the breaks crowd into the cluster of low values; the natural-breaks panel ends each class at the last district before a wide gap, so its lines sit on Patna, Betul and Wayanad.

Why the choice matters

• Equal interval is easy to read ("Rs 500 bands"), and honest about distances between values, but one outlier leaves most districts in one colour.
• Quantile uses every colour equally, which makes any map look varied. Two districts Rs 30 apart can land in different colours, and a reader will assume they differ.
• Natural breaks fits this particular dataset, so the classes change when the data change: two survey rounds classed this way cannot be compared colour for colour.
• For a series of maps (two rounds, or the same indicator across states), fix the breaks by hand with round numbers, and use the same breaks on every map.

Honest legends

A legend is part of the claim the map makes. In the Classes list of the Graduated renderer, double-click a label to edit it; the QGIS manual notes you can also set the legend format and its precision. An empty class, like the third equal-interval class in the last module, still shows in the legend: delete it or change the breaks.

• State the unit and the measure: "Mean monthly per capita expenditure (Rs), households surveyed". A bare "Expenditure" leaves the reader guessing.
• Round the class labels to what the data can support. "Rs 2,395.0 to 2,569.2" claims a precision that 24 households per district cannot give; "Rs 2,400 to 2,600" does not.
• Show no-data districts as no data, in grey, with their own legend entry. A blank district can read as zero. One way: put a second copy of the boundary layer underneath, styled with a single grey fill, and name it "No survey data".
• Use one hue, light to dark, for a quantity. Rainbow ramps have no natural order, and red-green ramps fail for many colour-blind readers.
• Say how many observations stand behind each value, in a note or a label. A district mean from 24 households is a rough estimate; map the sample size too if it varies.
• For survey estimates, use the weights. NFHS household weights are hv005 / 1,000,000 and women's weights v005 / 1,000,000; compute the weighted district value before you join it, in your statistics software. QGIS styles the numbers it is given and applies no survey weights.
• Write the source and the licence on the map: the survey, its year, and the boundary file's attribution line.

Publishing a map of India

Maps of India that show the national or state boundaries carry an extra requirement. The Department of Science and Technology's Guidelines for acquiring and producing Geospatial Data and Geospatial Data Services including Maps (F.No.SM/25/02/2020 (Part-I), dated 15 February 2021) say in paragraph xiii:

SoI is the Survey of India. Its website lists a Political Map of India, an Outline Map of India and state maps among its products. The same guidelines' paragraph xv says violations "will be dealt with under the applicable laws".

What that means for your map

• If your map shows India's outline or state boundaries, the outline must match the Survey of India's, including the boundaries in Jammu and Kashmir and Ladakh and in Arunachal Pradesh. Compare your boundary file with the Survey of India's political map before you publish.
• Global boundary sets do not necessarily follow it. GADM, Natural Earth and many web basemaps draw India's borders differently in places, so do not assume a downloaded file complies.
• DataMeet's repository has a Country folder whose README describes an India outline "in accordance with the Official boundary of India as per the Survey of India" (india-composite.geojson, licensed CC0) and a file india-soi.geojson, which it describes as "dissolved shapefiles from the official Indian shapefile data available" (licence given as CC-by-sa 2.5 / ODbL). Check either against the Survey of India map yourself.
• A map of a few districts with no national or state outline still benefits from a state boundary for context; take that boundary from a source that follows the official one.

Export a print layout

A print layout is a page on which you arrange the map, legend, scale bar, title and notes, and from which you export a PDF or an image. The steps follow the QGIS 3.44 manual's print layout overview and output pages.

1. Style the map in the main window first. Then choose Project ► New Print Layout… and give the layout a name.
2. Click Add map on the toolbar and drag a rectangle on the page. The current map view is drawn inside it.
3. Click Add legend and drag a rectangle. In the legend's item properties, remove layers that should not appear (the CSV table, for instance) and rename entries.
4. Click Add scale bar and click on the page. Add a north arrow from the same toolbar if your audience expects one.
5. Add labels for the title, the source line (survey, year, sample size) and the boundary attribution, for example DataMeet's suggested line with its licence.
6. Export with Layout ► Export as PDF… or Layout ► Export as Image…. The manual notes that a PDF export puts every page of the layout into one file, and that image export writes one file per page.
7. Save the project (Ctrl + S): the layout is saved inside it.

Where next