Double the far-zoom heat spots so they read as spots

At 10 km per pixel each fire was a marker-sized spot that faded into the
imagery from the globe view. A 2048 px grid with the same one-pixel blur
makes every spot exactly twice as wide (~60 km for a lone weak fire),
still separate, and the build is three times faster.

changelog.d/1426-fire-intensity-filter.md

@@ -4,6 +4,6 @@
   log-scale minimum-FRP slider whose track is the ramp, an All / Nominal+ /
   High confidence switch and an "N of M shown" counter to the layer row. Far
   zoom shows a Snap Map–style heatmap in the same colors instead of squares,
-  one marker-sized spot per fire, empty where nothing burns. The filter applies once before rendering, so markers, cards, the heatmap, context and the top-fires query agree; share
+  one distinct spot per fire, empty where nothing burns. The filter applies once before rendering, so markers, cards, the heatmap, context and the top-fires query agree; share
   links keep it (`w` token: `f`, `c`), and `get_active_fires` takes matching
   `min_frp_mw` and `confidence` arguments. Default 0 MW / All (gekh, #1426).

docs/CURRENT-STATE.md

@@ -2185,10 +2185,10 @@ accents read that ramp; confidence and brightness do not affect the color.
 
 At far zoom (camera above ~3,000 km) the layer drapes a Snap Map–style
 heatmap instead of individual markers: the shown detections are binned on a
-global 4096 × 2048 grid (≈10 km per pixel), each weighted by its FRP in megawatts (so heat is
+global 2048 × 1024 grid (≈20 km per pixel), each weighted by its FRP in megawatts (so heat is
 radiated power, and raising the FRP slider does not dim areas where strong fires
-burn), Gaussian-blurred by one pixel so each fire stays a spot about the size
-of its marker,
+burn), Gaussian-blurred by one pixel so each fire stays a distinct spot about
+twice the size of its marker,
 normalized on a log scale against the 99.5th percentile and mapped through the
 same hues, from a bright red edge through orange and gold to a pale
 white-hot core. The edge starts more than half opaque, so regions far below

src/layers/firms/heatImagery.test.mjs

@@ -88,7 +88,7 @@ test('destroy removes the layer', () => {
   assert.equal(h.layers.length, 0);
 });
 
-test('the shipped raster paints a lone weak fire as a marker-sized spot', () => {
+test('the shipped raster paints a lone weak fire as a distinct spot', () => {
   // A hot cluster elsewhere sets the scale, as Africa does on real data.
   const fires = [{ lat: 0, lon: 0, frp: 5 }];
   for (let i = 0; i < 40; i++) {
@@ -106,7 +106,7 @@ test('the shipped raster paints a lone weak fire as a marker-sized spot', () =>
   const kmPerPx = 40075 / r.width;
   assert.ok(painted.length > 0, 'the weak fire still shows');
   assert.ok(
-    painted.length * kmPerPx <= 80,
+    painted.length * kmPerPx <= 100,
     `spot ${Math.round(painted.length * kmPerPx)} km wide`,
   );
 });

src/layers/firms/policy.js

@@ -49,16 +49,17 @@ export const LOD_LEVELS = [
 ];
 
 /**
- * Global heat raster for the `heat` bands: ≈0.09° (≈10 km) per pixel and a
- * one-pixel Gaussian σ, so a lone fire is a spot about as wide as its marker
- * at the 3,000 km hand-off and neighbours merge only when they really touch
- * (2048 px with σ 3 smeared each fire over ~200 km). Built once per data
- * refresh or filter change, never per camera move.
+ * Global heat raster for the `heat` bands: ≈0.18° (≈20 km) per pixel and a
+ * one-pixel Gaussian σ, so a lone fire is a distinct spot about twice as wide
+ * as its marker at the 3,000 km hand-off and still reads from the
+ * whole-globe view; neighbours merge only when they nearly touch (σ 3 here
+ * smeared each fire over ~200 km). Built once per data refresh or filter
+ * change, never per camera move.
  */
 
 export const HEAT_RASTER = Object.freeze({
-  width: 4096,
-  height: 2048,
+  width: 2048,
+  height: 1024,
   sigmaPx: 1,
 });