Forest insect outbreak and bark beetle stress mapping

Bark beetle and defoliator outbreaks leave spectral fingerprints in red-edge and SWIR bands before visible browning begins. Sentinel-2 and Landsat time series can track infestation from green attack through grey snag, with an honest detection lag of four to eight weeks post-attack.

Sensors

Sentinel-2 MSI: 10 m visible and NIR bands; 20 m red-edge bands B5 (705 nm), B6 (740 nm), B7 (783 nm) and SWIR bands B11/B12. Five-day revisit at mid-latitudes with two satellites. Red-edge bands are the primary early-stress signal; SWIR responds to canopy moisture loss. Free archive from 2015.
Landsat 8/9 OLI: 30 m multispectral, 16-day single-satellite revisit (8-day combined). No dedicated red-edge band, but SWIR1 (1.57 µm) and SWIR2 (2.11 µm) track moisture stress and grey-attack stages reliably. Archive continuity back to 1972 via earlier Landsat missions enables long-term outbreak cycle analysis.
Planet SuperDove: 3 m resolution, daily revisit globally. Eight bands including a red-edge channel at ~717 nm. Useful for mapping individual crown-level mortality and validating Sentinel-2 classifications, though the red-edge placement differs from Sentinel-2 B5–B7 and spectral response functions must be harmonised before comparison.
MODIS MOD09: 250–500 m surface reflectance, daily global coverage. Too coarse for stand-level mapping but valuable for regional outbreak surveillance across millions of hectares and for detecting anomalous NDVI decline over entire mountain ranges in near real time.

What the canopy is actually signalling

A bark beetle attack kills the phloem. The tree's vascular system shuts down before the needles visibly discolour. During this green-attack phase, chlorophyll concentration begins to fall and canopy water content drops, but to the naked eye the crown still looks healthy. This is the window that remote sensing tries to exploit.

Red-edge reflectance, the steep slope between roughly 700 nm and 740 nm, is governed almost entirely by chlorophyll absorption. As chlorophyll degrades, reflectance rises in this range. Sentinel-2 bands B5, B6 and B7 sit precisely across this slope, making them more sensitive to early chlorophyll loss than broad red or NIR bands. SWIR bands respond to liquid water in needles; as the tree desiccates, SWIR reflectance rises independently of chlorophyll status, providing a second, partially uncorrelated stress signal.

The four to eight week problem

The detection lag is not a sensor limitation. It is a biological one. Chlorophyll degradation measurable from orbit typically requires four to eight weeks post-attack under field conditions, based on published studies of Ips typographus in European spruce and Dendroctonus ponderosae in North American pine. Before that window closes, spectral change is within normal seasonal variation and cloud-contaminated composites can mask it further.

Cloud cover compounds the problem at high latitudes and in maritime climates, where ten-day cloud-free Sentinel-2 composites may be unachievable for months at a time. Practical early-warning systems therefore combine dense time-series compositing, using all available clear pixels across a rolling window, with statistical change detection against a phenologically corrected baseline. Even then, the system is detecting early stress rather than confirming infestation; ground verification remains necessary before operational response.

The grey-attack and snag stages that follow are spectrally unambiguous and detectable at Sentinel-2's 20 m resolution. By that point, the beetles have typically moved on.

Spectral indices and what they measure

Several published indices are routinely applied to bark beetle mapping. The Red Edge Chlorophyll Index (CIre), calculated as (B7/B5) minus 1 in Sentinel-2 notation, correlates with chlorophyll content and declines measurably during green attack in published European studies. The Normalised Difference Red Edge index (NDRE, using B6 and B8A) offers a similar signal with slightly different saturation behaviour in dense canopies.

For moisture stress, the Normalised Difference Water Index (NDWI, using NIR and SWIR) and the Moisture Stress Index (MSI, SWIR1/NIR) both respond to canopy water loss. Combining a red-edge index with a moisture index reduces false positives from drought stress, which produces a partially overlapping spectral signature. Full spectral unmixing or machine-learning classifiers trained on labelled outbreak polygons can separate the two, but require training data from the specific forest type and region.

Mapping outbreak progression across time

A single image date tells you where dead trees are. A time series tells you how fast the outbreak is moving and in which direction. Annual Landsat composites, available free through USGS, can reconstruct outbreak histories going back decades, which matters because Ips typographus and Dendroctonus species show multi-year irruption cycles tied to drought, windthrow and stand age.

Sentinel-2's five-day revisit allows within-season tracking of the green-to-red-to-grey transition at individual stand level. Change vector analysis applied to sequential composites quantifies both the magnitude and direction of spectral change, distinguishing fresh mortality from recovering stands or from phenological artefacts. At Planet SuperDove's 3 m resolution, individual crown mortality becomes visible, which is useful for validating coarser classifications but generates data volumes that require careful management.

MODIS MOD09 sits at the other end of the scale. A single MODIS pixel at 250 m may contain hundreds of trees, so it cannot map outbreak boundaries precisely. Its value is regional: flagging anomalous NDVI decline across a national forest estate in near real time, which then triggers targeted Sentinel-2 tasking over the affected area.

