Tropical forest moisture stress and drought vulnerability mapping

Shortwave-infrared reflectance ratios reveal canopy water loss well before visible browning. Sentinel-2, Landsat and ECOSTRESS together give forest managers an early-warning signal for fire risk, carbon loss and drought-induced die-off.

Sensors

Sentinel-2 MSI: 10 m visible and NIR bands, 20 m SWIR bands (B11 at 1610 nm, B12 at 2190 nm). Five-day revisit at the equator with both satellites. The 20 m SWIR bands are the primary source for MSI and LSWI calculation. Free and open archive from 2015.
Landsat 8/9 OLI: 30 m multispectral including Band 5 (NIR, 865 nm) and Band 6 (SWIR-1, 1610 nm) and Band 7 (SWIR-2, 2200 nm). Eight-day combined revisit. Longer archive (Landsat 8 from 2013, Landsat 9 from 2021) supports multi-year drought-cycle comparisons.
ECOSTRESS (ISS-mounted): Thermal infrared radiometer, approximately 70 m resolution, measuring land surface temperature and evapotranspiration stress. Non-sun-synchronous orbit from the ISS gives variable overpass times, which is useful for capturing diurnal temperature range. Covers tropical latitudes well. Published ET stress products available via NASA Earthdata.
MODIS MOD13 / MYD13: 250 m to 500 m vegetation and water indices at 16-day composites. Lower spatial detail, but daily acquisitions allow near-real-time drought tracking at regional scale and long time series back to 2000. Useful for contextualising basin-wide drought episodes.

What a dry canopy gives away before the eye sees it

Water in leaf tissue absorbs strongly at 1610 nm and 2190 nm. When a canopy dries, liquid water content drops and those wavelengths reflect more energy back to the sensor. The shift is measurable weeks before visible wavelengths show any browning, because chlorophyll degradation lags behind moisture loss. This is the physical basis for using shortwave-infrared bands as an early-warning tool.

The Moisture Stress Index (MSI) is simply the ratio of SWIR reflectance to NIR reflectance. Higher values mean drier canopy. The Land Surface Water Index (LSWI) inverts that relationship: it subtracts SWIR from NIR and normalises, so lower values indicate stress. Both are derivable from Sentinel-2 Band 8A and Band 11, or from Landsat OLI Bands 5 and 6. Neither index requires anything beyond standard surface-reflectance products.

The cloud problem, and why it is not fully solved

Tropical forests sit under persistent cloud. In the humid tropics, Sentinel-2 cloud-free fractions can fall below 20 percent of acquisitions during the wet season, and even in the dry season a single scene may be partially obscured. This is the central operational limit of any optical moisture-stress workflow.

Several strategies reduce but do not eliminate the problem. Multi-temporal compositing, taking the least-cloudy pixel from a rolling 30-day window, smooths the signal but introduces a temporal lag. Sentinel-2 cloud masks (Sen2Cor or s2cloudless) miss thin cirrus over bright stressed canopy, which can be misread as low LSWI. MODIS provides daily coverage at coarser resolution and is genuinely useful for tracking drought fronts at the basin scale, even though 500 m pixels cannot resolve individual forest patches. SAR backscatter from Sentinel-1 is sensitive to canopy water content under certain conditions, but interpreting that signal for moisture stress alone is ambiguous without optical ground truth. Honest practice means reporting cloud fraction alongside any stress map.

Adding temperature: what ECOSTRESS contributes

Canopy temperature rises when stomata close to conserve water. A tree under severe moisture stress transpires less, so its surface temperature decouples from the surrounding humid canopy and climbs. ECOSTRESS measures this at roughly 70 m resolution, which is coarse enough to miss individual trees but fine enough to identify stressed forest patches of a few hectares.

The combination of an elevated MSI from Sentinel-2 and an elevated land surface temperature from ECOSTRESS is a stronger stress signal than either alone. The two sensors are not co-registered by default and their overpass times differ, so fusion requires careful temporal matching. ECOSTRESS also has gaps in coverage because the ISS orbit is not sun-synchronous and does not achieve systematic global revisit. Despite those caveats, the published ET stress product from ECOSTRESS has been used in peer-reviewed studies to detect forest water deficit in Amazonia and Southeast Asia, giving the method a documented track record.

From stress index to actionable risk layer

A moisture-stress map on its own is descriptive. The operational question is where stress is anomalous relative to the same location in previous years at the same point in the seasonal cycle. Anomaly detection requires a multi-year baseline, which is straightforward with Landsat (archive to 1972 for older sensors, to 2013 for OLI) and Sentinel-2 (2015 onwards). A pixel showing an MSI two standard deviations above its 10-year dry-season mean is a materially different signal from one that is merely in its normal seasonal trough.

Drought vulnerability maps add a second layer: which forests are structurally predisposed to stress? Factors include soil water-holding capacity, proximity to drainage, canopy height and species composition. These inputs come from ancillary data sources rather than the stress indices themselves. The output is a ranked map: high stress now, high vulnerability structurally, equals the areas where carbon loss or fire ignition is most probable in coming weeks. That is the product a forest manager or a government carbon-accounting team can act on.

