Update seasonal scores from sounding data, document commercial link patterns

Soundings (10,935): October has strongest gradients (-350 N/km) and
65.7% ducting, raised seasonal score 82→88. February ducting (20.2%)
exceeds January (14.8%), raised 32→40.

Commercial links (12,068 samples): 68 GHz shows 3.9 dB diurnal swing,
11 GHz shows 1.7 dB with inverted multipath pattern, 24 GHz remarkably
stable at 0.9 dB. Diurnal sensitivity is non-monotonic with frequency.

RAOB gradients avg -265 N/km vs HRRR -107 (2.5x stronger) — confirms
HRRR misses thin surface ducts that soundings resolve.

algo.md

@@ -572,6 +572,12 @@ Binned distance analysis of 37,925 HRRR-matched contacts confirms and refines Pa
 
 **Mode distance advantage quantified.** CW: 247 km avg, PH (SSB): 191 km avg, FM: 141 km avg at 10 GHz. CW's 29% advantage over SSB is consistent with the ~7 dB bandwidth difference theoretical prediction.
 
+**RAOB gradients are 2.5x stronger than HRRR.** 10,935 soundings show avg gradient -265 N/km (median -200) vs HRRR avg -107 at contact points. RAOB resolves thin surface ducts (50-100m) that HRRR's ~250m vertical resolution misses. The sharp ducting threshold from RAOB data is -200 to -300 N/km, with 91% ducting at -300 and near-zero below -150. This confirms that HRRR refractivity scoring is inherently limited.
+
+**October seasonal score raised.** RAOB data shows October has the strongest mean gradient of any month (-350 N/km) and 65.7% ducting (3rd highest behind Jun 70% and Jul 71%). The 10 GHz seasonal score was raised from 82 to 88. February raised from 32 to 40 (20.2% ducting exceeds January's 14.8%).
+
+**Commercial link diurnal patterns (12,068 samples).** 68 GHz (2.82 km) shows 3.9 dB diurnal swing — morning best (-50.7 dBm at 09 UTC), afternoon worst (-54.6 at 13 UTC). 11 GHz (5.66 km) shows inverted pattern: 1.7 dB swing with more multipath variability at night. 24 GHz (4.36 km) is remarkably stable at only 0.9 dB swing. Diurnal sensitivity is non-monotonic with frequency: 24 GHz is more stable than 11 GHz on LOS paths because H₂O absorption is a constant floor rather than a fluctuating variable.
+
 ---
 
 ## Part 3: Band Configuration

lib/microwaveprop/propagation/band_config.ex

@@ -57,7 +57,7 @@ defmodule Microwaveprop.Propagation.BandConfig do
       rain_alpha: 1.28,
       seasonal_base: %{
         1 => 38,
-        2 => 32,
+        2 => 40,
         3 => 22,
         4 => 55,
         5 => 68,
@@ -65,7 +65,7 @@ defmodule Microwaveprop.Propagation.BandConfig do
         7 => 95,
         8 => 75,
         9 => 78,
-        10 => 82,
+        10 => 88,
         11 => 78,
         12 => 25
       },

updates.md

@@ -99,11 +99,102 @@ The -150 to -75 range (the bulk of HRRR gradients) shows remarkably flat distanc
 
 All atmospheric correlations in this analysis are effectively summer-only findings. Seasonal scoring curves cannot be validated from this dataset. The physics-based seasonal tables remain the best available approach for non-summer months.
 
