Getting Started With Ocean Titan Mapping Tools
Most people who try to map large ocean zones hit the same wall within the first hour. The tile resolution drops too low, the coordinate system drifts, and you end up with overlapping artifacts that look like a broken mosaic. I learned this the hard way back when I was working on a coastal survey project that needed both surface bathymetry and subsea feature annotation in the same pass. What follows is the workflow I ended up using after three failed attempts. This toolset is designed around two core components. The Leviathan module handles large-scale regional mapping with low-resolution base layers, while the Kraken module takes over for high-detail zone capture. You don't run them separately. They feed each other in a pipeline that most beginners try to shortcut, which is exactly where things fall apart. Here is what actually works in practice. Start by pulling your base data from whatever source you have access to. GEBCO grid data, NOAA bathymetry exports, or satellite-derived terrain layers all work. Load them into Leviathan first. Set your region bounds to slightly wider than your area of interest, maybe five to ten percent on each side. This buffer matters more than people realize because the edge tiling behavior in this tool is not symmetric, and running right at your boundary will drop corner tiles silently.
Once Leviathan finishes its initial pass, export the mid-res mesh. This is where the handoff to Kraken happens. Import the mesh, switch the capture mode to detail, and re-run over your actual zone of interest. The merge step uses a weighted blend around the boundary region. Give it a blend radius of at least two hundred meters or you will see hard seams in the final output. I ran into a specific problem during a project last year where the Kraken module was generating duplicate vertices along the merge boundary. The output file size ballooned to forty gigabytes for a region that should have been under ten. I spent two days trying to figure out if it was a data issue or a config issue. It turned out to be a projection mismatch. Leviathan was outputting in EPSG:3857 and Kraken was expecting EPSG:4326 by default. The workaround was straightforward but not documented anywhere obvious. I had to open the Kraken config file and add a projection override flag before the merge step. Without that flag, the vertex duplication happens every time on boundary regions, and the tool does not throw an error, it just produces garbage geometry. Another thing nobody mentions is the memory threshold. Leviathan can handle large regions without issue, but Kraken needs at least sixteen gigabytes of RAM available, preferably more if you are working with high-density point clouds. If you are below that threshold, the tool silently downgrades to a lower resolution internally. You will not get an error message. You will just get a blurry result and waste hours wondering why your detail pass looks like it was rendered on a potato. Check your available memory before hitting run.
The export options are where most people get confused. There are three main formats. OBJ for general 3D use, LAS for LiDAR-compatible workflows, and the native .lmf format which preserves all the metadata tags the tool generates. If you need to bring the data into another GIS platform, use LAS. If you are just visualizing or annotating features, .lmf is the fastest option and keeps everything in one file. OBJ is the worst of both worlds and I only recommend it if a third-party application specifically requires it. There are limitations you should know about upfront. This tool is not built for real-time mapping. The full pipeline from raw data import to final export for a medium-sized ocean region takes between four and eight hours depending on your hardware and the density settings. If you need faster turnaround, you have to reduce the point density, which directly reduces the accuracy of subsea feature detection. There is no way around that trade-off. The tool also struggles with polar regions due to the projection assumptions built into the base tiling algorithm. If you are mapping areas above sixty-five degrees north or below sixty-five degrees south, you will need to manually reproject your source data before feeding it into Leviathan. Otherwise the tile boundaries will shift and you will get misaligned seams that are extremely difficult to fix in post-processing.
Get the Full Details
For anyone just starting out, I would recommend using the built-in sample dataset first. It takes about twenty minutes to run end to end on a decent machine and gives you a realistic sense of what the output looks like before you commit hours to your own data. The default settings on the sample are actually reasonable, which is unusual for tools in this space. Download availability varies by platform. The Windows build is the most stable and receives updates first. The Linux version works but has occasional issues with certain GPU acceleration paths depending on your driver version. Mac support exists but is the least tested, and there have been reported crashes when processing regions larger than five thousand square kilometers.
