test: add Mox deps + SnmpClient poll/3 tests, RoadProximity, Buildings.Loader

- Add {:mox, "~> 1.0", only: :test} to deps
- Commercial.SnmpClient: injectable snmp_runner for poll/3 tests (62→92%)
- Rover.RoadProximity: pure road_distances/3 tests with injectable fetch (37→82%)
- Buildings.Loader: ensure_loaded/1 with real tile file + quadkey tests (69→100%)
- Buildings.Index: records_near, mark_loaded, loaded? tests (72→96%)

Coverage: 77.96% → 78.93% (+0.97%)

2026-05-07 13:36:22 -05:00

21 KiB

Raw Blame History

Recalibration Analysis Report — 2026-05-04

Auto-generated by scripts/recalibrate_algo.py. This report is the input for algo.md updates — diff against the prior dated section to see what's actually moved.

Connection: postgres://prop:***@10.0.15.30:5432/prop Statement timeout: 20min

Row counts and date ranges

tbl	n	lo	hi
contacts	82084.000	1991-05-04	2026-05-02
hrrr_native_profiles	11472.000	2019-03-09	2025-09-21
hrrr_profiles	81302322.000	2016-09-17	2026-05-02
iemre_observations	28936.000
narr_profiles	395.000	1991-05-04	2014-09-19
nexrad_observations	7286.000	2019-08-17	2024-09-22
propagation_scores	nan
soundings	27745.000	1990-06-07	2026-05-04
surface_observations	648809.000	1991-05-03	2026-05-04
terrain_profiles	82084.000

Data gaps and ingestion health

Severity legend: BROKEN = table is empty and feeds an active scoring path; WARN = populated but coverage is low enough to skew per-band statistics; OK = within spec or informational. Each non-OK row carries a one-line remediation hint.

metric	value	severity	remediation
narr_rows	395	OK
narr_pre2014_rows	395	OK
hrrr_climatology_rows	0	BROKEN	Should self-heal within 24h via the daily 02:30Z cron that runs AdminTaskWorker(task=climatology). If still 0 after a day, check Oban dashboard for failed admin jobs.
hrrr_complete_contacts	81278	OK
hrrr_pending_contacts	806	OK
pre2014_contacts	207	OK

HRRR enrichment completeness: 99.0% (81,278 complete / 806 pending).

NARR coverage by year

Year-bucketed counts. NARR is supposed to span 1979 → 2014-10 at 3-hourly cadence (~2,920 timestamps per year per grid cell). Years far below the others are partial backfills the worker hasn't finished.

year	profiles
1991.000	31.000
1992.000	4.000
1993.000	18.000
1994.000	19.000
1996.000	3.000
1997.000	8.000
1998.000	2.000
1999.000	11.000
2000.000	10.000
2001.000	14.000
2002.000	12.000
2003.000	36.000
2004.000	23.000
2005.000	49.000
2006.000	20.000
2007.000	44.000
2008.000	15.000
2009.000	17.000
2010.000	31.000
2011.000	4.000
2012.000	15.000
2013.000	3.000
2014.000	6.000

HRRR enrichment status by year

Locate the pending backlog: a year with high queued/failed/unset counts is where HrrrFetchWorker needs to be rerun. Recent years dominating queued usually means the live Oban queue is stuck rather than a historical backfill gap.

year	complete	queued
2014.000	0.000	0.000
2015.000	0.000	0.000
2016.000	2.000	0.000
2017.000	26.000	0.000
2018.000	2658.000	0.000
2019.000	13036.000	0.000
2020.000	10300.000	0.000
2021.000	11266.000	0.000
2022.000	14487.000	2.000
2023.000	13461.000	5.000
2024.000	13378.000	0.000
2025.000	2640.000	0.000
2026.000	24.000	0.000

Contacts by band (≥50 MHz)

band_mhz	contacts	avg_km	max_km
222	7595	229.000	2306.000
432	9177	200.000	1752.000
902	1794	171.000	915.000
1296	2996	160.000	1752.000
2304	749	153.000	655.000
3400	351	151.000	643.000
5760	336	126.000	505.000
10000	54280	211.000	2393.000
24000	3815	96.000	710.000
47000	774	65.000	343.000
75000	117	57.000	289.000
122000	70	18.000	139.000
134000	5	65.000	157.000
142000	4	47.000	79.700
145000	2	40.000	79.600
241000	14	31.000	114.400
288000	1	1.000	1.246
322000	1	1.000	1.400
403000	1	1.000	1.400
411000	1	0.000	0.050