+## Finding 8: Commercial Link Diurnal Patterns (12,068 samples, 7 links)
+
+Data from March 30 - April 7, 2026. All links in Princeton TX area, LOS paths.
+
+### 68 GHz (af60, 2.82 km) — Strongest diurnal signal
+
+| UTC Hour | Avg RX (dBm) | Std Dev | Notes |
+|----------|-------------|---------|-------|
+| 09-10 | **-50.7** | 0.80 | Morning best |
+| 00-08 | -51.2 to -52.2 | 0.6-1.2 | Night — gradual improvement |
+| 11-12 | -52.1 to -52.3 | 4-5 | Weather events (high std) |
+| 13-14 | **-54.3 to -54.6** | 5-6 | Afternoon worst |
+| 15-23 | -52.8 to -53.4 | 0.8-3.7 | Evening recovery |
+
+**3.9 dB diurnal swing.** Afternoon degradation at 68 GHz is dominated by gaseous absorption increase with daytime heating and humidity rise, plus convective scintillation. The high stddev at 11-15 UTC indicates individual weather events (rain, convection) can cause 10+ dB excursions.
+
+### 11 GHz (core-new-hope, 5.66 km) — Inverted pattern
+
+| UTC Hour | Avg RX (dBm) | Std Dev | Notes |
+|----------|-------------|---------|-------|
+| 08-12 | **-49.7 to -49.8** | 2.1-2.5 | Morning best, but variable |
+| 01-07 | -49.8 to -50.1 | 1.7-2.0 | Night — slightly better mean, multipath |
+| 14-21 | -50.8 to -51.4 | 1.2-1.5 | Afternoon/evening — worse mean, less variable |
+
+**1.7 dB diurnal swing.** At 11 GHz, gaseous absorption is negligible. The diurnal pattern is driven by **multipath fading** from nocturnal inversions (higher stddev at night: 2.0-2.5 vs afternoon: 1.2-1.4). Mean signal is slightly better at night despite the multipath, because the refractivity enhancement aids the direct path.
+
+### 24 GHz (climax-to-core, 4.36 km) — Remarkably stable
+
+**Only 0.9 dB diurnal swing** (-63.4 to -64.3). The 22.235 GHz H₂O line creates a constant absorption floor that varies minimally with diurnal moisture changes on short paths. One anomaly at 12 UTC (stddev 2.51) is likely a single rain event.
+
+### Key insight: Diurnal sensitivity by frequency on LOS paths
+- 11 GHz: 1.7 dB (multipath-dominated)
+- 24 GHz: 0.9 dB (absorption floor, stable)
+- 68 GHz: 3.9 dB (gaseous absorption-dominated)
+
+This is **non-monotonic** — 24 GHz is more stable than 11 GHz. The H₂O absorption at 24 GHz is a constant penalty that doesn't fluctuate much diurnally, while 11 GHz refractivity effects create multipath that varies with inversion development.
+
+## Finding 9: RAOB Soundings Resolve Ducting HRRR Cannot (10,935 soundings)
+
+### RAOB vs HRRR gradient distributions
+
+| Statistic | RAOB | HRRR (at contacts) |
+|-----------|------|-------------------|
+| Average | -265.4 N/km | -107.2 N/km |
+| P10 | -521.7 | ~-200 |
+| P50 (median) | -199.5 | ~-75 |
+| P90 | -83.3 | ~-50 |
+| Min | -1969.9 | — |
+
+RAOB gradients are **2.5x stronger** than HRRR because soundings at ~10m vertical resolution capture thin surface ducts (50-100m) that HRRR's 25 hPa pressure level spacing (~250m) cannot resolve.
+
+### Sharp ducting threshold at -200 to -300 N/km
+
+| Gradient Bin | n | Ducting % | Avg PWAT (mm) |
+|-------------|---|-----------|---------------|
+| < -500 | 622 | **100%** | 26.6 |
+| -500 to -300 | 967 | **99.9%** | 26.9 |
+| -300 to -200 | 1,168 | **91.2%** | 29.9 |
+| -200 to -150 | 879 | 31.2% | 28.5 |
+| -150 to -100 | 1,061 | 1.8% | 28.7 |
+| -100 to -50 | 683 | 2.2% | 27.0 |
+| >= -50 | 159 | 1.9% | 20.0 |
+
+The transition from "always ducting" to "rarely ducting" occurs between -200 and -300 N/km. This is consistent with the theoretical ducting threshold of -157 N/km (where k goes negative), with the additional gradient needed to sustain a duct layer of practical depth.
+
+### Seasonal ducting rates from RAOB
+
+| Month | Total | Ducting % | Avg Gradient | Avg PWAT (mm) |
+|-------|-------|-----------|-------------|---------------|
+| Jun | 232 | **70.3%** | -339.9 | 28.5 |
+| Jul | 247 | **71.3%** | -308.0 | 28.1 |
+| Oct | 99 | **65.7%** | **-350.0** | 18.9 |
+| Sep | 2,073 | 56.0% | -282.3 | 26.3 |
+| Aug | 2,281 | 53.7% | -258.2 | 34.0 |
+| Nov | 99 | 49.5% | -248.9 | 10.8 |
+| May | 128 | 41.4% | -227.4 | 20.4 |
+| Apr | 86 | 32.6% | -155.1 | 13.7 |
+| Feb | 84 | 20.2% | -135.4 | 6.8 |
+| Jan | 61 | 14.8% | -126.3 | 8.9 |
+| Mar | 54 | 14.8% | -140.9 | 10.5 |
+| Dec | 100 | 12.0% | -138.7 | 8.0 |
+
+**October is the most underrated month.** It has the strongest average gradients of any month (-350 N/km), 65.7% ducting rate (3rd highest), but was scored only 82/100 in the 10 GHz seasonal table.
+
+**February was scored below January despite higher ducting.** Feb: 20.2% ducting, was scored 32. Jan: 14.8% ducting, scored 38.
+
+**Code changes:**
+- 10 GHz February seasonal score: 32 → 40 (now exceeds January, matching its higher ducting rate)
+- 10 GHz October seasonal score: 82 → 88 (3rd highest ducting, strongest gradients)
+
 ## Updated Dataset Statistics
 
 - **Contacts:** 58,367 total, 58,367 tropo (<3000 km)
 - **HRRR profiles:** 41,071,398 (37,925 matched to contacts, 65% coverage)
 - **Terrain profiles:** 58,361 (56,735 BLOCKED 97.1%, 1,284 CLEAR 2.2%, 342 FRESNEL_PARTIAL 0.6%)
+- **Soundings:** 10,935 (5,539 with gradient data, 2,964 with ducting detected)
+- **Commercial samples:** 12,068 (7 links, March 30 - April 7, 2026)
 - **IEMRE observations:** 15,092
 - **Weather stations:** 3,268
-- **Commercial samples:** 0 locally (available on production only)