Rain intensity gauge accuracy rankings explained based datasheets: tipping buckets, siphons, and weighing gauges

/

Required-depth analysis for rain-gauge short-window intensity

A transparent analytical-floor comparison of 52 rain-gauge intensity-relevant rows from Barani, OTT HydroMet, Lambrecht, KISTERS / HyQuest, Geolux, Vaisala, Texas/Campbell, Casella, Davis, Onset HOBO, Met One, RainWise, MicroStep / Meteoservis, AEM/FTS, CAE, Pronamic, Seven Sensor Solutions, and EML — including the Lambrecht rain[e] hi-resolution state-model rows. Each row is ranked by the minimum rainfall depth needed to reach a stated relative uncertainty target on a short reporting window during active rain. The full input parameters, formulas, and worked sample calculations are in the appendix so any reader can verify the numbers independently. Companion accumulation totals ranking is published separately.

Conflict-of-interest disclosure (read first). This analysis is prepared by Jan Barani, founder and CEO of Barani Design Technologies, manufacturer of the MeteoRain product family that appears in the comparison set. To mitigate bias: (i) every input value is drawn from publicly available manufacturer datasheets and operating manuals — no proprietary or vendor-supplied data — so any reader can re-evaluate the framework against the same inputs; (ii) conservative modelling parameters are applied uniformly across all products (U95 = 2σ_total, γ_s = 1 for siphon TBRs, B3/B3 only for IoT³ phase-aware rows when the reporting window is interior to active rain — short-window intensity is exactly that case); (iii) no vendor of any compared product has funded, reviewed, or approved this analysis. The full formulas and per-category sample calculations are in the appendix below; readers who suspect bias are encouraged to substitute alternative parameter values and verify the results.

What this ranking is and is not

What it shows

An intrinsic analytical floor from the instrument architecture and transmitted output state for short-window intensity readings during active rain (1-min, 10-min, rolling 30-min, and longer windows interior to an event). Useful for short-listing products that can reach a stated accuracy target under best-case conditions when the reporting window sits inside a rain event.

What it does not show

Field accuracy. Wind under-catch, evaporative loss, debris, shielding, calibration drift, siting effects, spatial representativeness, response time, filter bandwidth, output delay, durability, cost, power, and maintenance burden are not included. Field-measured short-window intensity at any given site will always be worse than the analytical floor — sometimes substantially worse for native real-time intensity products that depend on weighing-gauge dynamic response.

Read this before reading the table. The rankings below come from a deterministic mathematical model (analytical floor) using only public datasheet inputs. The same model would produce the same numbers if any other party ran it. Within any product class, the row that wins is the one with the smallest combined boundary-state and per-tip stochastic uncertainty. The Lambrecht hi-resolution state-model rows sit at the top of the table because they have the smallest mathematical state resolution (0.001 mm), but they are model-potential layers — defensible real-time intensity claims require measured weighing response time, filter bandwidth, and self-emptying-correction interval, which the public datasheet does not constrain. The defensible Lambrecht row at this horizon is the public envelope further down. Field performance is a separate question not addressed here.

Why intensity needs its own ranking

Short-window intensity is a live measurement problem. The reporting window — a 1-minute, 10-minute, or rolling 30-minute interval — usually sits inside an active rain event, with both window boundaries bracketed by observed tips. For a tipping bucket that transmits a B-bit phase code on each tip (the MeteoRain IoT³ family transmits B = 3, resolving the inter-tip residual to 1/8 of a tip), this means the boundary residual at both ends of the window is resolved to the phase-code resolution rather than the full tip depth. This is the B3/B3 interior bound in the model, and it is dramatically smaller than the wired equivalent of the same hardware. For accumulation totals over a daily window — where the rain event start and end sit in dry conditions with no adjacent tips to anchor the residual — the phase code does not help (see the companion accumulation ranking).

30 minutes is genuinely dual-use. A rolling 30-minute intensity reading during active rain belongs in this intensity ranking (B3/B3, both ends bracketed by tips). A fixed 4:00–4:30 PM accumulation slot — for example a typical Florida summer afternoon thunderstorm — belongs in the accumulation ranking (B0/B0, event start/end unresolved). The two rankings give different positions for the IoT³ family even at the same nominal 30-minute horizon.

Overall ranking — all 52 rows in one table

Sorted by required depth at 1%. Lower is better. The category badge in the product column shows the row family. Lambrecht hi-resolution state-model rows are included with yellow shading to signal "model only — requires response gating before being treated as a real-time claim." Ties marked with a "T" suffix on the rank.

