[general]
name=SDM
qgisMinimumVersion=4.0
qgisMaximumVersion=4.99
description=A guided wizard for species distribution modeling, from aligning modeling variables to ensemble predictions and generating SDM reports.
version=1.2.1
author=John Paul Payopay
email=jpaul.payopay@gmail.com

about=SDM is a guided wizard for species distribution modeling in QGIS. It takes presence-only or presence/absence records plus a stack of predictor rasters and walks you through coordinate cleaning and thinning, background point generation, stepwise VIF predictor selection, spatial-block or k-fold cross-validation, and model fitting with nine algorithms (logistic regression, GAM, Random Forest, GBM, XGBoost, SVM, MLP, MaxEnt and ENFA), validating your inputs at every step. It reports AUC, TSS and Boyce across replicate runs, writes continuous and binary suitability rasters, response curves and variable importance, combines the models into a weighted or unweighted ensemble with an uncertainty map, optionally projects onto a second raster stack with MESS and MOP extrapolation flags, and bundles the results into a single HTML report, with every score, response curve and importance value also written out as CSV. Requires QGIS 4.0 or newer.

changelog=1.2.1
    - Every number in the report and the figures is now also written as CSV: model scores per algorithm, replicate and cross-validation fold (metrics_summary.csv, metrics_detail.csv), the response curve values (response_curves.csv, response_curves_summary.csv) and the variable importance values (variable_importance.csv, variable_importance_summary.csv).
    - Each cross-validation fold is now scored on its own held-out points, so a replicate's score can be checked against the folds it came from — most useful under spatial-block CV, where every fold is a different region.
    - The Plugins menu entry now shows the plugin icon.
    1.1.0
    - Predictor and projection raster pages now report each layer's data type, CRS, size, resolution, NoData and extent when the rasters don't line up, and a "Fix predictor layers" button resamples them onto one grid you choose (CRS, extent, resolution, plus per-layer output type, NoData and resampling method).
    - Background points: new "ratio to presences" method (count scales with the presence total), new "SRE" method (pseudo-absences drawn from outside the presences' environmental envelope), and the buffered method is now a disk taking a minimum and maximum distance from the nearest presence.
    - Selecting an option on the background, cross-validation, ensemble and algorithm pages now shows what it does, when it is the right choice, and the values people normally use.
    - Cross-validation methods reordered with k-fold first; spatial block remains the default.
    - Wording pass across every page.
    1.0.1
    - Initial public release.

tracker=https://github.com/jp-payopay/SDM/issues
repository=https://github.com/jp-payopay/SDM
category=Analysis
tags=species distribution modeling, sdm, habitat suitability, niche modeling, ensemble, machine learning, ecology, biodiversity, conservation
homepage=https://github.com/jp-payopay/SDM
icon=icon.png
experimental=False
deprecated=False
license=GPL-3.0-or-later