Monthly sounding ducting

month	soundings	ducting_pct	avg_min_grad	avg_pwat
1	103	27.200	-179.200	10.400
2	131	29.000	-163.100	7.900
3	93	17.200	-157.200	8.900
4	330	24.200	-155.500	11.300
5	360	48.100	-269.200	23.900
6	396	68.900	-352.900	31.600
7	304	67.100	-292.500	29.600
8	5045	55.200	-255.800	35.100
9	2345	56.300	-280.000	27.300
10	179	55.300	-304.800	18.300
11	194	45.400	-237.300	11.600
12	158	25.900	-182.400	10.800

HRRR native-profile duct distribution (best supportable band)

band_bin	profiles	avg_inv_top_m	avg_theta_e_jump	avg_richardson
15-30 GHz	10	2091.000	5.900	8.690
30-75 GHz	4	3860.000	5.000	12.120
5-15 GHz	98	4408.000	25.700	12.550
<5 GHz	1432	2097.000	5.400	18.420
none	9928	11497.000	80.500	38.270

Contact ↔ HRRR per-band Pearson correlations (matched n=81547)

band_mhz	n	rho_pr	rho_dpc	rho_pwat	rho_sref	rho_grad	rho_tc	rho_hpbl
222.000	7518.000	-0.014	0.077	0.068	0.087	0.040	-0.081	-0.039
432.000	9073.000	0.074	0.134	0.111	0.132	0.008	-0.088	-0.063
902.000	1754.000	0.080	0.097	0.050	0.090	-0.039	-0.070	-0.029
1296.000	2951.000	0.025	0.103	0.053	0.110	-0.077	-0.041	-0.051
2304.000	719.000	0.079	0.165	0.130	0.192	-0.126	-0.059	-0.086
3400.000	335.000	-0.026	0.154	0.150	0.221	0.061	-0.061	-0.024
5760.000	320.000	0.156	0.076	0.099	0.091	0.003	-0.068	-0.057
10000.000	54150.000	-0.086	-0.022	-0.029	0.000	0.030	-0.004	0.007
24000.000	3794.000	-0.221	-0.338	-0.362	-0.241	0.023	-0.211	-0.009
47000.000	757.000	-0.256	-0.215	-0.233	-0.054	-0.021	-0.113	-0.079
75000.000	109.000	-0.768	-0.699	-0.620	-0.582	0.434	-0.680	0.365
122000.000	63.000	-0.041	-0.129	-0.127	-0.248	0.355	0.039	0.400

Per-band HPBL bin distance distribution

One table per band with ≥50 matched contacts. A band-specific effect here means HPBL belongs in the scorer for that band.

222 MHz:

bin	n	avg_km	p50_km
<200	1224	266.100	205.500
200-500	1556	225.600	167.000
500-1000	1754	213.300	167.500
1000-1500	1426	223.700	170.000
1500-2000	1239	224.700	178.000
>=2000	319	233.800	158.000

432 MHz:

bin	n	avg_km	p50_km
<200	1579	224.500	178.000
200-500	1867	205.800	153.000
500-1000	2114	189.900	139.800
1000-1500	1683	196.800	148.000
1500-2000	1475	192.600	145.000
>=2000	355	171.700	124.000

902 MHz:

bin	n	avg_km	p50_km
<200	271	198.300	162.000
200-500	366	164.900	138.000
500-1000	433	166.500	131.600
1000-1500	356	166.800	136.500
1500-2000	282	174.900	152.000
>=2000	46	153.500	126.500

1.296 GHz:

bin	n	avg_km	p50_km
<200	443	174.500	145.000
200-500	612	173.100	133.500
500-1000	735	152.500	116.000
1000-1500	585	158.200	131.000
1500-2000	479	152.600	122.000
>=2000	97	143.800	108.000

2.304 GHz:

bin	n	avg_km	p50_km
<200	101	167.000	151.000
200-500	131	182.200	157.000
500-1000	188	140.800	122.000
1000-1500	159	146.700	135.000
1500-2000	123	148.300	135.000
>=2000	17	137.400	140.000

3.4 GHz:

bin	n	avg_km	p50_km
<200	54	143.100	122.000
200-500	60	178.200	141.000
500-1000	74	145.600	108.000
1000-1500	84	149.400	133.000
1500-2000	48	165.100	138.700
>=2000	15	121.800	128.000

