Home / GIS and Open Map Data for Beginners

New Layer From Selection in QGIS: Turn Chosen Features Into Their Own Dataset

Drone shot of lush green fields intersected by waterways

Photo: AJ Ahamad / Pexels

Selection is the verb at the heart of GIS: these parcels, those readings, the features inside this boundary. But a selection has the lifespan of a highlight. It lives in the project's memory, it survives exactly as long as you do not clear it, close the file, or select something else, and it cannot be emailed, styled independently, or handed to a colleague. The move that turns an answer into an asset is promoting the selection into a layer of its own, and it is a move beginners postpone for embarrassingly long because nobody shows them the door. This guide shows both doors, after a short review of making the selection itself. It completes a trilogy in this batch: the attribute table teaches asking questions of the data, select by location teaches asking questions of geography, and this article teaches keeping the answers. Its sibling on the export side, exporting a shapefile, covers sending a selection away as a file; here the goal is a layer that stays and works.

Interactive tool

A Selection Is Not a Layer Yet: The Checklist

Check off each step as you finish it. Your progress stays on this page.

0 of 7 complete

0% done

Make the Selection First: Four Doors Into the Same Room

Every route to a new layer begins with a selection worth keeping, and QGIS offers four ways to make one, all producing the same highlighted set. By hand on the map, clicking and dragging, for the handful of features your eye chooses. By table, selecting rows in the attribute table, where the records are readable. By expression, a written rule, the district combined with the size combined with the status, executed by select by expression. By location, the spatial questions, within, intersecting, containing, that our select by location guide teaches in full, whose own article is careful to note that a selection creates nothing, which is precisely the gap this article fills. Whichever door you used, pause before promoting and audit the catch: read the selected count in the layer panel or the table header, zoom to the selection, and eyeball it against the question you asked. A promoted selection freezes its mistakes into a dataset with a name and a future, so the thirty seconds of audit happen now, while unselecting is free. When the set is right, promote it by either of the next two routes.

Route One: Save Selected Features As

The direct route runs through the export machinery. Right-click the source layer, choose Export, then Save Selected Features As, and the familiar dialog opens with the decisive checkbox already implied by the menu item's name: the export will contain only the selected features. Choose the format with the layer's future in mind. For a layer that stays in your QGIS life, GeoPackage is the right home: one file, full field names, no legacy limits. Give it a name that states what the selection was, north_district_large_parcels rather than layer_copy_final2, because in three months the name is the only documentation of the question that made it. Set the CRS, normally the source layer's own, and confirm. The new layer appears in your project as a first-class citizen: its own style, its own table, its own future selections. Note what it is and is not. It is a copy: from this moment, the new layer and the source evolve independently, and editing one never touches the other. That independence is the point, a frozen answer, and also the responsibility: if the source data is corrected tomorrow, your promoted layer keeps yesterday's truth, and its name and your notes are how the future learns which truth it holds.

Drone shot of lush green fields intersected by waterways
Drone shot of lush green fields intersected by waterways

Route Two: Extract Selected Features in Processing

The second route lives in the Processing toolbox, and suits users who build workflows rather than click dialogs. Open Processing, search for Extract selected features, and run it on your layer: the algorithm writes the selected features to a new layer, in memory or to a file you specify, as one step that can sit inside a graphical model or a script alongside buffers, clips, and joins. The result is the same promoted layer as route one; the difference is the company it keeps. In a model, the extraction can feed directly into the next operation, so the selection becomes the input of an analysis chain rather than a manual artifact, reproducible the next time the source data updates. That reproducibility answers the copy problem of route one in the way processing always does: do not maintain the frozen layer, maintain the recipe that rebuilds it. Choose route one for the one-off answer you need to keep and send; choose route two when the question will be asked again, on fresher data, by you or by whoever inherits the project. Both are respectable. The only disreputable option is leaving a crucial selection as a highlight and hoping the project never closes.

The Scratch Layer Trap: Temporary Is a Warning

There is a third way a selection becomes a layer, and it bites everyone once. Copy features and paste them as a new layer, or let various tools create outputs without specifying a file, and QGIS makes a scratch layer, held in memory, marked with its own icon, and deleted when the project closes. Scratch layers are wonderful for experiments and fatal for results, because the program's warning, the dialog that lists temporary layers when you close and asks whether to make them permanent, arrives at the exact moment of least attention, at the end of the day, one misclick from oblivion. Build the reflex: any layer you would mourn gets a file. Right-click a scratch layer and choose Make Permanent, choose its GeoPackage home and name, and it crosses from memory into the world of things that exist tomorrow. The same reflex applies to processing outputs left in temporary form. GIS projects accumulate layers the way desks accumulate paper, and the discipline of permanence, real file, real name, at the moment of creation, is what separates a project that can be reopened with confidence from one that opens onto a small archaeology of loss.

