Logging road erosion and sediment delivery risk assessment
Newly cut logging roads are the dominant sediment source in logged catchments, yet they rarely appear in carbon or biodiversity assessments. Satellite-derived road networks, slope models and rainfall erosivity layers can quantify sediment delivery risk to downstream water bodies before field crews arrive.
Sensors
- Sentinel-1 SAR (C-band, ESA): 10 m ground range detected imagery, 6-day repeat at the equator with both satellites active. C-band backscatter lineaments reveal cleared road surfaces even under persistent cloud cover; double-bounce from road edges against standing trees adds a secondary detection signal.
- Planet SkySat (optical, sub-metre): 0.5 m panchromatic, ~0.8 m multispectral. Tasked on demand for road-width measurement, surface condition classification and confirmation of SAR-detected lineaments. No free archive; commercial tasking required.
- Maxar WorldView-3 (optical, very high resolution): 0.31 m panchromatic, 1.24 m multispectral, 3.7 m SWIR. Resolves road drainage structures, cut slopes and bare-soil extent. Revisit 1-4.5 days depending on latitude and tasking priority; archive depth back to 2014.
- TanDEM-X DEM (X-band radar, DLR): 12 m posting, absolute vertical accuracy better than 10 m, relative accuracy around 2 m over forested terrain. The primary slope and contributing-area source for sediment delivery modelling; superior to SRTM in forested catchments because acquisition geometry reduces layover artefacts.
- SRTM (C/X-band, NASA/NGA): 30 m posting globally, freely available. Adequate for catchment-scale slope classification where TanDEM-X is not licensed; vertical error in steep forested terrain can reach 6-10 m, which propagates into slope uncertainty and should be declared in any risk map.
Why roads, not felling, drive the sediment budget
The intuition that tree removal is the main sediment risk in a logged landscape is wrong. Published catchment studies from Borneo, the Pacific Northwest and the Amazon consistently show that unpaved logging roads contribute the majority of suspended sediment load, often by an order of magnitude more than skid trails or felling gaps, because roads intercept subsurface flow, concentrate it and discharge it directly to stream channels through culverts and cross-drains. A road on a 30-degree slope with poorly designed drainage can produce sediment yields per unit area that are 100 to 200 times higher than the surrounding undisturbed forest floor.
The problem compounds over time. Freshly cut roads are worst in the first two to three wet seasons, when the exposed subsoil has not yet crusted or revegetated. Culvert failures and road-surface rilling during high-intensity rainfall events then deliver pulses of sediment that can travel tens of kilometres downstream, raising turbidity, smothering spawning gravels and increasing treatment costs for downstream water utilities. None of this appears in a standard deforestation alert or carbon-stock assessment.
Detecting roads from space: what the physics allows
SAR backscatter at C-band (Sentinel-1, 5.4 GHz) responds to road surfaces differently from forest canopy. Bare mineral soil and compacted gravel return low backscatter in VV polarisation, creating dark lineaments against the brighter surrounding canopy. Where road embankments create a vertical face adjacent to standing trees, a double-bounce mechanism can produce locally elevated returns, making the road edge detectable even when the road surface itself is narrow. At 10 m resolution, roads wider than roughly 4-5 m are consistently detectable; narrower skid trails are not. Cloud cover is irrelevant to SAR, which matters enormously in tropical logging regions where optical revisit can be blocked for weeks.
Sub-metre optical imagery from SkySat or WorldView-3 resolves road width to within half a metre and can distinguish bare cut slopes from revegetating ones using near-infrared reflectance. The SWIR bands on WorldView-3 (eight bands between 1.19 and 2.37 micrometres) are particularly useful for identifying exposed clay minerals in fresh road cuts, which correlates with erodibility. The honest limit is cost: commercial tasking over large, remote concessions is expensive, so optical data is best used to validate and characterise SAR-detected lineaments rather than to run primary detection across entire watersheds.
From road network to sediment delivery ratio
Detecting roads is the first step. Translating their presence into a sediment delivery risk number requires three additional layers: slope, soil erodibility and rainfall erosivity. Slope is derived from the TanDEM-X or SRTM DEM using standard hydrological conditioning (pit-filling, flow-direction, contributing-area calculation). The Universal Soil Loss Equation (USLE) and its revised form (RUSLE) provide the published framework for combining these inputs; the road segment replaces the standard slope-length factor with a road-specific drainage-length term that accounts for whether runoff is routed to a stream or dispersed into the forest floor.
Sediment delivery ratio, the fraction of eroded material that actually reaches a water body, is then estimated from the road network topology relative to the stream network. A road that drains directly into a first-order stream via a failing culvert has a delivery ratio close to 1. A road on a broad ridge with vegetated buffer strips between it and the nearest channel may have a ratio below 0.1. This topology can be extracted automatically from the DEM-derived stream network and the detected road lineaments, producing a per-segment risk score that can be aggregated to sub-catchment or concession level.
Rainfall erosivity (the R-factor in RUSLE) is typically sourced from published global datasets such as the Global Rainfall Erosivity Database (GloREDa), which covers most tropical logging regions. Where local rain-gauge data exists, it should replace the global estimate; the global product carries uncertainty of roughly 20-30 percent in data-sparse regions, which is worth stating explicitly in any delivered risk map.
