Forest balance
& canopy change.

A forest can keep its land-use label while losing its canopy. These maps measure different parts of the story.

ForTy maps less forest in both massifs. Copernicus shows a small crown-area decrease in Făgăraș and an increase in Retezat. The periods and measurements differ.

Făgăraș & Retezat · whole-massif studyPublished 28 October 2026Mapped exploratory estimates

Three measurements, two mountain massifs
Measurement & period	Făgăraș	Retezat
Forest area · net changeForTy v2 beta · 2017–2025	−1,018.44 ha	−97.62 ha
Canopy disturbance · lossHansen GFC · 2018–2025	2,545.69 ha	180.12 ha
Crown-covered area · net changeCopernicus density · 2018–2024	−75.10 ha	+294.97 ha

These totals follow published massif polygons, unlike the landscape explorer’s rectangular study areas. Different products’ losses and gains cannot be combined into a net figure.

What happened
to the canopy?

Canopy disturbance and forest land use answer different questions. Hansen detects loss events; Copernicus measures changes in tree-covered extent and crown density.

Hansen · disturbance on a 30 m grid

2018–2025

Canopy loss and overlap with ForTy endpoint loss, hectares
Massif	Hansen loss	Both products	ForTy loss only	Hansen loss only
Făgăraș	2,545.69	1,102.83	207.79	1,442.86
Retezat	180.12	85.34	46.24	94.78

Hansen loss may reflect harvest, mortality, fire, or other disturbance. The comparison uses ForTy forest in 2017 as its baseline, with fractional 10 m support on an aligned 30 m equal-area grid. No year-2000 canopy-density threshold was applied, so these totals differ from default Global Forest Watch statistics. Repeated losses at one pixel are not fully represented.

Overlap shows cross-product agreement; it does not independently validate either product, the cause of loss, or permanent deforestation.

Copernicus · the area covered by tree crowns

2018–2024

Făgăraș

Net −75.10 ha

Crown decrease2,974.29 ha

Crown increase2,899.19 ha

Retezat

Net +294.97 ha

Crown decrease483.49 ha

Crown increase778.46 ha

Density changes mostly cancel in Făgăraș; Retezat has a net increase. Values are equivalent crown-covered hectares, not hectares of newly forested land. Both massifs use the same bar scale.

Density-weighted crown area, hectares · native 10 m support
Massif	2018 crowns	2024 crowns	Decrease	Increase	Net change
Făgăraș	92,121.63	92,046.53	2,974.29	2,899.19	−75.10
Retezat	23,984.78	24,279.76	483.49	778.46	+294.97

Tree Cover Density measures vertical crown projection: density / 100 × 0.01 ha per 10 m cell. Positive and negative differences describe spatial change between 2018 and 2024, rather than cumulative annual events.

Where tree-covered extent changed

20 m support

Tree Cover Presence Change and matching-period ForTy net change, hectares
Massif	Period	Canopy lost	Canopy gained	Canopy net	ForTy forest net
Făgăraș	2018–2021	501.60	39.88	−461.72	−452.45
Făgăraș	2021–2024	239.44	77.12	−162.32	−259.41
Retezat	2018–2021	23.52	2.52	−21.00	−28.72
Retezat	2021–2024	9.00	181.84	+172.84	−33.51

Copernicus marks the presence-change product as not validated. Presence describes tree-covered extent; ForTy describes forest land use, which can persist through temporary canopy removal. The two three-year periods remain separate.

The forest-area
balance.

The same annual beta product measures both directions. Endpoint loss and gain compare 2017 with 2025 on one common valid-pixel mask.

Făgăraș

Net −1,018.44 ha

Lost1,310.62 ha

Gained292.18 ha

Retezat

Net −97.62 ha

Lost131.58 ha

Gained33.96 ha

Both massifs have less mapped forest at the endpoint. Net change equals gained area minus lost area. Both massifs use the same bar scale.

ForTy forest classes 1, 2, 3, 4 and 7 · 2017–2025 · hectares
Massif	2017 forest	2025 forest	Lost	Gained	Net	Net / baseline
Făgăraș	116,075.72	115,057.28	1,310.62	292.18	−1,018.44	-0.877%
Retezat	31,839.06	31,741.44	131.58	33.96	−97.62	-0.307%

These are changes in modelled forest classes. They are not validated forest-inventory estimates, canopy-density measurements, or proof of permanent deforestation. Tree crops and agroforestry remain outside the forest total.

Nine years.
More than endpoints.

A cell forested in both 2017 and 2025 can lose its forest label and recover between them. Annual history reveals changes that an endpoint balance leaves out.

All forestNatural forestChange from 2017 · ha

Făgăraș

Each panel uses its own vertical scale. Solid: all forest. Dashed: natural classes 1, 2 and 7.