Overall
rank		 Gauge / output product P_min
at 1% (mm)		 P_min
at 2% (mm)		 1%
@1 min (mm/h)		 1%
@10 min (mm/h)		 1%
@30 min (mm/h)		 1%
@1 h (mm/h)		 1%
@3 h (mm/h)		 Δ_res
(mm)		 Evidence
layer		 State
detail		 A
(cm²)		 d / d_eq
(mm)		 V_s
(mL)		 Why it ranks here
1 Model only Lambrecht rain[e]400 — 0.001 mm hi-res scale-state model 0.635 0.166 38.13 3.81 1.27 0.635 0.212 0.001 rain[e] hi-res state model scale state: ≈6.64 bits 400 0.1 — Hi-resolution state-resolution model layer. Mathematical potential only; requires measured weighing response time, filter bandwidth, and self-emptying-correction interval before being treated as a real-time intensity claim.
2 Model only Lambrecht rain[e]314 — 0.001 mm hi-res scale-state model 0.804 0.207 48.27 4.83 1.61 0.804 0.268 0.001 rain[e] hi-res state model scale state: ≈6.99 bits 314 0.1274 — Hi-resolution state-resolution model layer. Mathematical potential only; requires measured weighing response time, filter bandwidth, and self-emptying-correction interval before being treated as a real-time intensity claim.
3 IoT³ MeteoRain IoT³ 533 Classic (gen3) 1.03 0.446 62.10 6.21 2.07 1.03 0.345 0.0094 / 0.075 IoT³ TBR hybrid product short windows: 3-bit state; daily/event totals: raw boundary 533 0.075 — IoT³ phase-aware: B=3 transmitted phase code resolves boundary residuals to 1/8 of a tip during active rain, dramatically reducing the boundary term for short-window intensity.
4 Model only Lambrecht rain[e]400 — 0.01 mm hi-res pulse-state model 1.19 0.494 71.22 7.12 2.37 1.19 0.396 0.01 rain[e] hi-res state model pulse state: ≈3.32 bits 400 0.1 — Hi-resolution state-resolution model layer. Mathematical potential only; requires measured weighing response time, filter bandwidth, and self-emptying-correction interval before being treated as a real-time intensity claim.
5 Model only Lambrecht rain[e] / rain[e]LP / rain[e]H3 — 0.001 mm hi-res scale-state model 1.25 0.318 75.30 7.53 2.51 1.25 0.418 0.001 rain[e] hi-res state model scale state: ≈7.64 bits 200 0.2 — Hi-resolution state-resolution model layer. Mathematical potential only; requires measured weighing response time, filter bandwidth, and self-emptying-correction interval before being treated as a real-time intensity claim.
6 Model only Lambrecht rain[e]314 — 0.01 mm hi-res pulse-state model 1.31 0.52 78.36 7.84 2.61 1.31 0.435 0.01 rain[e] hi-res state model pulse state: ≈3.67 bits 314 0.1274 — Hi-resolution state-resolution model layer. Mathematical potential only; requires measured weighing response time, filter bandwidth, and self-emptying-correction interval before being treated as a real-time intensity claim.
7 IoT³ MeteoRain IoT³ 400 Aero (gen3) 1.38 0.594 82.80 8.28 2.76 1.38 0.46 0.0125 / 0.100 IoT³ TBR hybrid product short windows: 3-bit state; daily/event totals: raw boundary 400 0.1 — IoT³ phase-aware: B=3 transmitted phase code resolves boundary residuals to 1/8 of a tip during active rain, dramatically reducing the boundary term for short-window intensity.
8 Model only Lambrecht rain[e] / rain[e]LP / rain[e]H3 — 0.01 mm hi-res pulse-state model 1.65 0.593 99.18 9.92 3.31 1.65 0.551 0.01 rain[e] hi-res state model pulse state: ≈4.32 bits 200 0.2 — Hi-resolution state-resolution model layer. Mathematical potential only; requires measured weighing response time, filter bandwidth, and self-emptying-correction interval before being treated as a real-time intensity claim.
9 IoT³ MeteoRain IoT³ 200 Pro 0.2 mm (gen3) 2.76 1.19 165.60 16.56 5.52 2.76 0.92 0.0250 / 0.200 IoT³ TBR hybrid product short windows: 3-bit state; daily/event totals: raw boundary 200 0.2 — IoT³ phase-aware: B=3 transmitted phase code resolves boundary residuals to 1/8 of a tip during active rain, dramatically reducing the boundary term for short-window intensity.
10 IoT³ MeteoRain IoT³ 200 Pro 0.254 mm / 0.01 in (gen3) 3.13 1.43 187.86 18.79 6.26 3.13 1.04 0.0318 / 0.254 IoT³ TBR hybrid product short windows: 3-bit state; daily/event totals: raw boundary 200 0.254 — IoT³ phase-aware: B=3 transmitted phase code resolves boundary residuals to 1/8 of a tip during active rain, dramatically reducing the boundary term for short-window intensity.
11 TBR MeteoRain 533 Classic 6.36 3.12 381.78 38.18 12.73 6.36 2.12 0.075 pulse-output TBR raw pulse 533 0.075 — Wired pulse output (B=0); no phase-code advantage. For short-window intensity, ranked by tip depth and collector area only.
12 TBR Texas/Campbell TR-525M / TE525MM 245 mm 0.1 mm 8.39 4.14 503.58 50.36 16.79 8.39 2.80 0.1 pulse-output TBR raw pulse 471.4 0.1 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
13T TBR CAE PG4i 400 cm² 0.1 mm 8.48 4.16 508.98 50.90 16.97 8.48 2.83 0.1 pulse-output TBR raw pulse 400 0.1 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
13T TBR Casella TBRG 400 cm² 0.1 mm 8.48 4.16 508.98 50.90 16.97 8.48 2.83 0.1 pulse-output TBR raw pulse 400 0.1 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
13T TBR MeteoRain 400 Aero 8.48 4.16 508.98 50.90 16.97 8.48 2.83 0.1 pulse-output TBR raw pulse 400 0.1 — Wired pulse output (B=0); no phase-code advantage. For short-window intensity, ranked by tip depth and collector area only.
16 TBR Texas TR-525-W2 200 mm 0.1 mm 8.69 4.21 521.22 52.12 17.37 8.69 2.90 0.1 pulse-output TBR raw pulse 314.2 0.1 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
17T TBR Lambrecht 15189 0.1 mm 9.51 4.41 570.60 57.06 19.02 9.51 3.17 0.1 pulse-output TBR raw pulse 200 0.1 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
17T TBR MicroStep/Meteoservis MR2 200 cm² 0.1 mm option 9.51 4.41 570.60 57.06 19.02 9.51 3.17 0.1 pulse-output TBR raw pulse 200 0.1 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
19T Weighing Lambrecht rain[e] / rain[e]LP / rain[e]314 / rain[e]400 / rain[e]H3 public amount-total 10.00 5.00 600.00 60.00 20.00 10.00 3.33 — weighing public envelope public fixed floor — — — Weighing public envelope: ±0.1 mm or ±1% amount accuracy floor. The native real-time intensity product (Pluvio² S Intensity RT, rain[e] native 1-min) has a separate ±6 mm/h floor — see "Where weighing fits" below.
19T Weighing OTT Pluvio² L Accu NRT / Accu total NRT 10.00 5.00 600.00 60.00 20.00 10.00 3.33 — weighing public envelope NRT amount product — — — Weighing public envelope: ±0.1 mm or ±1% amount accuracy floor. The native real-time intensity product (Pluvio² S Intensity RT, rain[e] native 1-min) has a separate ±6 mm/h floor — see "Where weighing fits" below.
19T Weighing OTT Pluvio² S Accu NRT / Accu total NRT 10.00 5.00 600.00 60.00 20.00 10.00 3.33 — weighing public envelope NRT amount product — — — Weighing public envelope: ±0.1 mm or ±1% amount accuracy floor. The native real-time intensity product (Pluvio² S Intensity RT, rain[e] native 1-min) has a separate ±6 mm/h floor — see "Where weighing fits" below.
22 TBR Vaisala QMR102 / EML ARG100 / Campbell ARG100 500 cm² 0.2 mm 16.43 8.19 985.80 98.58 32.86 16.43 5.48 0.2 pulse-output TBR raw pulse 500 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
23 TBR Casella TBRG 400 cm² 0.2 mm 16.49 8.20 989.40 98.94 32.98 16.49 5.50 0.2 pulse-output TBR raw pulse 400 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
24T TBR RainWise wireless metric 8 in 0.2 mm 16.57 8.22 994.20 99.42 33.14 16.57 5.52 0.2 pulse-output TBR raw pulse 324.3 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
24T TBR Texas TR-525USW 8 in 0.2 mm 16.57 8.22 994.20 99.42 33.14 16.57 5.52 0.2 pulse-output TBR raw pulse 324.3 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
24T TBR Met One 370D 324 cm² 0.2 mm 16.57 8.22 994.20 99.42 33.14 16.57 5.52 0.2 pulse-output TBR raw pulse 324 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
27 TBR Texas TR-525-W2 200 mm 0.2 mm 16.59 8.23 995.40 99.54 33.18 16.59 5.53 0.2 pulse-output TBR raw pulse 314.2 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
28 TBR Davis 6466/6466M AeroCone 214 cm² 0.2 mm 16.88 8.30 1012.80 101.28 33.76 16.88 5.63 0.2 pulse-output TBR raw pulse 214 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
29T TBR Lambrecht 15189 0.2 mm 16.97 8.32 1018.20 101.82 33.94 16.97 5.66 0.2 pulse-output TBR raw pulse 200 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
29T TBR MeteoRain 200 Pro 0.2 mm 16.97 8.32 1018.20 101.82 33.94 16.97 5.66 0.2 pulse-output TBR raw pulse 200 0.2 — Wired pulse output (B=0); no phase-code advantage. For short-window intensity, ranked by tip depth and collector area only.
29T TBR MicroStep/Meteoservis MR2 200 cm² 0.2 mm 16.97 8.32 1018.20 101.82 33.94 16.97 5.66 0.2 pulse-output TBR raw pulse 200 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
29T TBR Seven Sensor Solutions 3S-RG 0.2 mm — no siphon 16.97 8.32 1018.20 101.82 33.94 16.97 5.66 0.2 pulse-output TBR raw pulse 200 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
29T TBR Vaisala QMR101 / Pronamic Professional 0.2 mm 16.97 8.32 1018.20 101.82 33.94 16.97 5.66 0.2 pulse-output TBR raw pulse 200 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
34 TBR Onset HOBO RG3-M 15.39 cm 0.2 mm 17.07 8.35 1024.20 102.42 34.14 17.07 5.69 0.2 pulse-output TBR raw pulse 186 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
35 Siphon KISTERS/HyQuest TB4 Series II 0.1 mm — siphon Vs=12 mL 17.73 8.83 1063.80 106.38 35.46 17.73 5.91 0.1 siphon TBR raw pulse + siphon 628.3 0.1 12 Bucket plus integrated siphon storage; γ_s = 1 conservative siphon residual. The siphon term is part of the short-window boundary uncertainty.
36 TBR EML ARG100 optional 0.25 mm 20.49 10.23 1229.40 122.94 40.98 20.49 6.83 0.25 pulse-output TBR raw pulse 500 0.25 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
37T TBR AEM/FTS 2408 8 in 0.01 in 20.93 10.42 1255.80 125.58 41.86 20.93 6.98 0.254 pulse-output TBR raw pulse 324.3 0.254 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
37T TBR RainWise Rainew 8 in 0.01 in 20.93 10.42 1255.80 125.58 41.86 20.93 6.98 0.254 pulse-output TBR raw pulse 324.3 0.254 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
37T TBR Texas/Campbell TR-525WS / TE525WS 8 in 0.01 in 20.93 10.42 1255.80 125.58 41.86 20.93 6.98 0.254 pulse-output TBR raw pulse 324.3 0.254 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
40 TBR Davis 6466/6466M AeroCone 214 cm² 0.01 in 21.17 10.48 1270.20 127.02 42.34 21.17 7.06 0.254 pulse-output TBR raw pulse 214 0.254 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
41 TBR MeteoRain 200 Pro 0.254 mm / 0.01 in 21.24 10.49 1274.40 127.44 42.48 21.24 7.08 0.254 pulse-output TBR raw pulse 200 0.254 — Wired pulse output (B=0); no phase-code advantage. For short-window intensity, ranked by tip depth and collector area only.
42 TBR Onset HOBO RG3 15.39 cm 0.01 in 21.32 10.51 1279.20 127.92 42.64 21.32 7.11 0.254 pulse-output TBR raw pulse 186 0.254 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
43 TBR Texas/Campbell TR-525 / TE525 6 in 0.01 in 21.34 10.52 1280.40 128.04 42.68 21.34 7.11 0.254 pulse-output TBR raw pulse 182.4 0.254 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
44 Siphon Geolux RG400 0.1 mm — siphon Vs≈16 mL, 4 mL/tip 33.98 16.91 2038.80 203.88 67.96 33.98 11.33 0.1 siphon TBR raw pulse + siphon 400 0.1 16 Bucket plus integrated siphon storage; γ_s = 1 conservative siphon residual. The siphon term is part of the short-window boundary uncertainty.
45 Siphon KISTERS/HyQuest TB4 Series II 0.2 mm — siphon Vs=12 mL 35.45 17.66 2127.00 212.70 70.90 35.45 11.82 0.2 siphon TBR raw pulse + siphon 314.2 0.2 12 Bucket plus integrated siphon storage; γ_s = 1 conservative siphon residual. The siphon term is part of the short-window boundary uncertainty.
46 Siphon Geolux RG200 0.2 mm — siphon Vs≈8 mL, 4 mL/tip 37.15 18.41 2229.00 222.90 74.30 37.15 12.38 0.2 siphon TBR raw pulse + siphon 200 0.2 8 Bucket plus integrated siphon storage; γ_s = 1 conservative siphon residual. The siphon term is part of the short-window boundary uncertainty.
47 Siphon KISTERS/HyQuest TB4 / Campbell CS700/CS700H 0.01 in — siphon Vs=12 mL 37.65 18.78 2259.00 225.90 75.30 37.65 12.55 0.254 siphon TBR raw pulse + siphon 314.2 0.254 12 Bucket plus integrated siphon storage; γ_s = 1 conservative siphon residual. The siphon term is part of the short-window boundary uncertainty.
48 TBR Casella TBRG 400 cm² 0.5 mm 40.89 20.43 2453.40 245.34 81.78 40.89 13.63 0.5 pulse-output TBR raw pulse 400 0.5 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
49 TBR MicroStep/Meteoservis MR2 200 cm² 0.5 mm option 41.08 20.48 2464.80 246.48 82.16 41.08 13.69 0.5 pulse-output TBR raw pulse 200 0.5 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
50T Weighing Lambrecht rain[e]one Modbus public amount-total — 5.00 — — — — — — weighing public envelope public fixed floor — — — Weighing public envelope: ±0.1 mm or ±1% amount accuracy floor. The native real-time intensity product (Pluvio² S Intensity RT, rain[e] native 1-min) has a separate ±6 mm/h floor — see "Where weighing fits" below.
50T Weighing OTT Pluvio² L Intensity RT — — — — — — — — weighing public envelope RT 1-min intensity — — — Weighing public envelope: ±0.1 mm or ±1% amount accuracy floor. The native real-time intensity product (Pluvio² S Intensity RT, rain[e] native 1-min) has a separate ±6 mm/h floor — see "Where weighing fits" below.
50T Weighing OTT Pluvio² S Intensity RT — — — — — — — — weighing public envelope RT 1-min intensity — — — Weighing public envelope: ±0.1 mm or ±1% amount accuracy floor. The native real-time intensity product (Pluvio² S Intensity RT, rain[e] native 1-min) has a separate ±6 mm/h floor — see "Where weighing fits" below.