Treat the closing dialog, when it comes, as an audit rather than an interruption. It lists exactly the layers that exist only in memory, and reading it slowly, once, teaches you which of your habits produces them: the quick paste, the processing run left on temporary, the experimental extraction you meant to evaluate and discard. Some deserve the discard; the dialog is not a guilt trip. But any layer on that list whose loss would cost you an hour belongs in a file, and the minute spent making it permanent is the cheapest insurance in the whole program.

After the Promotion: Style It, Document It, Use It

A promoted layer earns its keep in the work that follows. Style it distinctly from its parent, because two identically drawn layers stacked in one project invite the oldest confusion in GIS, editing or measuring the wrong one. Record its provenance where the future will find it: the layer name carries the question, and a line in the project notes or the layer's metadata carries the date, the source, and the rule, the expression or the spatial test that defined it. Then use it as the independent thing it now is: the mailing list of chosen parcels, the study set for the fieldwork, the input to the buffer, the exhibit in the report. When the day comes to send it away, it exports like any layer, through Save Features As to shapefile if the receiver demands that format, with its provenance note riding in your message. And when the source data improves and the layer's frozen truth goes stale, the honest options are the two this article has given you: rebuild it by hand through route one, or rebuild it by recipe through route two. What you may not do is quietly edit the copy to resemble the new truth, because a layer whose history nobody can reconstruct is worse than a stale one whose history is written on its face.

Sharing the layer deserves the same care as making it. A GeoPackage layer travels as a single file: send the file, and the receiver opens the exact dataset, fields, names, and all, with none of the shapefile family's separation risks that our export guide guards against. Send its provenance note with it, the question, the rule, the date, the source, because a layer arriving without its story invites the receiver to invent one, and invented stories about data have a way of becoming project facts. The promoted layer, documented and whole, is the unit in which GIS work is actually handed over: not a screenshot, not a highlight, not a memory of a selection, but a dataset that says what it is.

Three Promotion Patterns Worth Naming

The promotion move recurs in shapes regular enough to name, and naming them makes the next project faster. The study area set: a spatial selection, everything within the boundary, inside the buffer, along the corridor, promoted into the layer the analysis then owns, the parcels to survey, the wells to sample, the segments to inspect. Its provenance note records the boundary and the date, because study areas are redefined and the layer must say which definition made it. The mailing list set: an attribute selection, the owners in this district, the customers in that tier, the properties meeting a rule, promoted so it can be joined, exported, and handed to the people who will act on it, with the expression that defined it written into the notes verbatim, since a list whose rule is lost cannot be defended or refreshed. The evidence set: the selection that supports a claim in a report, the features inside the flood outline the year of the event, the readings above the threshold during the incident, promoted and frozen precisely because its frozenness is the point, the exhibit that must not drift as the source data is corrected. Three patterns, three different reasons to freeze, one shared discipline: the name states the question, the note states the rule, and the layer, once promoted, is treated as a document, cited rather than casually edited. Recognize which pattern you are in and the decisions this article has scattered, format, naming, route, permanence, mostly make themselves.

Related Guides

This guide does not stand alone. The GIS and open map data guides below take up the questions it leads into:

Frequently Asked Questions

What is the difference between a selection and a layer?

A selection is a temporary highlight held in the project's memory; it vanishes when cleared or when the project closes, and it cannot be styled, saved, or shared on its own. A layer is a dataset. Promoting a selection through Save Selected Features As or Extract Selected Features copies the chosen features into a real layer with its own file and future.

Does the new layer update when the source layer changes?

No. Both promotion routes make a copy, and the copy is frozen at the moment of creation. If the source is corrected later, rebuild the layer, by hand for a one-off or through a processing model when the question will be asked repeatedly. The layer's name and notes should record when and how it was made.

What is a scratch layer, and why did my layer disappear?

A scratch layer is a temporary, memory-only layer, created by pasting features as a new layer or by tools whose output was not saved to a file. Scratch layers are deleted when the project closes. Use Make Permanent on any scratch layer you need to keep, and give it a GeoPackage file and a real name.

Which format should the new layer be?

GeoPackage for anything staying in your GIS work: one file, full-length field names, no size ceiling worries. Use shapefile only when a receiver specifically needs that format, and then export a copy for them rather than demoting your working layer.

Can I create a layer from a selection made by location?

Yes, and it is one of the most common workflows in GIS. Select by location answers the spatial question, and either promotion route then turns the result into a dataset. Audit the selection's count and geography first, because promotion preserves mistakes as faithfully as it preserves answers.

Photo credits: AJ Ahamad (Pexels); Sasa Peterkovic (Pexels). Photos are served from this site; follow the credit to the library page for the original.