RETROCHANGE DATASET METADATA - UC3 / RETROFRAG
============================================================

Dataset title:
RETROCHANGE UC3 RETROFRAG Forest Fragmentation Example Maps

Purpose:
Forest fragmentation, forest edges, and core-forest continuity
analysis using finished forest / non-forest maps.

Study areas:
Ruské and Havešová, Poloniny National Park, Slovakia;
Brzegi, Bieszczady region, Poland.

Map reference years/labels: 1975 and 2017
Modern source product period: 2017-2019
Data provider: Technical University in Zvolen, Slovakia
Project: RETROCHANGE
Documentation edition: November 2026
Dataset version: No separate dataset version is recorded in the files.
The accompanying RETROFRAG workflow is version 1.0 (October 2026).


DATA INVENTORY AND SPATIAL PROPERTIES
------------------------------------------------------------
All four files are single-band GeoTIFF classification maps:
- CRS: EPSG:5514 - S-JTSK / Krovak East North
- Coordinate units: meters
- Pixel size: 0.5 x 0.5 m; north-up, no rotation
- Pixel interpretation: Pixel is area
- Stored sample type: Unsigned 8-bit integer (UInt8)
- Compression: LZW
- Declared NoData value: 255

File                    Columns x rows  Footprint       Bounding area
Brzegi_1975.tif          10000 x 10000   5,000 x 5,000 m  2,500 ha
Havesova3x3_1975.tif       6000 x 6000   3,000 x 3,000 m    900 ha
Ruske_1975.tif           10000 x 10000   5,000 x 5,000 m  2,500 ha
Ruske_2017.tif           10000 x 10000   5,000 x 5,000 m  2,500 ha

Outer pixel-edge bounds, in EPSG:5514 meters:
File                   xmin       ymin       xmax       ymax
Brzegi_1975.tif       -167600   -1199000    -162600   -1194000
Havesova3x3_1975.tif  -184250   -1212375    -181250   -1209375
Ruske_1975.tif        -184250   -1200600    -179250   -1195600
Ruske_2017.tif        -184250   -1200600    -179250   -1195600

Bounding areas describe rectangular raster extents. Valid coverage,
common coverage between dates, forest area, and core-forest area use
different denominators and must be reported separately.

The two Ruské rasters have matching dimensions, origin, pixel size,
and CRS in their headers. No claim of independent positional accuracy
is implied by grid alignment or by the 0.5 m sampling interval.


CLASS DEFINITIONS AND SHADOW HANDLING
------------------------------------------------------------
Brzegi_1975.tif, Ruske_1975.tif, Ruske_2017.tif:
  0   non-forest
  1   forest
  2   shadow_excluded (as labeled in the embedded CLASSES metadata)
  255 no data (declared raster NoData)

Havesova3x3_1975.tif:
  0   non-forest, including deep shadow
  1   forest
  255 no data

Class definitions above come from embedded metadata, not from a new
pixel-value census or independent validation.

The analysis notebook defaults to FOREST_VALUE = 1 and NODATA_VALUE = 255.
Its reader also honors the raster's validity mask. It does not explicitly
exclude class 2: where such pixels are valid in the file mask, they are
treated as non-forest under the default settings.

To honor the shadow_excluded label, set NODATA_VALUE = 2 (No-data value
in the interface). The file-declared 255 remains excluded as well.
Alternatively, recode 2 to 255 in a separate analysis copy. Record this
choice because it changes valid coverage and edge-related indicators.
Havešová's deep shadow is already included in class 0; it cannot be
separated from non-forest using this class map alone.


SOURCE AND PROCESSING PROVENANCE
------------------------------------------------------------
1975 maps: Derived from historical KH-9 HEXAGON imagery, as described
in the package README and user guide. Exact acquisition dates, mission
identifiers, and scene identifiers are not supplied for these maps.

Modern source for Ruske_2017.tif:
Ortofotomozaika 2017-2019 © GKÚ Bratislava, NLC.
The filename retains the map label 2017. The source mosaic covers
2017-2019; the exact acquisition date for this subset is not supplied.
A source tile identifier and source-specific license notice are not
included in this package.

Embedded METHOD metadata records:
- Brzegi 1975 and Ruské 1975: NB3d, k = 14, manually selected forest
  clusters, post-processing, road-edge veto, and cluster-specific halos.
- Ruské 2017: NB3d, k = 14, manually selected forest clusters and
  post-processing.
- Havešová 1975: K-means, k = 14, using image brightness and local
  standard deviation (11-pixel window), cluster selection, shadow
  reclamation, and morphological/minimum-area cleanup. The tag records
  forest/opening minimum mapping units of 500/200 m2.

The full METHOD strings remain embedded in the GeoTIFF files.
Producing these classifications is outside the RETROFRAG workflow.
Raster properties and class/method tags were inspected directly;
the classifications and fragmentation results were not recomputed
for this documentation update.


USE AND LIMITATIONS
------------------------------------------------------------
Use one map for a fragmentation baseline; use the two Ruské maps
for two_dates, or two different sites for two_areas. Two-date analysis
uses their common valid coverage. Match grain, cleaning rules,
connectivity, edge depth, and exclusion treatment when comparing results.

Input classification errors, shadow treatment, map boundaries, and
minimum-area cleanup affect fragmentation and core-forest estimates.
A core-forest indicator is defined by the chosen edge-depth rule;
it is not independent evidence of habitat condition.


LICENSE, CITATION, AND CONTACT
------------------------------------------------------------
CC BY 4.0, to the extent of rights held by the Technical University
in Zvolen; see LICENSE.txt for scope, source rights, and attribution.
https://creativecommons.org/licenses/by/4.0/

Recommended citation:
Technical University in Zvolen (2026).
RETROCHANGE UC3 RETROFRAG Forest Fragmentation Example Maps.
RETROCHANGE project. https://gis.tuzvo.sk/retrochange

Workflow author/contact (from the user guide):
Ladislav Šumichrast, Technical University in Zvolen
xsumichrastl@tuzvo.sk

RETROCHANGE - Retrospective Analysis of Forest and Landscape
Changes from Historical Satellite Imagery
https://gis.tuzvo.sk/retrochange
