0.15.0 Barriers can be crossed where people can walk. A carved
motorway, railway or river corridor blocked movement along its whole
length, with no bridge, level crossing or station footbridge anywhere,
so it severed land a planner can plainly reach: at the Crews Hill
scenario 238 hectares of developable ground sat 75 metres from the
station and unreachable at any distance. The OSM download now writes a
walkable-ways layer (footpaths, and the streets a pedestrian may use,
excluding motorways and trunk roads), and the run dialog takes it as
an input you can edit or replace with your own. It is applied after
the unbuildable carve and after the slope limit, so a footbridge marks
the motorway beneath it crossable. Those cells stay closed to
development: a bridge does not make a railway a building site, only a
route across it. Scenario panels draw a crossing as a gap in the
barrier rather than as a fourth class of ground, and barriers are
drawn solid and dark, since a reader who cannot see the wall reads the
gap as available land.
0.14.0 A slope limit, barriers carved by cell centre, walks that
cannot cut a corner, and centres that are earned rather than drawn.
The run dialog gains a steep-slopes layer and a slope limit in
degrees: bands at or above the limit are unbuildable, which the
shipped scenarios described but the plugin never applied, so a
hillside scenario ran without its terrain constraint. Unbuildable land
and barrier corridors are now burned by cell centre like every other
layer instead of by every touched cell, which had carved up to a cell
beyond each edge and, at Cambourne, doubled the carved area; existing
built fabric is no longer carved away by a corridor that clips it. A
walk can no longer pass diagonally between two blocked cells, so a
carved corridor, a river or a steep band blocks it properly. Centres
are earned, not drawn, and a settlement may thicken into its own bay.
The minimum green span protected any short run of green between the
candidate cell and the nearest built land, in each direction
independently. Where growth formed a C, U or N, the arms of the same
settlement read as the far side of a green corridor, so the bay they
enclosed could never be filled: runs left narrow fingers with a
permanent green slot on either side, and windows reported capacity
findings that were artefacts of the rule. The span is now protected
only where it separates the candidate from a DIFFERENT settlement,
which is what a corridor between developments means; a run ending at
the window edge no longer counts either. The park area rule and the
green walk still govern what is built. Case studies gain what the rule
was withholding: Cambourne reaches 100 percent of target (was 23),
Crews Hill 102 (was 81), Pajarito 101 (was 81), and the transit
corridor run no longer costs population. Centres are earned, not
drawn. New centres previously appeared by a probability drawn per cell
per iteration, which tied the number of settlements to the size of the
window, the resolution of the grid and the pace of growth rather than
to the population: a Cambourne window drew twenty-four times as many
satellite centres as the published model's area-scaled rate, and
post-processing then pruned three quarters of the grown population as
unviable. Each new home is now charged to the centre it is nearest,
and a centre founds its successor once that population reaches the
service viability threshold, or once it has no room left to grow.
Growth therefore creates the centres post-processing is willing to
keep: at the Cambourne demonstration the plan now houses 100 percent
of the target where it held 23 percent, and pruning removes a tenth of
a percent of the grown population instead of three quarters. Three
settings go with the change. The build probability and the
neighbouring-centre probability are no longer exposed: the build draw
paces growth and varies the ensemble, and no longer shapes the plan.
Dispersed development becomes a choice of where an earned centre may
go, attached to existing development or detached from it, in place of
the off/moderate/aggressive probabilities. The service viability
threshold now governs settlement size directly, so raising it yields
fewer, larger settlements. Scenario files record allow_detached in
place of dispersal and build_prob. A centre drawn as an area counts as
one centre rather than one per covered cell, the nearest-centre
assignment breaks ties on position so a layer's feature order cannot
change a plan, and the build-out condition is tested directly rather
than inferred from a quiet iteration, which at a low build rate used
to found settlements that the population had not earned.
0.13.1 The centre-seeding cap is removed. New centres previously
stopped seeding once a run reached 80 percent of its population
target, a guard that predates the viability rules: post-processing
already removes a centre that ends up serving fewer than the service
viability threshold, and prunes the growth no remaining centre
reaches. Seeding now continues for the whole run and viability alone
decides which centres survive. Windows whose late growth was dispersed
keep less of the target than before, because centres that seed late
gather too few homes and are pruned with the growth around them;
concentrating growth (a transit corridor preference, lower dispersal,
or a wider centre walk) recovers it.
0.13.0 Transit-oriented growth, attached infill counted, doomed
pockets ruled out up front. Two transit inputs now shape a run.
