Missing values permitted a five-point mean depth reduction in four Atlantic and Pacific coral samples from 2003 to 2019
We tested whether missing coral bleaching values can hide a mean five-percentage-point reduction in deeper water. We used 1,387 archived Reef Check events at both depths in the Atlantic and Pacific from 2003 to 2019. Each site had equal weight. Shallow depths were 0-6 m and deeper depths were 8-20 m. There were 256 incomplete events. If each missing percentage can be from 0 to 100, all four ocean-period ranges include a mean difference of -5 points. The smallest absolute event-difference caps that permit this target are 63.485, 47.057, 17.014, and 36.533 points. A separate Pacific sample from 2012 to 2019 has 418 events and bounds of -0.412 to +0.173 points. It cannot hide the same target. High-heat samples do not give a common result. These are exact ranges for selected records, not confidence intervals or evidence that a depth refuge occurred. Missing sites, species differences, measurement error, and unknown reporting processes limit inference. The package gives a paired evidence frame, code, exact endpoint assignments, and independent linear-program checks.
coral bleaching · missing outcomes · depth refuge · observation bias · partial identification · marine heat stress · Reef Check · paired surveys
References
8- usesdoi:10.26008/1912/bco-dmo.773466.2
Raw public coral and environmental records, version 2.
- usesdoi:10.1038/s41597-022-01121-y
Source database description and survey structure.
- usesdoi:10.25921/ffw7-cs39
Supplied CoRTAD thermal fields define a secondary high-dose subset.
- usesdoi:10.1016/j.ijar.2004.10.006
Bounded-outcome reasoning for missing data; we apply it to finite paired records.
- reviewsdoi:10.1088/1748-9326/ad89df
Prior depth and heat modelling motivates an audit of record selection; our depth range and target differ.
- reviewsMatched coral surveys gave no clear depth refuge from bleaching at 0-20 m in three oceans in 1998-2019 sci:sha256:5c974d1e8f746429bf1b48b3297c0773f6c23fe4ce90bba7ed367307a96c5463
Prior complete-case depth study uses overlapping records. We bound a larger archived event frame with missing outcomes.
- reviewsMatched Reef Check surveys showed small mean depth contrasts in coral bleaching at 0-30 m in 1998-2019 sci:sha256:4a789955ad022aafa7fda357e761144e3c74ab8788fe72415b219ba6fe1192b4
Prior exact-date depth contrasts use overlapping records. We add missing-outcome ranges and minimum sensitivity caps.
- reviewsBleaching values were scarce before 2003 in the Reef Check part of BCO-DMO 773466 version 2 sci:sha256:287644233b61dd382329bfa085108546c44a59b8953b98ab6f3ccaa671225346
Prior reporting audit and period bounds motivate the post-2003 restriction. We extend the audit to paired depth means and explicit tipping assignments.
Cited by
5 publicationsPublic catalog entries on Sciscape.
- reviewsMissing percentages permit both heat slope signs for severe coral bleaching in four Atlantic and Pacific samples in 2003-2019
Prior paired depth bounds motivate an exact missing-outcome audit for a different adjusted coefficient.
- reviewsMissing follow-ups permit gains and losses in four high-heat coral cohorts in 1998-2018
Related missing-value bounds for a different bleaching depth outcome.
- reviewsPriority recovery halved the calculated depth bound width with 149 reef values under equal cost in 2003-2019
Published archive tests motivate a source-record recovery design. Observations overlap.
- reviewsMissing percentages permitted both heat-slope signs in three of four Atlantic and Pacific coral samples in 2003-2019
Prior bounds for a depth contrast motivate a different continuous heat-coefficient target.
- reviewsZero bleaching accounted for 996 of 1,052 exact depth matches in Reef Check surveys from 1998 to 2019
The missing-value study addresses a separate observation limit in overlapping records.