5.76 GHz:

bin	n	avg_km	p50_km
<200	37	148.800	108.000
200-500	54	138.400	135.000
500-1000	85	113.300	108.000
1000-1500	74	130.900	127.500
1500-2000	54	121.200	122.000
>=2000	16	124.200	125.000

10 GHz:

bin	n	avg_km	p50_km
<200	10466	212.500	182.600
200-500	14330	205.900	179.500
500-1000	16463	208.700	190.300
1000-1500	8712	212.400	198.500
1500-2000	2985	201.200	179.000
>=2000	1191	228.700	193.500

24 GHz:

bin	n	avg_km	p50_km
<200	731	100.200	92.200
200-500	1079	91.500	87.700
500-1000	1179	94.800	87.700
1000-1500	534	98.500	95.700
1500-2000	196	99.900	99.400
>=2000	75	74.300	31.800

47 GHz:

bin	n	avg_km	p50_km
<200	172	67.100	63.600
200-500	237	62.600	61.300
500-1000	253	61.900	65.400
1000-1500	58	71.000	70.900
1500-2000	29	33.500	23.700
>=2000	8	33.900	32.300

75 GHz:

bin	n	avg_km	p50_km
<200	19	18.600	19.600
200-500	31	44.300	36.700
500-1000	53	58.800	36.700
1000-1500	6	56.100	29.800

122 GHz:

bin	n	avg_km	p50_km
<200	7	4.400	5.000
200-500	27	8.800	8.000
500-1000	18	13.600	7.000
1000-1500	6	78.500	85.500
1500-2000	3	10.000	12.000
>=2000	2	5.000	5.000

Per-band pressure bin distance distribution

Surface pressure tends to proxy synoptic-scale stagnation, so we want to see longer distances under the 1015-1025 mb ridge at every band that ducts.

222 MHz:

bin	n	avg_km	p50_km
<990	2720	227.200	178.000
990-1000	1926	255.100	201.000
1000-1010	1799	233.800	186.000
1010-1020	953	181.400	131.000
>=1020	120	166.100	132.000

432 MHz:

bin	n	avg_km	p50_km
<990	3536	188.800	141.000
990-1000	2278	226.700	178.000
1000-1010	1832	210.900	156.500
1010-1020	1264	174.100	128.000
>=1020	163	159.000	121.000

902 MHz:

bin	n	avg_km	p50_km
<990	590	154.100	127.000
990-1000	510	203.200	157.500
1000-1010	369	189.200	148.000
1010-1020	249	139.200	143.000
>=1020	36	79.700	51.500

1.296 GHz:

bin	n	avg_km	p50_km
<990	1038	155.400	125.000
990-1000	800	183.100	146.000
1000-1010	620	165.400	117.000
1010-1020	444	132.400	114.000
>=1020	49	119.800	101.000

2.304 GHz:

bin	n	avg_km	p50_km
<990	190	146.500	134.000
990-1000	250	155.000	122.000
1000-1010	144	172.600	134.000
1010-1020	121	147.600	152.000
>=1020	14	127.600	137.000

3.4 GHz:

bin	n	avg_km	p50_km
<990	95	154.400	136.000
990-1000	159	159.200	134.000
1000-1010	45	166.000	140.000
1010-1020	27	111.700	98.000
>=1020	9	114.600	131.000

5.76 GHz:

bin	n	avg_km	p50_km
<990	86	106.000	105.500
990-1000	139	133.200	111.000
1000-1010	51	136.100	104.000
1010-1020	34	143.200	152.000
>=1020	10	140.600	139.000

10 GHz:

bin	n	avg_km	p50_km
<990	27142	221.800	213.500
990-1000	13469	181.200	158.600
1000-1010	7236	204.000	175.200
1010-1020	4658	218.700	175.300
>=1020	1645	229.300	207.600

24 GHz:

bin	n	avg_km	p50_km
<990	1485	100.900	96.500
990-1000	1462	87.200	81.000
1000-1010	463	114.400	113.400
1010-1020	258	70.900	72.100
>=1020	126	102.700	89.600

47 GHz:

bin	n	avg_km	p50_km
<990	279	68.500	70.900
990-1000	297	58.000	57.300
1000-1010	63	69.600	68.000
1010-1020	89	60.600	65.400
>=1020	29	43.400	42.000

