Forest Health Monitoring with NDVI and SAR Vegetation Indices

The Challenge of Forest Monitoring

Tropical forests — where deforestation rates are highest — are also among the cloudiest regions on Earth. Persistent cloud cover over the Amazon, Congo Basin, and Southeast Asian forests can prevent any usable optical satellite observations for weeks at a time. But deforestation does not pause for clouds.

A comprehensive forest monitoring system needs to:

1. Detect clearing events rapidly (within days to weeks)
2. See through clouds when necessary
3. Distinguish complete clearing from selective logging or degradation
4. Provide early warning of stress before visible damage occurs

No single satellite sensor achieves all four requirements. Combining Sentinel-2 optical data with Sentinel-1 SAR data comes close.

What Each Index Measures in Forest

NDVI (Sentinel-2) — Photosynthetic Activity

NDVI measures the fraction of absorbed photosynthetically active radiation — essentially, how actively the canopy is photosynthesizing. For forests:

• High stable NDVI (0.7–0.9) → Healthy closed-canopy forest
• Declining NDVI → Canopy thinning, leaf loss from drought or disease, or partial clearing
• Abrupt drop to low values → Complete clearing
• Slowly recovering NDVI → Regrowth after disturbance

The limitation: clouds. NDVI is useless on cloudy days, and many observations will be contaminated by cloud even after masking.

RVI — Radar Vegetation Index (Sentinel-1)

SAR RVI measures the ratio of cross-polarized to total backscatter:

RVI = (4 × VH) / (VV + VH)

In forest, multiple scattering through the canopy volume produces high VH backscatter → high RVI. When clearing removes the canopy:

• VH drops (no more volume scattering)
• VV may remain moderate (ground and edge scattering)
• RVI drops sharply

Key advantage over NDVI: Works through clouds. An RVI observation is available regardless of weather.

Key limitation: Does not distinguish healthy forest from very dense grassland or shrubland that also produces volume scattering. Context and baseline knowledge are important.

RFDI — Radar Forest Degradation Index (Sentinel-1)

RFDI is specifically designed to detect partial forest degradation — the selective logging, edge effects, and low-intensity disturbances that are invisible to complete-clearing detectors:

RFDI = (VV − VH) / (VV + VH)

In undisturbed dense forest: VH is high (volume scattering) → RFDI is low After selective logging or edge degradation: VH decreases, VV relatively stable → RFDI increases

This makes RFDI sensitive to:

• Selective logging that removes large trees without complete clearing
• Edge degradation around clearing perimeters
• Windthrow and storm damage that reduces canopy closure
• Progressive thinning before complete clearing

VV and VH Intensity

Raw backscatter values provide the underlying data for the derived indices. Monitoring VV and VH separately is useful for:

• Cross-checking index anomalies
• Detecting flooding under the forest canopy (CR increases significantly)
• Monitoring forest management activities (logging creates temporary high VV from slash piles)

The Multi-Index Approach

Setting up forest monitoring with multiple indices simultaneously gives you different "views" of the same ecosystem:

Index	What It Detects	Cloud Impact
NDVI	Photosynthetic activity, leaf cover	Full (no observation in clouds)
RVI	Canopy volume, structure	None (all-weather)
RFDI	Degradation, partial disturbance	None (all-weather)

Strategy: Use NDVI as the primary vegetation health indicator when cloud-free observations are available. Use RVI/RFDI to fill cloud gaps and confirm optical anomalies.

Confirmation logic:

• NDVI drop + RVI drop → Likely real vegetation loss
• NDVI drop + RVI stable → Possible cloud contamination or non-structural change
• RVI drop + NDVI cloudy → Real change probable; wait for next cloud-free observation to confirm

Characteristic Forest Monitoring Signatures

Complete Deforestation

Optical (NDVI): Sharp drop from ~0.7+ to <0.2, permanent SAR (RVI): Sharp drop from ~0.6 to <0.3, permanent SAR (RFDI): Moderate increase as VH drops

The combination of simultaneous NDVI and RVI drops is a very high-confidence deforestation signal.

Selective Logging

Optical (NDVI): Small to moderate drop, recovery over months as surrounding canopy grows in SAR (RVI): Moderate drop proportional to canopy removed SAR (RFDI): Clear increase — this index is most sensitive to partial disturbance

Selective logging often shows a clear RFDI signal before NDVI shows any significant change.

Drought Stress

Optical (NDVI): Suppressed values during dry season, more pronounced than historical average SAR (RVI): Minor change (canopy structure mostly intact, but may decrease slightly) SAR (RFDI): Slight increase if drought causes leaf drop or canopy thinning

Drought stress primarily shows in optical data; SAR is less sensitive to water stress without structural change.

Fire and Burn

Optical (NDVI): Sharp drop during/after fire, followed by gradual recovery SAR (VV/VH): Decrease after fire (canopy removed) then gradual increase as vegetation regrows Optical (NBR): Most sensitive fire damage indicator; monitor NBR + NDVI together

Post-Fire Recovery

One of the most ecologically important monitoring applications is tracking forest recovery after fire. NBR and NDVI together reveal:

• Rate of initial green-up from herbaceous regrowth
• Whether woody vegetation is recovering at expected rates
• Whether recovery stalls (indicating soil damage or repeat disturbance)

Setting up monitoring that starts 3–6 months before a known fire event and continues for 2–3 years afterward gives a full picture of disturbance and recovery dynamics.

Monitoring in Temperate vs. Tropical Forests

Temperate Deciduous Forest

Strong seasonal NDVI cycle (low in winter, high in summer) is normal. Forest health anomalies appear as:

• Suppressed summer peak
• Earlier-than-normal autumn NDVI decline
• Failure to reach expected summer NDVI levels

SAR time series in temperate deciduous forest also shows seasonal variation as leaf-on/leaf-off changes volume scattering.

Tropical Evergreen Forest

Relatively flat NDVI year-round at high values (~0.7–0.9). Anomalies are easier to detect because the stable baseline means any deviation is significant. SAR indices are also relatively stable, making change detection more straightforward.

Practical Considerations

Polygon size for forest monitoring: For a meaningful forest patch assessment, you typically want a polygon covering at least 1–5 km². Smaller polygons are dominated by edge effects and single-pixel noise.

Start date for trend analysis: 2–3 years of history provides a good baseline for detecting subtle degradation trends. For event-based monitoring (recent fire or logging event), starting 6–12 months before the event is sufficient.

Combining with official data: Where available, deforestation alert systems can be overlaid to cross-check monitoring results. Comparing automated alert timing with your SAR time series confirms how quickly different methods detect the same event.

Summary

Comprehensive forest health monitoring combines Sentinel-2 NDVI (sensitive to photosynthetic activity, cloud-limited) with Sentinel-1 RVI and RFDI (sensitive to canopy structure and degradation, all-weather). NDVI detects clearing and drought stress when clouds allow observation; RVI fills cloud gaps with structural information; RFDI specifically responds to partial degradation before complete clearing. The two-sensor approach provides both cloud-free coverage and two independent change signals, making false alarms rare and real deforestation events nearly impossible to miss.