Tiếng ViệtTalk to us
Measured result · 31 May – 13 Aug 2026 · pilot GMX-NB-2026-01

Dawn dissolved oxygen, two populations that barely overlap

Same farm, same water source, same feed lot, same crop. One variable changed: how oxygen gets into the water. Everything below is computed at build time from the pilot dataset — the numbers on this page cannot drift from their source.

Nanobubble ponds (n=2)Control ponds (n=3 of 4 instrumented)Below 4 mg/L
BELOW 4 mg/L — SUB-OPTIMAL FOR GROW-OUT0246810mg/L1 Jun1 Jul1 Aug13 AugNANOBUBBLE mean 8.36 mg/LCONTROL mean 5.48 mg/L
Fig. 1 — Daily mean dawn (05:00–07:00 local) dissolved oxygen, one line per pond. A line breaks at every missing day — nothing is bridged or interpolated, including the farm-wide gateway outage on 17 Jul. Mean water temperature over the window 28.3 °C. Hover the chart for the underlying readings.
234567891011mg/L at dawnNANOBUBBLE113 pond-days · median 8.42 mg/LCONTROL164 pond-days · median 5.64 mg/L< 4 mg/L
Fig. 2 — The same data as a distribution: every pond-day is one dot. Control ponds fall below 4 mg/L on 9 of 164 pond-days and below 3 mg/L on 3; nanobubble ponds do neither. Lowest instantaneous dawn reading 2.27 mg/L in a control pond, 5.92 mg/L in a nanobubble pond.

The gap survives re-specification

Three different ways of asking the same question, all from the same dataset.

+2.82 / +2.99
mg/L advantage in cohort A and cohort B computed independently — two separate nursery batches, same answer.
72 of 72
calendar days on which the nanobubble group mean was higher, out of every day where both groups reported. No day reversed.
+2.79
mg/L mean advantage on a strict day-matched comparison (SD 0.96, smallest daily gap +0.75).

Pond by pond

PondPond-daysCoverageMean mg/LSDLowest daily meanDays < 4Days ≥ 6
NANOBUBBLEPond 16799%8.35±0.657.050.0%100%
NANOBUBBLEPond 54661%8.37±0.826.810.0%100%
CONTROLPond 24965%5.42±0.843.316.1%35%
CONTROLPond 35698%5.38±0.743.181.8%25%
CONTROLPond 45997%5.61±0.992.858.5%47%
4 mg/LPond 5 · NB8.37Pond 1 · NB8.35Pond 4 · CTL5.61Pond 2 · CTL5.42Pond 3 · CTL5.38
Fig. 3 — Mean dawn dissolved oxygen by pond, sorted. The dashed line is the 4 mg/L grow-out threshold.

Summary statistics

NanobubbleControl
Pond-days113164
Mean dawn DO (mg/L)8.365.48
Median dawn DO (mg/L)8.425.64
Standard deviation0.720.87
Lowest daily mean (mg/L)6.812.85
Lowest instantaneous reading (mg/L)5.922.27
Pond-days < 4 mg/L09
Pond-days < 3 mg/L03
Pond-days ≥ 6 mg/L113 (100%)59 (36%)
MeasuredData as at 13 Aug 2026 · Source data/pilot_data.json

Data provenance, including what went wrong. Every figure on this page is computed from the live pilot database rather than transcribed from a report: 113 nanobubble and 164 control pond-days, from four parameters per pond read every 30 seconds — of the order of 11,500 values per pond per day. Coverage is the share of days with valid dawn data inside each pond’s own measurement span: two ponds lost their sensor feed for about four weeks in mid-2026 (Pond 2, 12 Jun – 6 Jul; Pond 5, 11 Jun – 6 Jul), and a farm-wide gateway outage removed 17 July from all five — that, not selection, is what the 65% and 61% coverage figures record. Six of Pond 2’s missing days had logged 0.0 mg/L and were dropped under one numeric rule — a dawn mean below 0.5 mg/L is treated as dropout — applied identically to both groups across the whole dataset. Nothing is interpolated. Pond 6 is uninstrumented and is excluded from all sensor analysis.