What the maps cannot tell you
Satellite-derived sediment risk assessments have real limits that any honest briefing must name. First, road surface condition matters: a well-gravelled, properly drained road on a steep slope may produce less sediment than a poorly maintained road on a gentle one. Satellite imagery can proxy surface condition through spectral indices and texture, but cannot replace a field inspection for culvert sizing or drainage structure integrity.
Second, DEM vertical error propagates directly into slope uncertainty. On slopes between 15 and 25 degrees, a 2 m vertical error in TanDEM-X translates to a slope error of roughly 2-4 degrees, which is meaningful when applying RUSLE thresholds. SRTM errors in steep forested terrain can be larger. Third, the approach models potential sediment delivery, not measured loads. Validation against turbidity sensors or suspended sediment gauges is strongly recommended before using outputs for regulatory or compensation purposes. The maps are best understood as a prioritisation tool: they identify which road segments and sub-catchments warrant urgent field attention, not what the sediment load in kilograms per year actually is.
Fitting this into a concession management workflow
The practical value of a satellite-derived road erosion assessment is speed and coverage. A field team might survey 20 km of road per day; a SAR-based road detection pass over a 50,000-hectare concession can map hundreds of kilometres of road network in a single processing run, updated after every six-day Sentinel-1 repeat. Newly cut roads, identifiable by the absence of the lineament in the previous-period image, can be flagged automatically and assigned a risk score before the next wet season arrives.
Outputs feed directly into reduced-impact logging (RIL) compliance checks, watershed protection planning and downstream water-quality risk disclosures. For concessions operating under FSC or PEFC certification, documented road erosion risk assessments are increasingly expected as part of the environmental management plan. Satellize runs this type of multi-layer catchment analysis on open and commercial imagery stacks; the Tonga crop-estimation programme is a different domain, but the underlying workflow of combining DEM-derived terrain attributes with change-detected surface features is directly transferable. Analysts wanting to scope a catchment-level assessment can request a pilot area run covering a defined sub-catchment with a delivered risk map and methodology note.
Typical figures
| Road detection spatial resolution | 10 m (Sentinel-1 SAR); 0.5 m panchromatic (SkySat); 0.31 m panchromatic (WorldView-3) |
| Minimum detectable road width (SAR) | Approximately 4-5 m in C-band backscatter lineament analysis; narrower tracks require sub-metre optical confirmation |
| DEM vertical accuracy | TanDEM-X: ~2 m relative, <10 m absolute; SRTM: 6-10 m in steep forested terrain |
| DEM spatial posting | TanDEM-X: 12 m; SRTM: 30 m globally |
| SAR revisit (Sentinel-1) | 6 days at equator (both satellites); single-satellite revisit 12 days |
| Optical tasking revisit (WorldView-3) | 1-4.5 days depending on latitude and tasking priority |
| Spectral bands used | SAR C-band VV/VH; optical visible, NIR, SWIR (WorldView-3 8-band SWIR, 1.19-2.37 µm) |
| Cloud penetration | SAR: full cloud penetration. Optical: blocked by cloud; not suitable as primary sensor in persistently cloudy tropical regions |
| Archive depth | Sentinel-1: from 2014; SRTM: single acquisition 2000; TanDEM-X: 2010-2015 acquisition; WorldView-3: from 2014 |
| Rainfall erosivity source | Global Rainfall Erosivity Database (GloREDa); uncertainty ±20-30% in data-sparse tropical regions |
Analytics Satellize can run
| Logging road network map | SAR backscatter lineament extraction (VV/VH ratio thresholding and directional filtering) combined with sub-metre optical edge detection | GIS polygon and centreline layer (GeoPackage or Shapefile) with road-width attribute and detection confidence score |
| New road detection alert | Bi-temporal SAR change detection: lineaments present in current epoch absent in baseline epoch flagged as newly cut | Automated alert layer delivered within 48 hours of Sentinel-1 acquisition, with area and estimated road length per alert polygon |
| Per-segment RUSLE erosion potential score | Revised Universal Soil Loss Equation applied to road segments using TanDEM-X slope, published K-factor grids and GloREDa R-factor | Road segment GIS layer with RUSLE LS-factor, K-factor and composite erosion potential score per 50 m segment |
| Sediment delivery ratio map | DEM-derived flow-path topology linking road drainage points to stream network; delivery ratio estimated from buffer width and slope break between road and channel | Sub-catchment raster and summary table of estimated sediment delivery ratio, aggregated to concession block level |
| High-risk road segment prioritisation report | Composite scoring of erosion potential, delivery ratio and proximity to protected water bodies; ranked segment list for field inspection | PDF report with ranked segment map, methodology note and declared uncertainty ranges; field-ready coordinate list for ground-truthing |
| Temporal road condition change series | Multi-date SAR backscatter and optical NDVI time series over detected road surfaces to proxy revegetation and surface stabilisation | Annual GIS update layer showing road segments classified as active, partially revegetated or stabilised, with date of last status change |
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.