Within-category rankings

The same 52 rows grouped by family, so within-class comparisons are immediate. Within-class ranks are computed within that family only.

Column legend (applies to overall and per-category tables)

Within each category, the table is sorted by required depth at 1% (lower is better). The within-class rank shows where each product sits within its own family. Δ_res is the unresolved leftover-water resolution (mm). A is collector area (cm²). d / d_eq is tip depth or equivalent self-emptying event depth (mm). V_s is siphon storage volume (mL). The 1% @ 1 min through @ 3 h columns convert the 1% required depth into average rainfall rate (mm/h) over that horizon — the rate a true rainfall would have to sustain over that window for the analytical floor to deliver 1% accuracy. Columns with the slightly darker grey background are technical audit data; the main accuracy comparison is in the leftmost numerical columns.

Tipping bucket gauges (raw pulse and IoT³ phase-aware) — 35 rows

For short-window intensity during active rain, the IoT³ rows use the B3/B3 interior bound (both window edges bracketed by transmitted phase-coded tips). The raw-pulse (wired B=0) rows use the same hardware floor as accumulation — no phase-code advantage. Within this category, IoT³ phase-aware rows on a given hardware (e.g. 533 Classic) sit dramatically below the wired equivalent for short-window intensity; for accumulation totals they tie (see the companion accumulation ranking).