Known gaps in the record

Every interruption in the sensor feed, as recorded in the dataset. These are the breaks you see in Fig. 1.

PondFromToDays
Pond 57 Jun 20268 Jun 20262
Pond 511 Jun 20266 Jul 202626
Pond 212 Jun 20266 Jul 202625
Pond 46 Jul 20266 Jul 20261
Pond 117 Jul 202617 Jul 20261
Pond 217 Jul 202617 Jul 20261
Pond 317 Jul 202617 Jul 20261
Pond 417 Jul 202617 Jul 20261
Pond 517 Jul 202617 Jul 20261

Excluded pond-days (dawn mean < 0.5 mg/L → dropout rule): Pond 2 · 12 Jun · Pond 2 · 13 Jun · Pond 2 · 14 Jun · Pond 2 · 15 Jun · Pond 2 · 16 Jun · Pond 2 · 17 Jun

Method

How the trial is built

  • Six ponds, one farm. 1,500 m² each at 1.2 m depth (1,800 m³), lined, outdoor. Penaeus (Litopenaeus) vannamei in two phases: 20-day nursery, then grow-out at 178 animals/m².
  • Two treatment, four control, one variable. Cohort-balanced pairs from the same nursery batch; oxygen delivery is the only thing that differs. Seed source, nursery SOP, water source, feed lot, probiotics and staff rotation are held common by protocol.
  • Commissioned before stocking. A four-test capacity SOP with written pass/fail criteria, including a safety cut-off at 13.5 mg/L, was run on empty ponds and signed off before the grow-out cohort was stocked.
  • Pre-specified analysis. Mixed-effects model with pond nested in cohort; hypotheses, model form and power were fixed in a versioned internal protocol before stocking (v2.5, fourteen logged versions). It is not lodged in a public registry.
  • External laboratory. PCR panel (WSSV, IMNV, EHP, TSV), hepatopancreas assessment, Vibrio counts and haematology are contracted out on the schedule in the protocol.
  • Published standards. ISO 5814 for dissolved oxygen, APHA Standard Methods for nutrients and chlorophyll-a, WOAH protocols for pathogen PCR.
Oxygen generation and nanobubble units commissioned pond-side
Oxygen generation and nanobubble units commissioned pond-side, May 2026. Commissioning is a written test with pass/fail criteria, not a demonstration.
Pond-side installation work at the pilot farm
Installation, Bac Lieu
Pond infrastructure and distribution lines at the pilot farm
Distribution lines across the pond
Panoramic view of the lined intensive pond
Grow-out pond, 1,500 m²

Evidence register

MeasuredGMX-NB-2026-01 · 31 May – 13 Aug 2026

Dawn dissolved oxygen, nanobubble versus control

Same farm, same water source, same feed lot, same crop. One variable changed: how oxygen gets into the water. Everything below is computed at build time from the pilot dataset — the numbers on this page cannot drift from their source.

Pending harvestGMX-NB-2026-01-ADDENDUM · Readout 29 Aug 2026

Production results at harvest

Production results read out at harvest, 29 August 2026; a data addendum will follow. Survival, feed conversion, growth, yield, energy and economics are deliberately not claimed on this site until the harvest figures have been reviewed and approved. This entry will be replaced by the addendum, with the same provenance rules as the dawn-oxygen dataset above.

Global Shrimp Forum 2026 · Utrecht · 1–3 September

Bring us a pond. We will bring the instruments.

We would rather show you a dataset from your own water than argue about ours. If you run intensive ponds and want one of them measured properly, that conversation takes fifteen minutes.

Email Ngo Thanh Son
Ngo Thanh Son
General Director · Co-founder
sonnt@novamed.vn
+84 917 383 668
Hanoi · Bac Lieu, Vietnam