75 GHz:

bin	n	avg_km	p50_km
<990	55	65.300	57.300
990-1000	32	17.500	10.400
1000-1010	12	71.800	89.800
1010-1020	8	18.300	19.600
>=1020	2	11.700	11.700

122 GHz:

bin	n	avg_km	p50_km
<990	22	12.000	7.000
990-1000	19	27.300	7.000
1000-1010	10	8.400	6.000
1010-1020	10	13.400	8.000
>=1020	2	12.000	12.000

Contact ↔ NARR per-band Pearson correlations (pre-2014 only, matched n=201)

Historical sanity check: if the HRRR-era signs and magnitudes survive on 30+ years of NARR reanalysis, the scoring factors are physical rather than HRRR-artefact. Diverging signs = flag it.

band_mhz	n	rho_pr	rho_dpc	rho_pwat	rho_sref	rho_grad	rho_tc	rho_hpbl
10000.000	103.000	-0.402	-0.402	-0.402	-0.234	-0.222	-0.402	-0.402

NEXRAD composite reflectivity vs distance (matched n=20525)

Rain-attenuation sanity check — one table per band with ≥50 matched contacts. At rain-sensitive bands (24+ GHz) higher max_dbz should map to shorter contacts; at 10 GHz and below the effect should be barely detectable.

10 GHz:

bin	n	avg_km	p50_km
none (<5)	13803	207.900	193.500
drizzle (5-20)	1270	201.800	200.400
light (20-30)	1696	209.600	187.900
moderate (30-45)	1594	229.500	213.900
heavy (>=45)	675	191.600	146.300

24 GHz:

bin	n	avg_km	p50_km
none (<5)	868	91.500	83.700
drizzle (5-20)	74	108.300	123.200
light (20-30)	151	88.200	70.900
moderate (30-45)	76	109.300	122.400
heavy (>=45)	49	86.600	102.300

47 GHz:

bin	n	avg_km	p50_km
none (<5)	186	52.400	56.000
light (20-30)	30	59.300	66.300
moderate (30-45)	9	75.100	81.600
heavy (>=45)	15	71.400	82.800

hrrr_native_profiles.best_duct_band_ghz vs distance (matched n=56845)

Validates the 1.15× boost in Scorer.score_refractivity/4: contacts where the native duct supports the target band should run longer than those where it does not. One table per band with ≥50 matched contacts.

10 GHz:

bin	n	avg_km	p50_km
none	44655	211.200	192.400
<5 GHz	7073	200.000	169.300
5-10 GHz	439	199.000	174.200
10-24 GHz	146	204.700	242.700
24-47 GHz	5	176.600	133.300
>=47 GHz	23	147.800	131.500

24 GHz:

bin	n	avg_km	p50_km
none	3183.000	94.800	92.200
<5 GHz	473.000	98.600	84.900
5-10 GHz	36.000	97.500	113.700
10-24 GHz	2.000	44.700	44.700
>=47 GHz	6.000	131.500	131.500

47 GHz:

bin	n	avg_km	p50_km
none	552.000	64.300	65.400
<5 GHz	131.000	56.800	61.800
5-10 GHz	6.000	98.100	113.700

75 GHz:

bin	n	avg_km	p50_km
none	77.000	54.200	36.700
<5 GHz	15.000	13.900	14.600

Commercial-link rx_power degradation vs contemporaneous DFW contacts (matched n=12)

Pearson(degradation_db, distance_km) = 0.318

bin	n	avg_km	p50_km
negative/0 (no fade)	10	134.500	122.500
3 dB noise floor	2	119.500	119.500

21 KiB Raw Blame History Unescape Escape

Recalibration Analysis Report — 2026-05-04

Row counts and date ranges

Data gaps and ingestion health

NARR coverage by year

HRRR enrichment status by year

Contacts by band (≥50 MHz)

Monthly sounding ducting

HRRR native-profile duct distribution (best supportable band)

Contact ↔ HRRR per-band Pearson correlations (matched n=81547)

Per-band HPBL bin distance distribution

Per-band pressure bin distance distribution

Contact ↔ NARR per-band Pearson correlations (pre-2014 only, matched n=201)

NEXRAD composite reflectivity vs distance (matched n=20525)

hrrr_native_profiles.best_duct_band_ghz vs distance (matched n=56845)

Commercial-link rx_power degradation vs contemporaneous DFW contacts (matched n=12)

21 KiB

Raw Blame History