In-class
rank		 Gauge / output product P_min
at 1% (mm)		 P_min
at 2% (mm)		 1%
@1 min (mm/h)		 1%
@10 min (mm/h)		 1%
@30 min (mm/h)		 1%
@1 h (mm/h)		 1%
@3 h (mm/h)		 Δ_res
(mm)		 Evidence
layer		 State
detail		 A
(cm²)		 d / d_eq
(mm)		 V_s
(mL)		 Why it ranks here
1 IoT³ MeteoRain IoT³ 533 Classic (gen3) 1.03 0.446 62.10 6.21 2.07 1.03 0.345 0.0094 / 0.075 IoT³ TBR hybrid product short windows: 3-bit state; daily/event totals: raw boundary 533 0.075 — IoT³ phase-aware: B=3 transmitted phase code resolves boundary residuals to 1/8 of a tip during active rain, dramatically reducing the boundary term for short-window intensity.
2 IoT³ MeteoRain IoT³ 400 Aero (gen3) 1.38 0.594 82.80 8.28 2.76 1.38 0.46 0.0125 / 0.100 IoT³ TBR hybrid product short windows: 3-bit state; daily/event totals: raw boundary 400 0.1 — IoT³ phase-aware: B=3 transmitted phase code resolves boundary residuals to 1/8 of a tip during active rain, dramatically reducing the boundary term for short-window intensity.
3 IoT³ MeteoRain IoT³ 200 Pro 0.2 mm (gen3) 2.76 1.19 165.60 16.56 5.52 2.76 0.92 0.0250 / 0.200 IoT³ TBR hybrid product short windows: 3-bit state; daily/event totals: raw boundary 200 0.2 — IoT³ phase-aware: B=3 transmitted phase code resolves boundary residuals to 1/8 of a tip during active rain, dramatically reducing the boundary term for short-window intensity.
4 IoT³ MeteoRain IoT³ 200 Pro 0.254 mm / 0.01 in (gen3) 3.13 1.43 187.86 18.79 6.26 3.13 1.04 0.0318 / 0.254 IoT³ TBR hybrid product short windows: 3-bit state; daily/event totals: raw boundary 200 0.254 — IoT³ phase-aware: B=3 transmitted phase code resolves boundary residuals to 1/8 of a tip during active rain, dramatically reducing the boundary term for short-window intensity.
5 TBR MeteoRain 533 Classic 6.36 3.12 381.78 38.18 12.73 6.36 2.12 0.075 pulse-output TBR raw pulse 533 0.075 — Wired pulse output (B=0); no phase-code advantage. For short-window intensity, ranked by tip depth and collector area only.
6 TBR Texas/Campbell TR-525M / TE525MM 245 mm 0.1 mm 8.39 4.14 503.58 50.36 16.79 8.39 2.80 0.1 pulse-output TBR raw pulse 471.4 0.1 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
7T TBR CAE PG4i 400 cm² 0.1 mm 8.48 4.16 508.98 50.90 16.97 8.48 2.83 0.1 pulse-output TBR raw pulse 400 0.1 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
7T TBR Casella TBRG 400 cm² 0.1 mm 8.48 4.16 508.98 50.90 16.97 8.48 2.83 0.1 pulse-output TBR raw pulse 400 0.1 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
7T TBR MeteoRain 400 Aero 8.48 4.16 508.98 50.90 16.97 8.48 2.83 0.1 pulse-output TBR raw pulse 400 0.1 — Wired pulse output (B=0); no phase-code advantage. For short-window intensity, ranked by tip depth and collector area only.
10 TBR Texas TR-525-W2 200 mm 0.1 mm 8.69 4.21 521.22 52.12 17.37 8.69 2.90 0.1 pulse-output TBR raw pulse 314.2 0.1 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
11T TBR Lambrecht 15189 0.1 mm 9.51 4.41 570.60 57.06 19.02 9.51 3.17 0.1 pulse-output TBR raw pulse 200 0.1 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
11T TBR MicroStep/Meteoservis MR2 200 cm² 0.1 mm option 9.51 4.41 570.60 57.06 19.02 9.51 3.17 0.1 pulse-output TBR raw pulse 200 0.1 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
13 TBR Vaisala QMR102 / EML ARG100 / Campbell ARG100 500 cm² 0.2 mm 16.43 8.19 985.80 98.58 32.86 16.43 5.48 0.2 pulse-output TBR raw pulse 500 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
14 TBR Casella TBRG 400 cm² 0.2 mm 16.49 8.20 989.40 98.94 32.98 16.49 5.50 0.2 pulse-output TBR raw pulse 400 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
15T TBR RainWise wireless metric 8 in 0.2 mm 16.57 8.22 994.20 99.42 33.14 16.57 5.52 0.2 pulse-output TBR raw pulse 324.3 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
15T TBR Texas TR-525USW 8 in 0.2 mm 16.57 8.22 994.20 99.42 33.14 16.57 5.52 0.2 pulse-output TBR raw pulse 324.3 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
15T TBR Met One 370D 324 cm² 0.2 mm 16.57 8.22 994.20 99.42 33.14 16.57 5.52 0.2 pulse-output TBR raw pulse 324 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
18 TBR Texas TR-525-W2 200 mm 0.2 mm 16.59 8.23 995.40 99.54 33.18 16.59 5.53 0.2 pulse-output TBR raw pulse 314.2 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
19 TBR Davis 6466/6466M AeroCone 214 cm² 0.2 mm 16.88 8.30 1012.80 101.28 33.76 16.88 5.63 0.2 pulse-output TBR raw pulse 214 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
20T TBR Lambrecht 15189 0.2 mm 16.97 8.32 1018.20 101.82 33.94 16.97 5.66 0.2 pulse-output TBR raw pulse 200 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
20T TBR MeteoRain 200 Pro 0.2 mm 16.97 8.32 1018.20 101.82 33.94 16.97 5.66 0.2 pulse-output TBR raw pulse 200 0.2 — Wired pulse output (B=0); no phase-code advantage. For short-window intensity, ranked by tip depth and collector area only.
20T TBR MicroStep/Meteoservis MR2 200 cm² 0.2 mm 16.97 8.32 1018.20 101.82 33.94 16.97 5.66 0.2 pulse-output TBR raw pulse 200 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
20T TBR Seven Sensor Solutions 3S-RG 0.2 mm — no siphon 16.97 8.32 1018.20 101.82 33.94 16.97 5.66 0.2 pulse-output TBR raw pulse 200 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
20T TBR Vaisala QMR101 / Pronamic Professional 0.2 mm 16.97 8.32 1018.20 101.82 33.94 16.97 5.66 0.2 pulse-output TBR raw pulse 200 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
25 TBR Onset HOBO RG3-M 15.39 cm 0.2 mm 17.07 8.35 1024.20 102.42 34.14 17.07 5.69 0.2 pulse-output TBR raw pulse 186 0.2 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
26 TBR EML ARG100 optional 0.25 mm 20.49 10.23 1229.40 122.94 40.98 20.49 6.83 0.25 pulse-output TBR raw pulse 500 0.25 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
27T TBR AEM/FTS 2408 8 in 0.01 in 20.93 10.42 1255.80 125.58 41.86 20.93 6.98 0.254 pulse-output TBR raw pulse 324.3 0.254 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
27T TBR RainWise Rainew 8 in 0.01 in 20.93 10.42 1255.80 125.58 41.86 20.93 6.98 0.254 pulse-output TBR raw pulse 324.3 0.254 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
27T TBR Texas/Campbell TR-525WS / TE525WS 8 in 0.01 in 20.93 10.42 1255.80 125.58 41.86 20.93 6.98 0.254 pulse-output TBR raw pulse 324.3 0.254 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
30 TBR Davis 6466/6466M AeroCone 214 cm² 0.01 in 21.17 10.48 1270.20 127.02 42.34 21.17 7.06 0.254 pulse-output TBR raw pulse 214 0.254 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
31 TBR MeteoRain 200 Pro 0.254 mm / 0.01 in 21.24 10.49 1274.40 127.44 42.48 21.24 7.08 0.254 pulse-output TBR raw pulse 200 0.254 — Wired pulse output (B=0); no phase-code advantage. For short-window intensity, ranked by tip depth and collector area only.
32 TBR Onset HOBO RG3 15.39 cm 0.01 in 21.32 10.51 1279.20 127.92 42.64 21.32 7.11 0.254 pulse-output TBR raw pulse 186 0.254 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
33 TBR Texas/Campbell TR-525 / TE525 6 in 0.01 in 21.34 10.52 1280.40 128.04 42.68 21.34 7.11 0.254 pulse-output TBR raw pulse 182.4 0.254 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
34 TBR Casella TBRG 400 cm² 0.5 mm 40.89 20.43 2453.40 245.34 81.78 40.89 13.63 0.5 pulse-output TBR raw pulse 400 0.5 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.
35 TBR MicroStep/Meteoservis MR2 200 cm² 0.5 mm option 41.08 20.48 2464.80 246.48 82.16 41.08 13.69 0.5 pulse-output TBR raw pulse 200 0.5 — Pulse-output tipping bucket (B=0); ranked by collector area and tip depth.

Siphon-equipped tipping bucket gauges — 5 rows

Siphon-equipped tipping buckets carry the siphon-storage residual into short-window intensity uncertainty just as they do into accumulation uncertainty. The conservative γ_s = 1 assumption applies; a measured γ_s would shift these thresholds proportionally.

In-class
rank		 Gauge / output product P_min
at 1% (mm)		 P_min
at 2% (mm)		 1%
@1 min (mm/h)		 1%
@10 min (mm/h)		 1%
@30 min (mm/h)		 1%
@1 h (mm/h)		 1%
@3 h (mm/h)		 Δ_res
(mm)		 Evidence
layer		 State
detail		 A
(cm²)		 d / d_eq
(mm)		 V_s
(mL)		 Why it ranks here
1 Siphon KISTERS/HyQuest TB4 Series II 0.1 mm — siphon Vs=12 mL 17.73 8.83 1063.80 106.38 35.46 17.73 5.91 0.1 siphon TBR raw pulse + siphon 628.3 0.1 12 Bucket plus integrated siphon storage; γ_s = 1 conservative siphon residual. The siphon term is part of the short-window boundary uncertainty.
2 Siphon Geolux RG400 0.1 mm — siphon Vs≈16 mL, 4 mL/tip 33.98 16.91 2038.80 203.88 67.96 33.98 11.33 0.1 siphon TBR raw pulse + siphon 400 0.1 16 Bucket plus integrated siphon storage; γ_s = 1 conservative siphon residual. The siphon term is part of the short-window boundary uncertainty.
3 Siphon KISTERS/HyQuest TB4 Series II 0.2 mm — siphon Vs=12 mL 35.45 17.66 2127.00 212.70 70.90 35.45 11.82 0.2 siphon TBR raw pulse + siphon 314.2 0.2 12 Bucket plus integrated siphon storage; γ_s = 1 conservative siphon residual. The siphon term is part of the short-window boundary uncertainty.
4 Siphon Geolux RG200 0.2 mm — siphon Vs≈8 mL, 4 mL/tip 37.15 18.41 2229.00 222.90 74.30 37.15 12.38 0.2 siphon TBR raw pulse + siphon 200 0.2 8 Bucket plus integrated siphon storage; γ_s = 1 conservative siphon residual. The siphon term is part of the short-window boundary uncertainty.
5 Siphon KISTERS/HyQuest TB4 / Campbell CS700/CS700H 0.01 in — siphon Vs=12 mL 37.65 18.78 2259.00 225.90 75.30 37.65 12.55 0.254 siphon TBR raw pulse + siphon 314.2 0.254 12 Bucket plus integrated siphon storage; γ_s = 1 conservative siphon residual. The siphon term is part of the short-window boundary uncertainty.