Honest limits of the method

Mixed pixels are a persistent problem at 20 m and 30 m resolution. A Sentinel-2 pixel over a partially attacked stand will average healthy and stressed crowns, depressing the spectral signal below detection thresholds until mortality exceeds roughly 20–30 percent of canopy cover within the pixel. This means the method systematically underestimates outbreak extent in the early stages and in forests with heterogeneous structure.

Drought stress, nutrient deficiency and fungal infection produce overlapping spectral signatures. Without ancillary data (soil maps, meteorological records, species maps), spectral classification alone cannot reliably attribute canopy stress to bark beetles specifically. Hyperspectral sensors can narrow the ambiguity, but spaceborne hyperspectral missions with the necessary spatial resolution and revisit are not yet routinely available at the scale needed for national monitoring. Shadow effects in steep terrain further complicate radiometric analysis, and topographic correction is mandatory in mountain forests where outbreaks are often most severe.

Satellize runs Sentinel-2 and Landsat time-series analytics for forestry clients, applying phenologically corrected change detection and red-edge index stacking; the same analytical framework that underpins the Kingdom of Tonga crop-estimation programme transfers directly to canopy stress monitoring at national scale.

From spectral map to operational decision

The output of a bark beetle mapping programme is only useful if it reaches forest managers in time to act. Salvage logging of beetle-killed timber is economically viable only within a narrow window before wood quality degrades; in European spruce, that window is roughly one to two seasons post-mortality. A monitoring system that produces annual maps is too slow. Monthly or sub-monthly alert layers, delivered as GIS-ready polygons with confidence scores, allow managers to prioritise field inspection and harvesting crews.

Alert thresholds need calibration against local forest type, stand age and historical outbreak data. A threshold tuned for Norwegian spruce will produce different false-positive rates in Scots pine or Douglas fir. Building a calibrated, locally validated system takes one to two full outbreak seasons of paired satellite and ground data. That investment is substantial, but it is modest against the economic losses that large-scale undetected outbreaks cause.

Typical figures

Best spatial resolution (operational)	10–20 m (Sentinel-2); 30 m (Landsat 8/9)
High-resolution validation	3 m (Planet SuperDove)
Revisit frequency	5 days (Sentinel-2, mid-latitudes); 8 days (Landsat 8+9 combined); daily (MODIS, Planet)
Key spectral bands	Red-edge 705–783 nm (Sentinel-2 B5/B6/B7); SWIR 1.57 µm and 2.11 µm; NIR 842 nm
Minimum detectable mortality fraction	Approximately 20–30% canopy loss within a 20 m pixel (Sentinel-2); lower with Planet 3 m data
Detection lag post-attack	4–8 weeks (green-attack phase); grey-attack detectable within one to two Sentinel-2 revisit cycles
Cloud sensitivity	High; cloud-free compositing required; 10-day windows often insufficient at high latitudes
Archive depth	Sentinel-2 from 2015; Landsat from 1972 (earlier sensors); MODIS from 2000
Typical area coverage per run	Sentinel-2 tile: 100 × 100 km; national mosaics feasible with automated processing
Delivery formats	GeoTIFF stress index rasters; vector mortality polygons (GeoPackage/Shapefile); alert feeds (GeoJSON)

Analytics Satellize can run

Green-attack early stress layer	Red-edge chlorophyll index (CIre, NDRE) change detection against phenologically corrected Sentinel-2 baseline	Monthly GeoTIFF raster with per-pixel stress score; flagged polygons exceeding threshold, with confidence class
Outbreak progression map	Multi-date change vector analysis on Sentinel-2 or Landsat time series; green-to-grey spectral trajectory classification	Seasonal GIS polygon layer showing attack stage (green, red, grey) with area statistics by forest compartment
Regional anomaly alert	MODIS MOD09 NDVI anomaly detection against 20-year climatology; automatic flagging of declining pixels above area threshold	Near-real-time alert feed (GeoJSON) delivered to client GIS or dashboard, updated weekly
Mortality area estimate by stand	Supervised classification (random forest or support vector machine) trained on labelled Sentinel-2 imagery; SWIR and red-edge feature stack	Tabular report of estimated dead-tree area per management unit, with uncertainty bounds
Multi-year outbreak history reconstruction	Annual Landsat surface-reflectance composites; LandTrendr or equivalent temporal segmentation algorithm applied to SWIR and NDVI time series	Decadal disturbance stack (GeoTIFF) showing year of onset, duration and severity for each outbreak patch
Drought-versus-beetle attribution layer	Combined red-edge chlorophyll index and canopy water index (NDWI/MSI) stack; decision-tree separation of moisture-only versus chlorophyll-plus-moisture stress signatures	Classified raster distinguishing probable beetle stress, drought stress and mixed/uncertain pixels, with recommended ground-check locations

Sources

Who does the work

We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.