PlanX DataCube Lab 1.2.0 — QGIS Python Plugins Repository

Changelog

## [1.2.0] - 2026-06-17

### Added
- **Space-Time Cube Aggregation** (`space_time_cube`) — bin
point/feature events by centroid into a fishnet grid crossed with time
bins (count / sum / mean / max / min). Emits a long-format panel
polygon layer (cell_id, col, row, time_step, time_value, value, count)
that feeds directly into EHSA, Local Moran's I, Trend, Anomaly and
Forecast. Optional netCDF4 cube export.
- **Local Outlier Analysis (Anselin Local Moran's I)**
(`local_morans_i`) — High-High / Low-Low clusters and High-Low /
Low-High spatial outliers with a deterministic conditional-permutation
pseudo p-value, on a temporally aggregated panel (mean / last / sum /
max).
- **Temporal Trend Analysis** (`temporal_trend`) — Mann-Kendall
significance + Theil-Sen slope and intercept per location, classified
as Increasing / Decreasing / No significant trend.
- **Temporal Anomaly Detection** (`temporal_anomaly`) — per-location
anomalous time steps via a robust median/MAD modified z-score (or
classic mean/std), with anomaly count, rate, max |z| and
most-recent-step flag.
- **Forecast Backtest** (`forecast_backtest`) — hold out the last N
steps, score Exponential Smoothing / ARIMA / Random Forest by MAE,
RMSE and MAPE, and report the lowest-RMSE method per location.
- Automatic categorized symbology for the EHSA, Local Moran's I,
Trend, Anomaly and Backtest outputs.

### Changed
- **Emerging Hot Spot Analysis Getis-Ord Gi\*** now computes a *true*
local statistic using spatial neighbourhoods — fixed distance band
(auto or explicit) or K-nearest neighbours — instead of a global
self-only z-score. The previous behaviour remains available as "No
spatial weighting".
- Shared `core/spatial_weights.py`, `core/panel.py` and
`core/timeseries_stats.py` modules unify neighbourhood construction
and panel parsing across all spatiotemporal tools.
- Dashboard catalogue, Quick Actions and Modules grid expanded to the
full 8-algorithm toolbox (new "Forecasting" group).

### Tested
- End-to-end headless runs of all 8 Processing algorithms on QGIS 3.44
(LTR) and QGIS 4, with numerical assertions (Gi\* hot-cluster
z-scores, Local Moran's HH/LL recovery, Mann-Kendall trend sign,
anomaly recovery, backtest accuracy ordering).

Approved

yes

Author

geo140195philo

Uploaded

2026-06-17T01:02:32.381445+00:00

Minimum QGIS version

3.34.0

Maximum QGIS version

4.99.0

External dependencies (PIP install string)

None