Weighing gauges (public envelope, amount-derived) — 6 rows

These rows show the amount-derived public envelope (10 mm at 1%) for reference. The Pluvio² S Intensity RT and rain[e] native 1-minute intensity products are separate output products with a different floor (±6 mm/h native) — they are NOT amount-derived and are not directly comparable to the depth-threshold values in this column. See "Where weighing fits for intensity" below.

In-class
rank		 Gauge / output product P_min
at 1% (mm)		 P_min
at 2% (mm)		 1%
@1 min (mm/h)		 1%
@10 min (mm/h)		 1%
@30 min (mm/h)		 1%
@1 h (mm/h)		 1%
@3 h (mm/h)		 Δ_res
(mm)		 Evidence
layer		 State
detail		 A
(cm²)		 d / d_eq
(mm)		 V_s
(mL)		 Why it ranks here
1T Weighing Lambrecht rain[e] / rain[e]LP / rain[e]314 / rain[e]400 / rain[e]H3 public amount-total 10.00 5.00 600.00 60.00 20.00 10.00 3.33 — weighing public envelope public fixed floor — — — Weighing public envelope: ±0.1 mm or ±1% amount accuracy floor. The native real-time intensity product (Pluvio² S Intensity RT, rain[e] native 1-min) has a separate ±6 mm/h floor — see "Where weighing fits" below.
1T Weighing OTT Pluvio² L Accu NRT / Accu total NRT 10.00 5.00 600.00 60.00 20.00 10.00 3.33 — weighing public envelope NRT amount product — — — Weighing public envelope: ±0.1 mm or ±1% amount accuracy floor. The native real-time intensity product (Pluvio² S Intensity RT, rain[e] native 1-min) has a separate ±6 mm/h floor — see "Where weighing fits" below.
1T Weighing OTT Pluvio² S Accu NRT / Accu total NRT 10.00 5.00 600.00 60.00 20.00 10.00 3.33 — weighing public envelope NRT amount product — — — Weighing public envelope: ±0.1 mm or ±1% amount accuracy floor. The native real-time intensity product (Pluvio² S Intensity RT, rain[e] native 1-min) has a separate ±6 mm/h floor — see "Where weighing fits" below.
4T Weighing Lambrecht rain[e]one Modbus public amount-total — 5.00 — — — — — — weighing public envelope public fixed floor — — — Weighing public envelope: ±0.1 mm or ±1% amount accuracy floor. The native real-time intensity product (Pluvio² S Intensity RT, rain[e] native 1-min) has a separate ±6 mm/h floor — see "Where weighing fits" below.
4T Weighing OTT Pluvio² L Intensity RT — — — — — — — — weighing public envelope RT 1-min intensity — — — Weighing public envelope: ±0.1 mm or ±1% amount accuracy floor. The native real-time intensity product (Pluvio² S Intensity RT, rain[e] native 1-min) has a separate ±6 mm/h floor — see "Where weighing fits" below.
4T Weighing OTT Pluvio² S Intensity RT — — — — — — — — weighing public envelope RT 1-min intensity — — — Weighing public envelope: ±0.1 mm or ±1% amount accuracy floor. The native real-time intensity product (Pluvio² S Intensity RT, rain[e] native 1-min) has a separate ±6 mm/h floor — see "Where weighing fits" below.

Lambrecht rain[e] hi-resolution state-model layers — 6 rows

These rows are mathematical state-resolution model potential layers — what the analytical floor would predict if the advertised 0.001 mm scale-state or 0.01 mm pulse-state could be lifted to a real-time intensity product. They sit above all validated intensity rows in the model, but they require measured weighing time constant, filter bandwidth, and self-emptying-correction interval before they can be treated as defensible real-time intensity claims. The defensible Lambrecht row at this horizon is the public envelope above (10 mm at 1% amount-derived; or ±6 mm/h native intensity floor — see "Where weighing fits").

In-class
rank		 Gauge / output product P_min
at 1% (mm)		 P_min
at 2% (mm)		 1%
@1 min (mm/h)		 1%
@10 min (mm/h)		 1%
@30 min (mm/h)		 1%
@1 h (mm/h)		 1%
@3 h (mm/h)		 Δ_res
(mm)		 Evidence
layer		 State
detail		 A
(cm²)		 d / d_eq
(mm)		 V_s
(mL)		 Why it ranks here
1 Model only Lambrecht rain[e]400 — 0.001 mm hi-res scale-state model 0.635 0.166 38.13 3.81 1.27 0.635 0.212 0.001 rain[e] hi-res state model scale state: ≈6.64 bits 400 0.1 — Hi-resolution state-resolution model layer. Mathematical potential only; requires measured weighing response time, filter bandwidth, and self-emptying-correction interval before being treated as a real-time intensity claim.
2 Model only Lambrecht rain[e]314 — 0.001 mm hi-res scale-state model 0.804 0.207 48.27 4.83 1.61 0.804 0.268 0.001 rain[e] hi-res state model scale state: ≈6.99 bits 314 0.1274 — Hi-resolution state-resolution model layer. Mathematical potential only; requires measured weighing response time, filter bandwidth, and self-emptying-correction interval before being treated as a real-time intensity claim.
3 Model only Lambrecht rain[e]400 — 0.01 mm hi-res pulse-state model 1.19 0.494 71.22 7.12 2.37 1.19 0.396 0.01 rain[e] hi-res state model pulse state: ≈3.32 bits 400 0.1 — Hi-resolution state-resolution model layer. Mathematical potential only; requires measured weighing response time, filter bandwidth, and self-emptying-correction interval before being treated as a real-time intensity claim.
4 Model only Lambrecht rain[e] / rain[e]LP / rain[e]H3 — 0.001 mm hi-res scale-state model 1.25 0.318 75.30 7.53 2.51 1.25 0.418 0.001 rain[e] hi-res state model scale state: ≈7.64 bits 200 0.2 — Hi-resolution state-resolution model layer. Mathematical potential only; requires measured weighing response time, filter bandwidth, and self-emptying-correction interval before being treated as a real-time intensity claim.
5 Model only Lambrecht rain[e]314 — 0.01 mm hi-res pulse-state model 1.31 0.52 78.36 7.84 2.61 1.31 0.435 0.01 rain[e] hi-res state model pulse state: ≈3.67 bits 314 0.1274 — Hi-resolution state-resolution model layer. Mathematical potential only; requires measured weighing response time, filter bandwidth, and self-emptying-correction interval before being treated as a real-time intensity claim.
6 Model only Lambrecht rain[e] / rain[e]LP / rain[e]H3 — 0.01 mm hi-res pulse-state model 1.65 0.593 99.18 9.92 3.31 1.65 0.551 0.01 rain[e] hi-res state model pulse state: ≈4.32 bits 200 0.2 — Hi-resolution state-resolution model layer. Mathematical potential only; requires measured weighing response time, filter bandwidth, and self-emptying-correction interval before being treated as a real-time intensity claim.

Why MeteoRain IoT³ is dramatically ahead of the wired version in this ranking

For short-window intensity, the IoT³ phase code resolves the boundary residual at both ends of the reporting window to 1/8 of a tip rather than the full tip depth. The variance reduction is (1/8)² = 1/64 per boundary, equivalent to a roughly 64× lower boundary variance for a fully bracketed interior interval. In practical depth-threshold terms, this means the MeteoRain IoT³ 533 Classic at the 1% target needs ≈1.0 mm of rainfall to reach 1% short-window intensity uncertainty, vs ≈6.4 mm for the wired equivalent of the same hardware — a factor of about 6× improvement. The accumulation ranking shows the opposite picture: for daily totals where the event start and end are not bracketed by tips, the IoT³ and wired versions tie.