Resolution floors and what they mean for interpretation

At 20 m, Sentinel-2 SWIR pixels in a heterogeneous canopy are mixed. A pixel covering a stressed emergent tree surrounded by healthy canopy will return a moderate MSI value, not a high one. The index reliably detects patch-level stress at roughly 1 hectare and above. Sub-hectare features are below the practical detection floor for this method with these sensors.

Landsat at 30 m has a coarser floor still, but its longer archive makes it the right choice for trend analysis over decades. MODIS is appropriate only for regional or national-scale drought monitoring, not for site-level decisions. None of these sensors can substitute for field measurement of live fuel moisture content, which varies at the individual-tree scale. The satellite signal is best treated as a spatial prioritisation tool: it tells you where to look, not what you will find when you get there. Satellize applies this workflow on Sentinel-2 and Landsat open data, with ECOSTRESS thermal layers added where ISS coverage permits, in the same analytical environment used for the Tonga crop-estimation programme.

Calibration, validation and the limits of index-based methods

MSI and LSWI are empirical indices, not physical retrievals of equivalent water thickness. Their relationship to actual canopy water content varies by species, canopy structure and understorey contribution to the mixed pixel. Published studies have reported correlations between LSWI and field-measured live fuel moisture in the range of r² 0.5 to 0.75 depending on forest type, which is useful but not precise. For carbon accounting or insurance applications that require quantified uncertainty, index-based stress maps should be accompanied by explicit confidence intervals and validated against field campaigns or airborne hyperspectral reference data.

Atmospheric correction quality also matters. Surface-reflectance products from Sentinel-2 (Level-2A) and Landsat Collection 2 reduce aerosol effects, but residual haze over fire-affected or dusty regions can inflate apparent SWIR reflectance and produce false stress signals. Quality-flagging aerosol optical depth alongside the stress product is good practice. These are not reasons to avoid the method; they are reasons to be specific about what it measures and what it does not.

Typical figures

Primary spatial resolution (SWIR bands)	20 m (Sentinel-2), 30 m (Landsat 8/9 OLI)
Revisit frequency	5 days (Sentinel-2A+B combined at equator); 8 days (Landsat 8+9 combined); daily at 500 m (MODIS)
Thermal resolution (ECOSTRESS)	Approximately 70 m; non-systematic revisit due to ISS orbit
Key spectral bands	NIR ~865 nm, SWIR-1 ~1610 nm, SWIR-2 ~2190 nm; thermal 8.28–12.13 µm (ECOSTRESS)
Minimum detectable stressed patch	Approximately 1 ha at 20 m SWIR resolution under clear-sky conditions
Cloud-free acquisition fraction (humid tropics)	Typically 20–50% of scenes; compositing over 30-day windows recommended
Archive depth	Sentinel-2 from 2015; Landsat OLI from 2013; MODIS from 2000
Latency (open data)	Sentinel-2 Level-2A within ~3 hours of acquisition; Landsat Collection 2 within 12–24 hours
Coverage	Global tropical belt; ECOSTRESS limited to ISS inclination ±51.6° latitude
Delivery formats	GeoTIFF stress-index rasters, anomaly maps, vector risk-zone polygons, time-series CSV

Analytics Satellize can run

Canopy moisture stress map (MSI / LSWI)	Band-ratio indices from Sentinel-2 Level-2A surface reflectance (NIR and SWIR-1 bands); cloud masking via s2cloudless or Sen2Cor	GeoTIFF raster layer, per-scene or monthly composite, with cloud-fraction metadata
Drought anomaly layer	Z-score of current MSI or LSWI against multi-year seasonal baseline (Landsat or Sentinel-2 archive); pixel-wise statistics	Anomaly raster classified into stress severity tiers, delivered as GIS polygon layer with area statistics
Drought vulnerability index	Composite scoring of stress anomaly, soil drainage class, canopy height proxy and historical drought frequency; weighted overlay	Ranked vulnerability map in GeoTIFF and PDF report with methodology appendix
ECOSTRESS thermal stress overlay	NASA ECOSTRESS ET Stress Index product (ESI) fused with Sentinel-2 LSWI anomaly by spatial resampling and temporal matching	Multi-layer GeoTIFF combining optical and thermal stress signals; flagged where temporal gap between acquisitions exceeds 5 days
Regional drought monitoring bulletin	MODIS MOD13 NDVI and EVI anomaly at 500 m, 16-day composite; basin-scale drought front tracking	Fortnightly PDF bulletin with maps and time-series charts, suitable for national forest authority reporting
Fire-risk pre-conditioning alert	Threshold exceedance on MSI anomaly combined with antecedent rainfall deficit from published gridded precipitation products (e.g. CHIRPS); alert logic documented and auditable	Automated alert feed (GeoJSON polygon) when stress anomaly exceeds defined threshold in a specified area of interest

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.