Forest canopy roughness and aerodynamic resistance mapping for flux modelling
Spaceborne lidar and SAR texture can replace field-measured roughness lengths in Penman-Monteith and similar flux models, but sampling gaps and stand heterogeneity introduce real uncertainty that buyers must understand before operationalising.
Sensors
- GEDI (NASA Global Ecosystem Dynamics Investigation): Full-waveform lidar aboard the ISS. Footprints of approximately 25 m diameter, spaced 60 m along-track and 600 m across-track. Delivers height percentiles (RH25 to RH100), canopy cover and plant area index. Sampling is non-wall-to-wall; within-stand spatial gaps are the primary source of uncertainty for roughness parameterisation.
- ICESat-2 ATL08: Photon-counting lidar. Six 11-m beams, 100-m segment canopy and terrain height products. Complements GEDI in high-latitude forests where GEDI's ISS orbit does not reach above 51.6° latitude, and provides independent height validation.
- Sentinel-1 C-SAR: C-band (5.405 GHz) synthetic aperture radar. IW mode at 10 m spatial resolution, 6-day repeat at mid-latitudes with two satellites. Backscatter texture metrics (GLCM contrast, homogeneity, entropy) correlate with canopy structural complexity and are used to spatially interpolate GEDI height percentiles between sparse footprints.
- TanDEM-X: X-band interferometric SAR pair providing a global digital surface model at 12 m posting. The difference between TanDEM-X DSM and a terrain model yields canopy height estimates, though X-band penetration is limited (typically less than 5 m into closed tropical canopy), so heights are systematically underestimated in dense stands.
Why roughness length matters more than it sounds
Aerodynamic resistance to sensible heat and water-vapour transfer is not a fixed property of a forest; it changes with canopy height, the zero-plane displacement height and the roughness length for momentum. In the Penman-Monteith framework, aerodynamic resistance appears in the denominator of the latent-heat flux term. Get it wrong by a factor of two and your evapotranspiration estimate shifts by tens of millimetres per month across a landscape. That error propagates directly into regional water-balance and carbon-flux models.
Traditionally, these parameters are measured at eddy-covariance towers. There are perhaps a few hundred well-maintained towers globally, mostly in temperate and boreal forests. Tropical forests, where the flux uncertainty matters most for global carbon accounting, are severely under-instrumented. Spaceborne lidar and SAR offer a route to spatially continuous parameterisation, with known and quantifiable limitations.
From waveform percentiles to roughness parameters
The standard approximations used in micrometeorology relate roughness length (z0) and zero-plane displacement (d) to mean canopy height (h): d is typically taken as 0.67h and z0 as 0.1h, though published coefficients vary by forest type. GEDI's relative height metrics, particularly RH75 and RH100, provide the height inputs. RH75 (the height below which 75 per cent of returned energy falls) is a reasonable proxy for the effective aerodynamic height of a heterogeneous canopy, more so than the top-of-canopy RH100, which responds to individual emergent crowns.
The problem is GEDI's sampling geometry. A single overpass leaves 600 m gaps across-track. Fusing GEDI height percentiles with Sentinel-1 backscatter texture bridges those gaps through regression or machine-learning models trained on co-located footprints. GLCM (grey-level co-occurrence matrix) entropy and contrast computed from Sentinel-1 VV and VH backscatter at 10 m resolution capture canopy structural variation at scales relevant to roughness. Published studies report RMSE values for predicted canopy height of roughly 3 to 6 m in tropical forests using this approach, depending on stand density and terrain complexity.
TanDEM-X fills gaps; physics sets the ceiling
Where GEDI sampling is thin, TanDEM-X interferometric height provides a spatially continuous alternative. The caveat is real: C- and X-band radar signals scatter from the upper canopy surface rather than penetrating to the ground. In closed tropical forest, the phase centre of X-band returns typically sits 3 to 8 m below the true top-of-canopy, and the bias is not constant across moisture conditions or stand density. Corrections require either ground-truth lidar or empirical adjustment using co-located GEDI footprints.
For boreal and temperate forests with lower, more open canopies, X-band penetration bias is smaller and TanDEM-X height models are more directly usable. The practical workflow is to use GEDI and ICESat-2 ATL08 as the primary height reference, apply Sentinel-1 texture to spatially extend height predictions, and use TanDEM-X as a secondary constraint where lidar sampling is absent entirely.