Where weighing fits for intensity

The OTT Pluvio² S Intensity RT and Lambrecht rain[e] native 1-minute intensity products have a different floor than the amount-derived rows in the table above. Their published 1-minute intensity floor is ±6 mm/h (or ±1%, whichever is larger), which translates to a 1-minute intensity threshold of 600 mm/h at 1%, 300 mm/h at 2%, and 200 mm/h at 3%. These are not depth thresholds — they are floors on the 1-minute moving-sum intensity output of the weighing instrument, dominated by the load-cell noise filtering and dynamic response, not by an inter-tip stochastic floor.

For sub-hourly intensity at low-to-moderate rain rates, the transmitted bracketed-state TBR (MeteoRain IoT³ B3/B3) sits at a depth threshold of 1.0–2.8 mm at 1%, equivalent to 60–166 mm/h at the 1-minute horizon. That is roughly an order of magnitude below the published 600 mm/h native weighing-intensity floor. Native real-time weighing intensity and bracketed-state TBR intensity are different output products with different physics; comparing them on equal evidence basis requires the response-gating discussion in the Lambrecht hi-res row sub-section above.

Manufacturer envelope validity bands

The analytical-floor ranking in the main tables above assumes each gauge is operating within its published manufacturer envelope. Different rate bands have different published accuracy specifications, and above the published band the analytical floor is no longer the dominant uncertainty term. The matrix below summarizes the typical envelope behavior by row category. For a specific application, always check the candidate gauge's published datasheet for the exact rate-band specification.

Row category Low rate (0–100 mm/h) Moderate (100–250 mm/h) High (250–500 mm/h) Extreme (>500 mm/h) Failure mode above design rate
TBR Raw-pulse 0.1 mm tipping bucket Spec'd ±1–2% per datasheet Dynamic loss begins; vendor-specific corrections may apply Significant dynamic loss (5–15%+ undercounting); above published envelope for most products Out of spec; uncorrected output unreliable Bucket cannot tip fast enough between fills; some water arrives during the tip and is lost
TBR Raw-pulse 0.2–0.5 mm tipping bucket Spec'd ±2–3% per datasheet Same dynamic loss onset as 0.1 mm Out of spec for most products Out of spec Same as above; coarser bucket has slightly higher tip rate before loss
IoT³ Phase-aware tipping bucket (B=3) Spec'd ±1% (Barani field-performance statement) Phase code preserves boundary accuracy; per-tip dynamic loss begins as in raw-pulse equivalent Dynamic loss as in raw-pulse same hardware; phase code does not mitigate dynamic loss Out of spec for the tip mechanism, regardless of phase code The 3-bit phase code corrects boundary uncertainty, not mechanical tip rate. Underlying tip mechanism is the same as the wired version.
Siphon Siphon-equipped tipping bucket Spec'd ±2% Spec'd ±2% (designed band, e.g. KISTERS TB4) Spec'd ±3% to 500 mm/h (KISTERS TB4); other siphon TBRs vendor-specific Above published envelope; siphon overflow possible Siphon buffers flow into bucket; designed to extend valid rate band well beyond raw-pulse TBRs
Weighing Public envelope, amount-derived (Pluvio² Accu NRT, rain[e] amount) ±0.1 mm or ±1% amount, rate-independent Same; rate-independent Same; rate-independent Same; bounded by load-cell capacity (typically 1500–3000 mm depth) Amount-derived product is intrinsically rate-independent; failure mode is load-cell saturation at extreme cumulative depth, not at high instantaneous rate
Weighing Native real-time 1-min intensity (Pluvio² Intensity RT, rain[e] native 1-min) ±6 mm/h or ±1% within published linear range Same; 1-min moving sum filters short bursts Response lag becomes visible; published envelope still ±1% but lag relative to true rate increases Above published intensity range; lag and saturation dominate 60-second moving-sum filtering smears short high-rate bursts; load-cell dynamic response and self-emptying timing matter
Model only Lambrecht hi-resolution state-model layer Mathematical model potential only — not validated at any rate without measured weighing response time, filter bandwidth, dead-time, and self-emptying-correction interval. Defensible row at this horizon is the public envelope above (10 mm at 1% amount, or ±6 mm/h native intensity). Validation gap, not rate-dependent failure

Behavior at high intensity (above the published envelope)

The analytical-floor model used in the rankings above assumes each gauge is operating within its published envelope. Above the envelope, mechanical and dynamic effects start to dominate and the ranking framework breaks down. This is most relevant for short-window intensity readings during heavy rain — flash-flood warnings, urban-drainage runoff at peak intensity, and convective cells with rates above several hundred mm/h.

Tipping-bucket dynamic loss

Above approximately 100 mm/h, raw-pulse tipping buckets begin to lose tips because the bucket cannot tip and refill fast enough between fills. Some water that arrives during the tip motion is lost or counted incorrectly. The classical references are Calder & Kidd (1978), Marsalek (1981), and Niemczynowicz (1986), with later refinements by Habib, Krajewski & Kruger (2001). The reported uncorrected undercount is typically 5–15% at 100–200 mm/h and grows roughly with rate; at 500 mm/h, raw-pulse tipping buckets without dynamic correction can underestimate by 20–40%.

This dynamic loss is not captured by the analytical-floor model in this article. The floor assumes that every drop arriving in the reporting window is counted exactly once with σ_tip variance per tip. Above the dynamic-loss onset, that assumption is violated and the actual short-window error is larger than the analytical floor predicts. Vendors that publish dynamic-correction curves (or built-in firmware corrections) extend the usable rate band but do not change the analytical-floor floor.

Why siphon TBRs exist

Siphon-equipped tipping buckets — KISTERS / HyQuest TB4, Geolux RG200/RG400, the Campbell CS700 family — buffer the incoming water flow upstream of the bucket. This decouples the bucket tip rate from the incoming rate, which is exactly what mitigates the dynamic-loss onset. The KISTERS TB4 datasheet states ±2% from 0–250 mm/h and ±3% from 250–500 mm/h, a band that raw-pulse TBRs cannot match. The trade-off is the siphon-storage residual term γ_s discussed in the main rankings: at low and moderate rates the siphon adds an unresolved boundary-state term that places siphon TBRs near the bottom of the analytical-floor accumulation ranking, but at high rates the same siphon design extends the usable rate band substantially. Siphon TBRs are not "worse" gauges — they are designed for a different rate band than the analytical-floor ranking foregrounds.

Weighing-gauge dynamic response

Weighing gauges measure mass rather than counting discrete tips, so they do not exhibit tip-rate-limited dynamic loss. Their high-intensity failure mode is different: load-cell dynamic response time, multi-stage filtering for wind and shock rejection, and self-emptying timing all add lag between the true rainfall rate and the reported short-window intensity. The Lambrecht rain[e] family explicitly publishes "6 measurements per minute" feeding a 60-second trailing moving sum; the OTT Pluvio² Intensity RT product is also a filtered 1-minute output. At high rates of change, the reported 1-minute intensity lags the true rate by approximately the filter response time. The analytical floor (±6 mm/h or ±1% native) describes the noise floor of the filtered output, not its lag.

Practical implication

For application contexts at sustained high rates (flash-flood warning, convective cell tracking, tropical cyclone observations), three considerations override the headline intensity-floor ranking shown in the main table above:

  • Check the manufacturer envelope. Use the validity-band matrix above to confirm the candidate gauge is spec'd in the rate band you need. Out-of-band operation invalidates both the published envelope and the analytical floor.
  • Prefer siphon TBRs or weighing-amount products at high rates. Raw-pulse TBRs (including the bracketed-state IoT³ family for short-window intensity) remain accurate at low-to-moderate rates but lose tips at high rates. The siphon design directly mitigates this; weighing-amount products (Pluvio² Accu NRT, rain[e] amount-derived) are fundamentally rate-resilient up to load-cell saturation.
  • For real-time short-window intensity at high rates, accept the trade-off. No gauge in this comparison set provides simultaneously: (i) the analytical-floor headline accuracy of phase-aware short-window intensity, (ii) freedom from tip-rate dynamic loss, and (iii) zero filtering lag. Application contexts must pick which two of the three to optimize.