Transit hubs (rail and tram stations, or any point the planner adds)
each anchor a pinned mixed-use centre that seeds growth and survives
post-processing, so a designated centre can be placed by hand. Transit
corridors (bus stops as points, or a proposed route drawn as a line)
attract growth: a corridor preference of 0 to 1 scales the development
draws outside the walkable transit catchments (400 m of a stop by
default, 1,200 m of a hub), so growth concentrates along existing or
proposed transit; at the default of 0 the run is unchanged and the
layers feed the run report only. The run log warns when the catchments
cannot hold the target population. The Cambourne scenario ships a
hand-drawn proposed corridor and bus rapid transit stop demonstrating
both levers. Absorption is removed: a cluster of new development grown
against existing fabric was anchored on the town's centres by the
growth rules, so it stays new development whatever its size
(previously it was folded into the existing fabric and dropped from
the counts, which discarded large amounts of legitimate infill). Grid
preparation now reclassifies unviable pockets as protected green: a
connected patch of buildable land that touches no existing fabric and
is smaller than the minimum settlement can never host viable
development, so it is ruled out before the run instead of being grown
and pruned. Delivered population and served coverage rise
substantially in windows with existing fabric.
0.12.22 Four outputs replace the two clustering options: the raw run
(coloured by the density tiers it actually drew, in place), centres as
grown, optimised placement, and fewest centres. Every processed option
keeps every home within the centre walk: a placement move that would
strand a home is discarded, a centre is added wherever new development
lacks provision within the walk, and a centre is removed only while
every home stays covered. A contiguous cluster of new development
below the minimum settlement size is absorbed into the existing fabric
when it touches it (stray free cells inside a town are usually
unmapped roads or parks, not building sites) and reverts to green when
detached. Loading parameters now lists exactly which fields changed
with old and new values, and names any keys in the file it could not
use. Downloaded OSM layers arrive pre-styled in a fixed colour scheme
instead of random symbology. Requires engine 0.12.22 (the raw layer's
drawn densities are recovered by re-running the selected ensemble
member at its own seed).
0.12.21 The run report is a set of fixed-width tables: the plan
options side by side (population accommodated, coverage, walks,
per-person provision), the achieved density mix per tier, and the
centre audit with the weakest new centres; single-run mode now writes
a report too. Rail and tram stations seed growth as centres by default
(development gathers around a station, and post-processing pins and
grows its centre); stations and stops that land on a carved corridor
cell snap to the nearest walkable cell instead of dropping out of the
anchor and access accounting. The dialog rejects a CRS whose area of
use misses the extents layer and names the local UTM zone to pick,
instead of failing mid-run with a raw transform error.
0.12.20 Metadata made parseable by the plugin repository (a stray
percent sign rejected the 0.12.19 upload); the release check now
parses it the same way the repository does. A run's output layers load
as one group named for the output folder and run, so successive runs
stay organised in the layers pane.
0.12.19 The run dialog, the message bar and the guide all point at the
Isobenefit tab of the Log Messages panel, where a running simulation
reports per-stage progress and warnings.
0.12.18 Post-processing candidates run about twice as fast again
(component labelling joins the walk field in the engine; the placement
stack caches its columns), and the progress bar now gives the
selection stage the larger share of the bar it actually occupies.
0.12.17 Post-processing is engine-accelerated and barrier-aware: the
walk field is computed by the Rust engine (a large window
post-processes in under a minute), and scoring walks now detour around
water and carved corridors exactly as growth does (they previously
crossed them). Requires engine 0.12.17; the plugin prompts for the
upgrade.
0.12.16 With an OSM download in the project, the run dialog pre-fills
the output folder as a scenarios folder beside the downloaded data,
numbered upward when earlier runs exist. A chosen folder is never
overwritten.
0.12.15 One distance model: bounded grid walks everywhere, in growth
and in scoring. Street-network routing removed (a new settlement has
no streets yet, so a network metric measured new and existing fabric
on different terms); the streets layer stays as map context and the
source of carved barrier corridors. Centre placement is
settlement-local: a centre is positioned only by homes in its own
contiguous settlement, so it cannot be dragged to a periphery cell
facing a green gap. Scenario panels colour slope-excluded land tan,
distinct from water and corridors. Plugin unload no longer errors
after a finished run.
0.12.14 Centre refinement caches walk fields by coordinate across the
whole ensemble: post-processing a many-centre window drops from ~40 s
to ~3 s per candidate with identical results. Selection progress logs
every candidate.
0.12.13 The post-processing selection reports progress, honours
Cancel, and runs in two phases when a street network is supplied
(every run ranked by the grid walk, the leaders re-scored along the
network), so large windows finish in minutes rather than sitting
silent near the end of the bar.
0.12.12 Defaults grounded in practice: 800 m centre walk (a ten-minute
walk) and 400 m green walk (the WHO/Natural England everyday-green
reach); the minimum settlement is now a population (default 1,000
people) rather than an area; new development must earn its own
mixed-use centres (a nearby existing centre no longer stands in, so
growth cannot sprawl centre-free along existing fabric); scenario
downloads fixed on the website; one consolidated Dnipro scenario;
distances, barriers and station handling documented in the guide.
0.12.11 Growth no longer stands off carved road, rail and river
corridors: a green strip bounded by unbuildable land is exempt from
the minimum-span rule, so development can build up to a barrier while
green corridors between developments stay protected. Seeding diagrams
and probability mechanics documented on the website; growth
troubleshooting added to the guide.
0.12.10 First public beta, for QGIS 4. Earlier development history is
recorded in the repository.

None