Reading the Water: We Built a Gulf Stream Front Finder for Bimini

A single water temperature reading is close to useless for finding fish. It tells you the water is 30 degrees. Fine. What a captain wants is the edge — where 30 sits next to 28 within a couple of miles, because that seam concentrates bait and everything that eats it. One reading cannot show you that. You need the field. None of which replaces captains who already read this water for a living.

They do it off a plotter, a sea-surface chart and thirty years of pattern recognition. What a gradient map adds is a way to show a visitor what the thing they keep hearing about actually looks like from above — and, as it turns out, when it is worth looking for at all. It is a small piece of arithmetic and nobody appears to have run it for this island, so we did.

What a front is, precisely

A front is a gradient, not a temperature. Formally: the magnitude of the horizontal derivative of sea surface temperature, in degrees per unit distance. Informally: how fast the water changes as you drive across it. Thats the whole idea.

Off Bimini the setting is unusually favourable for one. The Gulf Stream runs north along the western wall carrying warm water up from the tropics, and immediately to the east sits the Great Bahama Bank — shallow, sun-baked, and thermally nothing like the Stream. Two water masses, a hard bathymetric boundary between them, and a current shearing along it. If a seam is going to form anywhere it forms there.

Two runs, one colour scale

We sampled a box from 25.5 to 26°N and 79.1 to 79.6°W — a bit over 30 miles a side, all of it water a boat out of Alice Town could reach in a morning. Two dates: 25 July 2026 and 25 January 2026. Both panels below use the same colour scale, which is the whole reason they are worth putting side by side.

The finding: winter seams are sharper

On the summer run the entire box spans 1.43 °C. Warmest corner to coldest corner, across thirty-odd miles of ocean, less than a degree and a half. The sharpest single cell manages 0.175 °C per km and the top five per cent of cells sit above 0.09. It is, in the technical sense, a fairly boring picture. The water is warm and its warm everywhere.

On the winter run the same box spans 4 °C and the top five per cent of cells sit above 0.171 °C per km — about 1.9 times the summer figure. In old money that is roughly 0.6 °F per nautical mile in the sharper water, which is the kind of change you notice on a gauge while the boat is moving.

In summer the whole box spans a degree and a half. In January it spans four.

Why it matters: the winter fishery here is wahoo, November through February, and wahoo work temperature edges. Our two runs do not prove a causal link and we are not going to pretend they do — two dates is two dates. But it does mean the advice “find the break” is much more actionable in the months people come for the wahoo than in the months they come for the mahi, which is the opposite of how most fishing content presents it. Wahoo work these seams in winter, and by contrast mahi sit on the weedline that forms along them in a season when, on this evidence, the seams themselves are faint.

How it is computed

Source data is NASA JPL’s MUR SST v4.1 analysis at 0.01 degree resolution, roughly a kilometre, served through NOAA CoastWatch ERDDAP. For each interior cell we take central differences in both directions, convert to degrees per kilometre using the real metres-per-degree at that row’s latitude, and take the magnitude of the vector. Longitude degrees are about ten per cent shorter than latitude degrees here, and skipping that correction would exaggerate every east-west gradient by the same amount.

Edge cells of the grid have no central difference available. They are recorded as null and drawn as nothing. Nothing is interpolated to fill them, and no cell is invented. 2601 of 2601 cells returned a value on both runs, which is what you would expect for a box that is entirely open water.

The whole derived field is downloadable as a licensed JSON file with the method and units attached, and the script that produces it lives in the repository. If you want to re-run it for a different date or a different box, that is the point.

What we can’t build, and why

This is the part we think is genuinely worth publishing, because the absence of these things explains a lot about why fishing apps for this region are so thin.

• There is no live “fish are biting” feed. Not at any price, from anyone. Every product that appears to offer one is either a user-report social feed or a repackaged weather model.
• There is no public OCEARCH API. The shark tracker everyone links to has no documented public interface, so nothing here should be architected around it.
• There is no NOAA tide station in Bahamian waters. None. The nearest is Virginia Key, Florida, about 50 nautical miles away on the far side of the Gulf Stream. We show that reading elsewhere on the site and we label it as what it is: a reference for the pattern, not a Bimini prediction.
• The nearest Bahamian buoy is not reporting. NDBC lists Settlement Point (SPGF1), and its real-time data file returned a 404 when we checked again on 26 July 2026. The nearest station actually returning data is 41122 off Hollywood Beach, Florida — again, the wrong side of the Stream.

So the honest position is that satellite SST is the best remote information available for this patch of ocean, and its a snapshot of the surface with no wind, no current shear and no depth in it. Treat what is above as one input. Planning around weather windows matters more than any of it, and the geology underneath is more permanent than either — we measured that in the depth transect piece, and this section’s wider argument about what makes this island fish sits behind both.