References for tipping-bucket dynamic-loss correction

  • Calder, I.R., and Kidd, C.H.R., 1978. A note on the dynamic calibration of tipping-bucket gauges. Journal of Hydrology, 39(3–4), 383–386. doi:10.1016/0022-1694(78)90013-6
  • Marsalek, J., 1981. Calibration of the tipping-bucket raingage. Journal of Hydrology, 53(3–4), 343–354. doi:10.1016/0022-1694(81)90010-X
  • Niemczynowicz, J., 1986. The dynamic calibration of tipping-bucket raingauges. Nordic Hydrology, 17(3), 203–214.
  • Habib, E., Krajewski, W. F., and Kruger, A., 2001. Sampling errors of tipping-bucket rain gauge measurements. Journal of Hydrologic Engineering, 6(2), 159–166. doi:10.1061/(ASCE)1084-0699(2001)6:2(159)
  • Vuerich, E., Monesi, C., Lanza, L. G., Stagi, L., and Lanzinger, E., 2009. WMO Field Intercomparison of Rainfall Intensity Gauges (Vigna di Valle, Italy). WMO IOM Report No. 99, WMO/TD-No. 1504.

Siphon TBR storage state is conservatively modelled

Siphon-equipped tipping buckets carry the siphon-storage residual into short-window intensity uncertainty just as they do into accumulation uncertainty (γ_s = 1 conservative residual fraction). Sensitivity reference: at γ_s = 0.5 the KISTERS TB4 0.1 mm 1% threshold falls from 17.7 mm to 11.3 mm; at γ_s = 0 it falls to 8.2 mm. See sample calculation 3 in the appendix.

Reference: Lambrecht 1-minute intensity timing in plain language

For readers who want to understand why the public Lambrecht envelope (or the ±6 mm/h native intensity floor), rather than the hi-resolution state model, is the defensible row at this horizon:

Publicly defined itemPlain-language meaningWhy it matters here
6 measurements per minuteThe 1-minute intensity is built from six 10-second measurement steps.The native 1-minute intensity product behaves like a 60-second trailing moving sum, not like an instantaneous inter-tip measurement.
Moving sum of the last 6 valuesEach new value is combined with the previous five values to form "intensity within the last minute."A new rainfall step appears gradually during the first minute and then exits the product after it is older than 60 seconds.
Self-emptying by tippingWhen one chamber is full, the vessel tips and weighing continues in the other chamber.The tip should not erase rainfall from the total amount, but it creates a correction event that must be timed and filtered correctly.
Multi-stage filteringWind, shock, and vibration are filtered before the weight increase is evaluated.Filtering reduces noise but can also delay or smear short-window intensity changes.
Variance over 4 sThe instrument can output a short-term variance/quality value.This acknowledges noise, but it is not the same as a published numerical load-cell noise specification.

Appendix — methodology, sample calculations, and source data

This appendix gives the full mathematical model, a worked sample calculation for each row category in the comparison set, two visualizations of the full ranking, and a list of source documents. Any reader can reproduce every number in the article above using the formulas and inputs below.

A1. Methodology in plain language

The model decomposes the total uncertainty at a reporting interval into two independent parts: a per-tip stochastic floor (random variation in the volume of water needed to tip the bucket, or equivalent state-resolution variation for a weighing product) and a boundary-state residual (unresolved water at the start and end of the reporting interval). The two parts add as variances. Expanded uncertainty uses the GUM convention U95 = 2σ_total. The minimum rainfall depth needed to reach a target relative uncertainty q is the positive root of a quadratic equation derived from U95(P)/P = q.

σ_tip = 0.05 / (0.1 · A) (per-tip stochastic floor; A = collector area in cm²)
σ_B² = (Δ_start² + Δ_end²) / 12 (boundary residual variance, uniform-distribution assumption)
σ_total²(P) = σ_B² + (P/d) · σ_tip² (total variance at reporting interval depth P)
U95(P) = 2 · σ_total(P) (expanded uncertainty, GUM coverage factor k = 2)
P_min = [4σ_tip²/d + √( (4σ_tip²/d)² + 16 q² σ_B² )] / (2 q²) (solve U95(P_min)/P_min = q)

For ordinary tipping buckets transmitting only pulses (B = 0), Δ_start = Δ_end = d (the tip depth). For a phase-aware tipping bucket transmitting a B-bit phase code with both window boundaries inside an active rain event (the "B3/B3 interior" case used in this intensity ranking for the MeteoRain IoT³ family), Δ = d / 2^B = d/8 for B = 3. For siphon-equipped TBRs, the effective boundary variance is increased by the siphon-storage residual: σ_B² = (d² + γ_s² · s²) / 6 where s = 10 · V_s / A is the depth-equivalent siphon storage and γ_s ∈ [0, 1] is the unresolved-residual fraction (γ_s = 1 conservative). For weighing public envelope amount-derived rows, U_amount = max(0.1 mm, 0.01 · P) → P_min = 10 mm at 1%. For native 1-minute weighing intensity rows, U_I = max(6 mm/h, 0.01 · I) → I_min = 600 mm/h at 1%. For Lambrecht hi-resolution state-model rows, Δ is set to the advertised state resolution (0.001 mm or 0.01 mm), not the tip depth.

A2. Worked sample calculations — one per row category

Calculation 1 — Raw-pulse tipping bucket (Lambrecht 15189 0.2 mm, 200 cm²)

Inputs: A = 200 cm², d = 0.2 mm, B = 0 (no phase code transmitted)
σ_tip: 0.05 / (0.1 · 200) = 0.05 / 20 = 0.0025 mm
σ_B²: Δ_start = Δ_end = d = 0.2 mm → σ_B² = 2(0.2²)/12 = 0.08/12 = 0.006667 mm²
Quadratic at q = 0.01: q² = 1×10⁻⁴; 4σ_tip²/d = 4(0.0025²)/0.2 = 1.25×10⁻⁴; 4σ_B² = 0.02667
Discriminant: (1.25×10⁻⁴)² + 4(1×10⁻⁴)(0.02667) = 1.069×10⁻⁵; √ = 3.270×10⁻³
P_min: (1.25×10⁻⁴ + 3.270×10⁻³) / (2 × 1×10⁻⁴) = 3.395×10⁻³ / 2×10⁻⁴
P_min(1%) = 16.97 mm — same value as in the accumulation ranking, because raw-pulse TBRs (B = 0) have no phase-code advantage; intensity and accumulation share the same floor on the same hardware.

Calculation 2 — IoT³ phase-aware TBR (MeteoRain IoT³ 400 Aero, B3/B3 short-window — the case for this ranking)

Inputs: A = 400 cm², d = 0.1 mm, B = 3 (3-bit phase code), both boundaries bracketed by tips (interior of active rain event)
σ_tip: 0.05 / (0.1 · 400) = 0.05 / 40 = 0.00125 mm
σ_B² (B3/B3 interior): Δ = d / 2³ = 0.1/8 = 0.0125 mm; σ_B² = 2(0.0125²)/12 = d²/384 = 0.01/384 = 2.604×10⁻⁵ mm²
Quadratic at q = 0.01: 4σ_tip²/d = 4(0.00125²)/0.1 = 6.25×10⁻⁵; 4σ_B² = 1.042×10⁻⁴
Discriminant: (6.25×10⁻⁵)² + 4(1×10⁻⁴)(1.042×10⁻⁴) = 4.56×10⁻⁸; √ = 2.135×10⁻⁴
P_min: (6.25×10⁻⁵ + 2.135×10⁻⁴) / 2×10⁻⁴ = 2.760×10⁻⁴ / 2×10⁻⁴
P_min(1%, B3/B3 short-window intensity) = 1.380 mm — applies only to ongoing intensity windows interior to active rain. For daily/event totals (B0/B0 case), repeat with σ_B² = d²/6 = 1.667×10⁻³ mm² → P_min(1%, B0/B0 daily) = 8.483 mm, identical to wired MeteoRain 400 Aero. Equivalent intensity at 1-minute window: 1.380 × 60 = 82.8 mm/h.