View Făgăraș annual measurements

Change from 2017 and annual forest turnover, hectares
Year	All forest Δ	Natural Δ	Annual lost	Annual gained
2017	0.00	0.00	0.00	0.00
2018	−85.29	−81.22	85.74	0.45
2019	−133.19	−124.50	101.11	53.21
2020	−460.23	−437.17	421.49	94.45
2021	−537.74	−503.82	133.66	56.15
2022	−591.43	−550.37	138.64	84.95
2023	−676.34	−620.29	161.82	76.91
2024	−797.15	−728.97	215.15	94.34
2025	−1,018.44	−920.71	246.31	25.02

Retezat

Each panel uses its own vertical scale. Solid: all forest. Dashed: natural classes 1, 2 and 7.

View Retezat annual measurements

Change from 2017 and annual forest turnover, hectares
Year	All forest Δ	Natural Δ	Annual lost	Annual gained
2017	0.00	0.00	0.00	0.00
2018	−6.32	−5.36	6.87	0.55
2019	−21.19	−17.29	20.16	5.29
2020	−31.19	−26.92	17.46	7.46
2021	−35.04	−29.46	7.89	4.04
2022	−37.85	−30.69	9.03	6.22
2023	−46.86	−37.47	18.29	9.28
2024	−68.55	−48.01	33.19	11.50
2025	−97.62	−76.37	31.29	2.22

Natural-forest endpoints and intervening recovery · 2017–2025 · hectares
Massif	Natural loss	Natural gain	Natural → planted	Forest recovered by 2025
Făgăraș	1,196.40	275.69	0.01	185.88
Retezat	102.20	25.83	0.00	11.25

Recovery follows cells forested in 2017 that return to a modelled forest class. It does not establish mature canopy, ecological restoration, or biodiversity recovery. Natural-to-planted conversion counts toward natural-forest loss but not total-forest loss. Later forest gains appear in annual turnover.

The massif
sets the boundary.

These scientific previews use the analytical 10 m rasters, clipped to published Romanian relief polygons. Open a figure at full resolution or download its GeoTIFF to inspect small changes.

Făgăraș: mapped 2017–2025 forest loss, gain, retention and intervening recovery within relief polygon 552. — Făgăraș · relief polygon 552 · 1,558.15 km². Loss and gain describe endpoint change; recovery records an intervening model change. Generalized massif boundaries are distinct from park and cadastral boundaries.

Retezat: mapped 2017–2025 forest loss, gain, retention and intervening recovery within relief polygon 576. — Retezat · relief polygon 576 · 412.97 km². Loss and gain describe endpoint change; recovery records an intervening model change. Generalized massif boundaries are distinct from park and cadastral boundaries.

The screening rule
changes the result.

Each rule has its own eligible support. These variations describe methodological sensitivity, rather than statistical confidence intervals or a local accuracy estimate.

Raw endpoints use all common valid cells. Stable endpoints require matching binary forest status in 2017/2018 and 2024/2025. Maximum native-class score filters apply at both endpoints, or all four years for stable endpoints. Excluded cells contribute to neither loss nor gain; scores are uncalibrated model confidence, never canopy percentages.

Făgăraș

Forest-area sensitivity, hectares
Rule	Score ≥	Lost	Gained	Net	Excluded
Raw endpoints	0.0	1,310.62	292.18	−1,018.44	0.00
Raw endpoints	0.5	1,275.85	286.55	−989.30	854.40
Raw endpoints	0.7	534.08	183.16	−350.92	26,226.75
Raw endpoints	0.9	No eligible cells			155,814.93
Stable endpoints	0.0	1,023.81	288.97	−734.84	356.89
Stable endpoints	0.5	992.07	277.76	−714.31	1,239.11
Stable endpoints	0.7	354.77	129.50	−225.27	26,567.59
Stable endpoints	0.9	No eligible cells			155,814.93

Retezat

Forest-area sensitivity, hectares
Rule	Score ≥	Lost	Gained	Net	Excluded
Raw endpoints	0.0	131.58	33.96	−97.62	0.00
Raw endpoints	0.5	129.42	31.54	−97.88	214.06
Raw endpoints	0.7	69.36	14.46	−54.90	7,772.76
Raw endpoints	0.9	No eligible cells			41,297.73
Stable endpoints	0.0	96.27	32.29	−63.98	40.79
Stable endpoints	0.5	92.53	28.84	−63.69	261.86
Stable endpoints	0.7	37.41	8.02	−29.39	7,824.73
Stable endpoints	0.9	No eligible cells			41,297.73

Fixed support.
Traceable evidence.

Every total has a defined measurement, period and geographical support. Input hashes, acquisition records and accounting gates accompany the downloadable results.