Where the method breaks and what to do about it
Forest edges are the hardest terrain for this workflow. A 25 m GEDI footprint straddling a canopy-to-clearing boundary produces a mixed waveform that overestimates mean height for the clearing and underestimates it for the forest interior. Edge-affected footprints should be flagged and excluded from the training dataset for any spatial interpolation model. The GEDI L2A quality flag (quality_flag = 1) removes many but not all problematic footprints.
Cloud is not a problem for lidar or SAR directly, but GEDI's ISS orbit means any given 25 m location may receive only a handful of footprints over the full mission lifetime, and not all will pass quality filters. In regions with persistent cloud, optical-derived canopy height products (such as those from the Global Land Analysis and Discovery group at the University of Maryland) can supplement the record, though their vertical accuracy is lower. Buyers should request uncertainty maps alongside any roughness-length layer, not just the point estimates.
Connecting the roughness map to a working flux model
A spatially distributed roughness-length layer enters Penman-Monteith or Surface Energy Balance System (SEBS) models as a gridded input alongside net radiation, air temperature, humidity and wind speed from meteorological reanalysis products such as ERA5. The spatial resolution of the roughness layer (typically 100 m to 500 m after aggregation from the 10 to 25 m sensor inputs) should be matched to the resolution of the flux model grid to avoid introducing spurious spatial gradients.
Validation against eddy-covariance tower data remains the only rigorous way to close the uncertainty budget. Where towers exist, the workflow should be calibrated so that modelled aerodynamic resistance at the tower footprint matches the tower-derived value within an accepted tolerance before the layer is extended across the landscape. Satellize's analytics pipeline can ingest GEDI, Sentinel-1 and TanDEM-X inputs and deliver calibrated roughness-parameter grids as GIS layers ready for direct ingestion into client flux models, following the same open-data methodology applied in the Tonga crop-estimation programme.
The honest summary: spaceborne roughness parameterisation reduces the cost of landscape-scale flux modelling substantially, but it does not eliminate uncertainty. It moves the dominant error source from unmeasured spatial variability to quantifiable sensor and model uncertainty, which is a meaningful improvement.
Typical figures
| GEDI footprint diameter | ~25 m; 60 m along-track spacing, 600 m across-track spacing |
| ICESat-2 ATL08 segment length | 100 m; 6 beams, ~17 m beam spacing within pairs |
| Sentinel-1 IW spatial resolution | 10 m (multi-looked); 6-day repeat at mid-latitudes (two-satellite) |
| TanDEM-X DSM posting | 12 m global; height accuracy ~2 m LE90 over flat terrain, higher error on slopes |
| Canopy height prediction RMSE (lidar + SAR fusion) | 3 to 6 m in tropical closed-canopy forest (published range; varies by method and site) |
| Roughness-length output resolution | Typically 100 to 500 m after spatial aggregation, depending on application |
| GEDI mission coverage | 51.6°S to 51.6°N (ISS orbit constraint); no polar coverage |
| TanDEM-X X-band penetration bias (closed tropical canopy) | 3 to 8 m systematic underestimate of top-of-canopy height |
| GEDI archive depth | April 2019 to present (ongoing); data available via NASA Earthdata |
| Deliverable formats | GeoTIFF (height percentiles, z0, d layers), NetCDF (flux-model-ready), uncertainty rasters |
Analytics Satellize can run
| Canopy height percentile mosaic | GEDI L2A RH metrics spatially extended using Sentinel-1 GLCM texture regression | GeoTIFF layers at 25 to 100 m resolution: RH25, RH50, RH75, RH100 with per-pixel uncertainty |
| Zero-plane displacement and roughness-length grid | Standard micrometeorology parameterisations (d = 0.67h, z0 = 0.1h) applied to fused height mosaic with forest-type adjustment | GeoTIFF layers of d and z0 ready for Penman-Monteith or SEBS model ingestion |
| Aerodynamic resistance surface | Ra computed from z0, d, wind speed from ERA5 reanalysis using logarithmic wind-profile equations | Monthly or seasonal NetCDF grids of aerodynamic resistance at client-specified resolution |
| Structural complexity index from SAR texture | GLCM entropy, contrast and homogeneity from Sentinel-1 VV/VH multi-temporal composites | GIS layer indicating within-stand heterogeneity; used to flag high-uncertainty zones in roughness maps |
| Edge-contamination mask | Forest-edge proximity analysis combined with GEDI quality-flag filtering | Binary GeoTIFF mask identifying footprints and pixels excluded from roughness parameterisation |
| Tower-calibrated roughness validation report | Comparison of modelled aerodynamic resistance against published eddy-covariance tower records at co-located sites | PDF report with bias, RMSE and scatter plots; used to bound landscape-scale uncertainty before operational deployment |
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.