Calculation 3 — Siphon TBR (KISTERS / HyQuest TB4 0.1 mm, V_s = 12 mL)

Inputs: A = 628.3 cm², d = 0.1 mm, V_s = 12 mL, γ_s = 1 (conservative)
σ_tip: 0.05 / (0.1 · 628.3) = 7.957×10⁻⁴ mm
Depth-equivalent siphon storage: s = 10 · V_s / A = 10 · 12 / 628.3 = 0.191 mm
σ_B² (bucket + siphon): (d² + γ_s² · s²) / 6 = (0.01 + 1·0.0365) / 6 = 0.007746 mm²
Quadratic at q = 0.01: 4σ_tip²/d = 2.532×10⁻⁵; 4σ_B² = 0.03098
Discriminant: 1.239×10⁻⁵; √ = 3.521×10⁻³
P_min: (2.532×10⁻⁵ + 3.521×10⁻³) / 2×10⁻⁴
P_min(1%, γ_s = 1) = 17.73 mm. At γ_s = 0.5: 11.3 mm. At γ_s = 0: 8.16 mm. Same value used for both intensity and accumulation rankings (the siphon term carries through). Equivalent intensity at 1 min: 17.73 × 60 = 1064 mm/h.

Calculation 4 — Weighing native real-time 1-minute intensity (OTT Pluvio² S Intensity RT)

Inputs: Public manufacturer envelope U_I = max(0.1 mm/min, 0.01 · I) = max(6 mm/h, 0.01 · I)
Solve U_I / I ≤ q at q = 0.01: max(6/I, 0.01) ≤ 0.01 → 6/I ≤ 0.01 → I ≥ 600 mm/h
I_min(1%, native 1-min intensity) = 600 mm/h. At q = 0.02: 300 mm/h. At q = 0.03: 200 mm/h. The native real-time intensity output uses the manufacturer specification directly; it is dominated by load-cell noise filtering and dynamic response, not by an inter-tip stochastic floor. For comparison, the amount-derived row (OTT Pluvio² S Accu NRT amount) gives 10 mm at 1%, equivalent to 600 mm/h at 1 minute (60 × 10 mm) — coincidentally the same value at the 1-minute horizon, but only because the amount-derived ratio happens to align with the native intensity ratio at this specific horizon.

Calculation 5 — Lambrecht hi-resolution scale-state model layer (rain[e]400, Δ_res = 0.001 mm, model only)

Inputs: A = 400 cm², d_eq = 0.1 mm, Δ_res = 0.001 mm (advertised scale resolution treated as effective state resolution in the model)
σ_tip: 0.05 / (0.1 · 400) = 0.00125 mm (same as MeteoRain IoT³ 400 Aero, because A is the same)
σ_B²: 2(Δ_res)²/12 = (0.001)²/6 = 1.667×10⁻⁷ mm²
Quadratic at q = 0.01: 4σ_tip²/d = 6.25×10⁻⁵; 4σ_B² = 6.667×10⁻⁷
Discriminant: (6.25×10⁻⁵)² + 4(1×10⁻⁴)(6.667×10⁻⁷) = 4.173×10⁻⁹; √ = 6.460×10⁻⁵
P_min: (6.25×10⁻⁵ + 6.460×10⁻⁵) / 2×10⁻⁴
P_min(1%, model layer) = 0.6355 mm — equivalent to 38.1 mm/h at 1-minute window. Caveat: this assumes the advertised 0.001 mm scale resolution can be lifted to a defensible real-time intensity product. In practice this requires a measured weighing time constant, filter bandwidth, dead-time, and self-emptying-correction interval. Without that validation, the row sits in its own category in the main table above with yellow shading and a "Model only" badge to signal that it is mathematical model potential, not a validated short-window intensity product.

A3. Visualizations

Bar chart of required depth at 1% target across 52 rows for short-window intensity, color-coded by category
Figure A1. Required rainfall depth at the 1% accuracy target for short-window intensity, across the 52-row comparison set, sorted by required depth and color-coded by row category. Lambrecht hi-resolution state-model rows are shown in yellow at the top of the plot — these are model-potential layers that require response-gating validation before being treated as real-time intensity products. The IoT³ phase-aware rows (blue) sit between the hi-res model rows and the raw-pulse TBR rows (green). Within the raw-pulse TBR class, the smallest tip depth and largest collector area win.
Required depth vs target uncertainty for representative short-window intensity products
Figure A2. Required rainfall depth as a function of the relative uncertainty target q for one representative row from each category, in short-window intensity context. The Lambrecht hi-res scale-state model curve sits at the bottom (lowest threshold), the IoT³ phase-aware B3/B3 short-window curve next, then siphon TBR, raw-pulse TBR, and finally the weighing public envelope (a hyperbola in q because the floor is a fixed amount-resolution rather than a stochastic per-tip floor).

A4. Conflict-of-interest and funding statement (full)

The author is the chief executive officer and a founder of Barani Design Technologies, the manufacturer of the MeteoRain product family included in the comparison. The MeteoRain IoT³ rows occupy several positions in the upper-middle of the ranking (blue category). The Lambrecht hi-resolution scale-state and pulse-state model rows (yellow category) sit above the MeteoRain IoT³ rows in the table because they have the smallest mathematical state resolution; they are model-potential layers that the author does not manufacture and that are presented with their full caveat about response-gating validation. The author has no financial relationship with the other manufacturers (KISTERS / HyQuest, OTT HydroMet, Lambrecht, Geolux, Vaisala, Texas/Campbell, Casella, Davis, Onset HOBO, Met One, RainWise, MicroStep / Meteoservis, AEM/FTS, CAE, Pronamic, Seven Sensor Solutions, EML).

The work was not funded by any vendor of any compared product. No external party reviewed, edited, or approved this article prior to publication. All inputs are publicly available datasheets and operating manuals; no proprietary, internal, or vendor-supplied data was used for any product, including the author's own products. Conservative modelling parameters are applied uniformly: U95 = 2σ_total (GUM coverage factor k = 2, more conservative than a triangular-distribution k = 1.90); γ_s = 1 for siphon TBRs; B3/B3 only for IoT³ phase-aware rows when the reporting window is interior to active rain (the assumption that defines short-window intensity in the first place — for accumulation totals where the assumption fails, the IoT³ rows revert to B0/B0 in the companion accumulation ranking).

Designing a measurement instrument requires a working uncertainty model of the same instrument; the author's role as instrument designer is therefore not separable from the author's role as analyst. The appropriate response to this dual role is transparency rather than recusal. Readers who suspect bias are encouraged to substitute alternative parameter values into the formulas in section A1, re-run any sample calculation in section A2, and verify whether the conclusions change. The full peer-reviewed manuscript with detailed derivations is available on request.

A5. Source notes and verification

BARANI MeteoRain product sizes and IoT³ descriptions: MeteoRain product overview and MeteoRain IoT.

KISTERS / HyQuest TB4 Series II rows: documented siphon storage volume V_s = 12 mL applied with γ_s = 1 conservative residual fraction. Reference: KISTERS TB4 Series II user manual (publicly available).

Geolux RG200/RG400 siphon rows: working values RG200 ≈ 8 mL and RG400 ≈ 16 mL, both at approximately 4 mL/tip. Seven Sensor Solutions 3S-RG is treated as a non-siphon 200 cm² 0.2 mm tipping bucket.

OTT Pluvio² S/L treatment: Intensity RT and Accu NRT / Accu total NRT are different output products with different floors. The Intensity RT product (native real-time 1-minute intensity, ±6 mm/h floor at 1%) is the row that matters most for short-window intensity; the Accu NRT amount-derived row (10 mm at 1%) is shown for reference because it defines the public-envelope amount floor.

Lambrecht rain[e] family: public envelope (0.1 mm or 1% amount accuracy floor) and operating manual (6 measurements per minute, 60-second moving sum, self-emptying compensation, multi-stage filtering, 4-second variance output). The hi-resolution 0.001 mm scale-state and 0.01 mm pulse-state model layers are presented with strong caveats about response gating.

All other listed competitors (Vaisala, Texas/Campbell, Casella, Davis, Onset HOBO, Met One, RainWise, MicroStep / Meteoservis, AEM/FTS, CAE, Pronamic, EML) are modelled from publicly available datasheets and product manuals using the formulas in section A1 above. No proprietary data is used.

The full mathematical model, derivation, per-horizon regime selection logic, and γ_s sensitivity analysis are documented in the v23 manuscript "Low-Power Rain-Gauge Uncertainty Models" (Barani, 2026), available on request. The companion accumulation ranking article uses the same model with the boundary regime swapped to B0/B0 (event-bound).