Published relief polygons · EPSG:3035 equal-area analysis
Massif	Polygon	Analytical area (ha)	Common coverage	Missing data (ha)
Făgăraș	552	155,814.93	100.000%	0.00
Retezat	576	41,297.73	100.000%	0.00

Boundaries & common coverage

The 2008 Romanian relief dataset follows Posea and Badea’s geographical regionalization. Its WGS84 download omits a projection file; the publisher’s GCS_WGS84 designation is used explicitly. These are generalized massif polygons.

All nine years share one valid-pixel mask, with at least 99% coverage required. Each 10 m cell represents 0.01 ha. Pixel centres determine polygon membership, so raster and vector areas differ slightly at the edges.

Classes & comparability

All years use ForTy v2 beta. Forest includes classes 1, 2, 3, 4 and 7; natural forest includes 1, 2 and 7. Tree crops stay separate. The national v1 baseline is never spliced into this series.

Sorted first-valid tile ownership prevents double counting. Model classification changes and temporarily unstocked forest can resemble loss or recovery. No field verification, causal attribution, legal determination or local accuracy estimate is included.

Copernicus coverage & grid details

Official 2018 rasters use an older ETRS89 datum serialization. Exact LAEA/GRS80 projection parameters are verified before normalizing the CRS to EPSG:3035.

Crown density uses 10 m support; presence change uses 20 m support
Massif	Product / period	Valid area (ha)	Coverage / excluded
Făgăraș	Density · 2018–2024	155,814.93	100.000%
Făgăraș	Presence · 2018–2021	155,816.76	0.00 ha excluded
Făgăraș	Presence · 2021–2024	155,816.76	0.00 ha excluded
Retezat	Density · 2018–2024	41,297.73	100.000%
Retezat	Presence · 2018–2021	41,296.68	0.00 ha excluded
Retezat	Presence · 2021–2024	41,296.68	0.00 ha excluded

Reproduce the research & read the rasters

From the project root: uv run forest-lab balance. With all inputs cached: FOREST_LAB_OFFLINE=1 uv run --offline forest-lab balance. Focused checks: uv run --offline pytest -q tests/test_balance.py.

Change-raster codes: 0 outside or excluded; 1 nonforest retained; 2 forest retained; 3 endpoint loss; 4 endpoint gain; 5 forest lost and recovered by 2025. Annual raster bands preserve native classes 1–7.

Sources & attribution

Romanian relief boundaries Bogdan Candrea, Petronela Candrea, Mihai Daniel Niță; regionalization after Posea and Badea (1984).
ForTy release & annual beta series Neumann and colleagues · Google DeepMind, Google Research, WRI and IIASA · 2026.
Official ForTy beta explorer Native band definitions and experimental annual maps.
Hansen Global Forest Change 2025 v1.13 University of Maryland, Google, USGS and NASA.
Copernicus Tree Cover Density European Union · EEA / JRC · 2018 and 2024.
Copernicus Tree Cover Presence Change 2018–2021 and 2021–2024 · product marked not validated.
WRI interpretation & comparability limits Read the definitions before comparing products.

Take the evidence
with you.

Original research artifacts, served locally. Tables retain their full numerical precision; GeoTIFFs retain the analytical resolution and coordinate system.

Forest balance
& canopy change.

What happened
to the canopy?

Hansen · disturbance on a 30 m grid

Copernicus · the area covered by tree crowns

Făgăraș

Retezat

Where tree-covered extent changed

The forest-area
balance.

Făgăraș

Retezat

Nine years.
More than endpoints.

Făgăraș

Retezat

The massif
sets the boundary.

Făgăraș

Retezat

The screening rule
changes the result.

Făgăraș

Retezat

Fixed support.
Traceable evidence.

Boundaries & common coverage

Classes & comparability

Sources & attribution

Take the evidence
with you.

Tables

Geospatial data

Sources & verification

Figures & report

Forest balance& canopy change.

What happenedto the canopy?

Hansen · disturbance on a 30 m grid

Copernicus · the area covered by tree crowns

Făgăraș

Retezat

Where tree-covered extent changed

The forest-areabalance.

Făgăraș

Retezat

Nine years.More than endpoints.

Făgăraș

Retezat

The massifsets the boundary.

Făgăraș

Retezat

The screening rulechanges the result.

Făgăraș

Retezat

Fixed support.Traceable evidence.

Boundaries & common coverage

Classes & comparability

Sources & attribution

Take the evidencewith you.

Tables

Geospatial data

Sources & verification

Figures & report

Forest balance
& canopy change.

What happened
to the canopy?

The forest-area
balance.

Nine years.
More than endpoints.

The massif
sets the boundary.

The screening rule
changes the result.

Fixed support.
Traceable evidence.

Take the evidence
with you.