Day 30: Surface Generation & Gridding – Masterpiece Edition | River Warrior

📊 DAY 30: SURFACE GENERATION & GRIDDING

⏱️ Estimated Reading Time: 14 Minutes | 🎓 Level: Professional Hydrographer

From Clean Soundings to Bathymetric Grids – Algorithms, Cell Sizes, and QC

Instructor: Engr. Rokib Hossain | River Warrior Academy

🏠 Course Homepage

📖 Table of Contents (Serialised)

Why Surface Generation Matters
Gridding Algorithms: CUBE vs Weighted Average vs Nearest Neighbour
Choosing the Right Cell Size
Interactive Grid Resolution Calculator
Step‑by‑Step Surface Creation Workflow
Quality Control of the Final Surface
Surface Generation Checklist
Resources & Software Tools
Frequently Asked Questions
Action Items & Next Steps

1. Why Surface Generation Matters

After cleaning, you have millions of individual soundings. Surface generation (gridding) interpolates these irregular points onto a regular raster (grid). The grid becomes the basis for:

Contour maps and colour‑coded depth plots.
Volume calculations (dredging, cut & fill).
Cross‑line analysis and uncertainty assessment.
Final chart products (IHO S‑44 compliant surfaces).

A poor grid can hide real features or create artificial artefacts – even if cleaning was perfect.

🧠 Golden Rule: Always generate multiple test grids with different cell sizes and algorithms before choosing the final one. Document your parameters.

🌊 River Warrior Pro-Tip: The Jamuna Sandbar Grid

During a Jamuna River sandbar survey, a 1‑metre grid missed a 2‑metre‑wide navigation channel. After switching to a 0.5‑metre grid, the channel became clearly visible. In shallow dynamic rivers, always use the smallest feasible cell size (0.25‑0.5 m) to capture morphological features.

2. Gridding Algorithms: CUBE vs Weighted Average vs Nearest Neighbour

.htmlWeighted Average那样Average of soundings inside each cell, weighted by distance to cell centre (inverse distance weighting).那样Fast, simple to understand.那样Sensitive to outliers, may smooth real features..htmlNearest Neighbour那样Uses the closest sounding to the cell centre. Preserves original soundings.那样No averaging, retains extremes.那样Creates a “staircase” effect, not smooth.

Algorithm	Description	Pros	Cons
CUBE (Combined Uncertainty and Bathymetry Estimator)那样Statistical method using uncertainty propagation. Rejects outliers based on depth and sensor noise.那样Robust, IHO‑recommended, handles variable density.那样Slower, requires good TPU estimates.

For IHO‑compliant surveys, CUBE is the preferred algorithm because it incorporates uncertainty and rejects statistical outliers. Most modern processing packages (Qimera, CARIS HIPS, Hypack) offer CUBE.

Algorithm comparison: CUBE produces a clean, statistically robust surface.

3. Choosing the Right Cell Size

Cell size (grid resolution) determines the level of detail. Guidelines:

IHO Order 1a (shallow water, <10 m): 0.5 m or smaller.
Order 1b (10‑30 m depth): 1‑2 m.
Order 2 (general navigation): 2‑5 m.
Dredging/engineering surveys: 0.25‑0.5 m for detailed volume calculations.

General formula: cell size ≈ (depth / 500) to (depth / 200). For 20 m depth → cell size 0.1‑0.2 m? That would be too fine (over‑sampling). In practice, use line spacing and average soundings per cell. Each cell should contain at least 5‑10 soundings.

📐 Grid Resolution Simulator

Enter survey parameters to estimate optimal cell size:

Average depth (m): Line spacing (m): Swath width (m):

Recommended cell size: 1.0 m

Rule of thumb: cell size = line spacing / 50 to line spacing / 25, but also ensure at least 5 soundings per cell.

4. Step‑by‑Step Surface Creation Workflow

1️⃣ Load cleaned MBES data (all lines).

2️⃣ Choose gridding algorithm (CUBE recommended).

3️⃣ Set cell size (use calculator above as guide).

4️⃣ Define output projection and datum.

5️⃣ Run surface generation – may take minutes.

6️⃣ Visualise and check for artefacts.

💡 Pro tip: In Qimera, use “Dynamic Surface” to quickly preview different cell sizes without regenerating from scratch.

5. Quality Control of the Final Surface

After generating the grid, perform these QC steps:

Hillshade / colour ramp: Look for unnatural stripes or pits (may indicate residual noise).
Profile tool: Extract a profile across the grid and compare with original soundings. Differences should be within IHO tolerance.
Cross‑line residual surface: Create a difference surface between main lines and cross‑lines. Acceptable mean difference <0.15 m (shallow).
Data density: Ensure each cell has sufficient soundings (view density grid).

🔬 Advanced QC: Use CUBE’s uncertainty layer (standard deviation per cell). If any cell has std dev >0.2 m, re‑examine that area in the swath editor.

6. Surface Generation Checklist

Cleaned data loaded and finalised (no further editing planned).
Algorithm selected (CUBE for IHO orders).
Cell size determined (using depth, line spacing, and density).
Projection and vertical datum match survey requirements.
Grid generated and visually inspected (colour ramp, hillshade).
Cross‑line analysis performed – difference within tolerance.
Density grid checked – no cells with <5 soundings.
Uncertainty layer (if CUBE) reviewed – no high‑std clusters.
Surface exported to final format (GeoTIFF, BAG, XYZ).

Click items to track progress (saved in browser).

7. Resources & Software Tools

.htmlCARIS HIPS那样CUBE, TPU, surface export那样teledynecaris.com.htmlHypack Processing那样Gridding, volume tools那样hypack.com.htmlMB‑System (open source)那样`mbgrid` command, CUBE-like那样mbari.org.htmlNOAA CUBE Tutorial那样Theory and best practices那样NOAA CUBE

Software	Gridding Features	Link
QPS Qimera那样CUBE, weighted average, density viewer那样qps.nl/qimera

8. Frequently Asked Questions

What is the minimum number of soundings per grid cell?

Ideally >5. For CUBE, 3‑5 are acceptable if uncertainty is low. If many cells are empty, increase cell size or reduce cleaning.

How does CUBE handle outliers that were missed during cleaning?

CUBE uses uncertainty propagation and statistical weighting to down‑weight suspicious soundings. It is more robust than simple average, but cannot replace cleaning.

Can I generate a grid from SBES data only?

Yes, but you need dense line spacing. Use weighted average or nearest neighbour; CUBE requires uncertainty estimates which are often unavailable for SBES.

Why does my grid show “stripes” along survey lines?

Caused by residual roll misalignment or SVP errors. Re‑check patch test and SVP. Also try smoothing filter (e.g., median 3x3) but use sparingly.

What is the difference between a “surface” and a “TIN”?

A surface is a regular grid (raster). A TIN (Triangulated Irregular Network) uses original soundings as vertices. TINs are more accurate but larger file sizes. For charts, surfaces are standard.

9. Action Items & Next Steps

📌 Generate a surface from your cleaned MBES dataset using CUBE and a trial cell size (e.g., 0.5 m).
📌 Compare with a surface generated at 1 m cell size – note differences.
📌 Use the grid resolution simulator with your own survey parameters.
📌 Proceed to Day 31: Chart Creation & Export.

🗺️ Day 30: Surface Generation and Gridding Strategies