Six years of NOAA forecasts, 160 measured oxygen records and Gulf fish research explain habitat loss, the 2026 forecast miss and what a good catch cannot tell us.
A small dead zone sounds like a good fishing forecast. A large one sounds like a warning that the fish will disappear. Both interpretations skip the most important part of the evidence: oxygen is habitat, and habitat changes with depth, location and time. An annual area estimate does not describe every layer of water, and a catch does not measure the health of the whole shelf.
The 2026 season makes that problem unusually clear. NOAA forecast 7,027 square miles of bottom hypoxia; the summer cruise measured 1,332. Our comparison of six consecutive forecast-and-survey pairs, a separate historical cruise’s 160 bottom-oxygen records, and published Gulf fish research examines why those numbers can diverge—and what their differences actually tell us.
Our conclusion: hypoxia can reduce usable bottom habitat and reorganize fish distribution and feeding even where anglers still catch fish. Nutrient inputs help establish the seasonal potential, while stratification and weather help determine the footprint a cruise encounters. The useful outlook is conditional on those processes. Neither a good catch nor one small annual map establishes that the underlying problem has gone away.
The 2026 mismatch is the central case
NOAA’s June forecast anticipated an above-average hypoxic area of 7,027 square miles. Its published 95% confidence interval extended from 3,895 to 10,021 square miles. The LSU/LUMCON shelfwide survey, conducted July 24–30, subsequently measured 1,332 square miles, the second-smallest measurement in NOAA’s 40-year record. The measurement was 5,695 square miles below the point forecast, or 81.0% smaller, and below the forecast interval’s lower bound. These are calculations from the reported values, rather than a new estimate of the hypoxic area.
NOAA attributes the smaller measured footprint to Tropical Storm Bertha passing through the northern Gulf immediately before the cruise. Strong winds and waves mixed oxygenated surface water downward. That process changes the oxygen balance quickly: the water can receive an oxygen supply that a seasonal forecast assuming average weather did not anticipate. NOAA’s August 2026 survey account.
This is an informative forecast miss. The prediction describes conditions expected from the model inputs and weather assumptions; the cruise describes the footprint encountered after an important intervening event. A nutrient-driven hypoxia problem and storm-driven temporary oxygen replenishment can both be real. Mixing does not, by itself, demonstrate that the watershed delivered fewer nutrients or that oxygen demand in the bottom system permanently disappeared.
There is another limitation in the event account: NOAA says the cruise did not capture the full extent of the zone. We therefore treat 1,332 square miles as the reported survey result, with its survey context, rather than a complete inventory of every low-oxygen place or every hypoxic day during summer. The source’s storm explanation is an agency interpretation of that event; our six-year arithmetic does not independently isolate a storm effect from all other drivers.
Six years show why a point forecast needs its uncertainty
We assembled every annual NOAA ensemble forecast and subsequent survey result from 2021 through 2026, using the original annual forecast sheets for interval endpoints. No intervening year was dropped. The comparison contains two years in which the survey area exceeded the point forecast and four in which it was smaller. Five of the six measurements fall inside their source’s published interval; 2026 is the exception.
The mean absolute difference between point forecasts and survey measurements is about 2,064 square miles across these six years. That statistic is the average size of the miss in either direction. It is useful for resisting false precision, but it is not a validation of all operational forecasts, an estimate of next year’s error, or a basis for ranking individual models. Six cases, evolving model ensembles and variable summer weather cannot supply those conclusions.
The relatively broad intervals are part of the finding. In 2024 the forecast was 5,827 square miles and the survey found 6,705; in 2025 the forecast was 5,574 and the survey found 4,402. Those opposite-signed errors occurred inside the respective published intervals. Reading only the central values would make an uncertain forecast look like a promise of one exact area. Reading the interval still does not make a forecast a map of which reef will have oxygen on a particular date.
A particularly revealing comparison is 2021 against 2026. The 2021 May nitrate load reported in the forecast sheet was 90,500 metric tons and the measured summer area was 6,334 square miles. In 2026 the reported May nitrate load was higher, at 134,500 metric tons, but the measured area was much smaller. That two-year contrast is not evidence that nitrate protects the shelf. It shows why a cross-year relationship cannot be interpreted while ignoring weather, stratification, survey timing and the other processes connecting a spring input to a summer observation.
The original sheets call their bounds “95% confidence intervals,” which is the label preserved in our table. We have not converted them into newly calculated prediction intervals or assumed that five inclusions out of six establish 95% long-run coverage. The full six-year source and nutrient-input table is available with the downloads.
Why freshwater can help create a bottom-oxygen problem
The dead zone involves both oxygen demand and restricted oxygen replacement. River-borne nitrogen and phosphorus can stimulate production in surface waters. Organic material subsequently sinks, and decomposition consumes oxygen. Meanwhile, relatively fresh surface water can remain above denser saline water, limiting exchange between the surface and bottom. Warm-season heating can strengthen density differences as well. A bottom oxygen deficit develops when consumption outpaces replacement.
This explains a result that is otherwise counterintuitive to an angler: surface activity and depleted bottom water can occur in the same general area. The surface receives oxygen through exchange with the atmosphere and photosynthesis. Conditions several metres below it may be very different. Water color, bait on top, a surface temperature reading and a bottom contour on the chart describe useful parts of the scene, but none is a dissolved-oxygen profile.
Freshwater volume and nutrient load also need to stay separate. NOAA reports that May 2026 Mississippi–Atchafalaya discharge was about 36% below its 1980–2025 average, while the nitrate load was about 4% above average and phosphorus about 6% below. A low-flow spring therefore did not mean all nutrient loads were low. A load measures the amount transported over time; both water volume and nutrient concentration affect it. NOAA’s June 2026 forecast context.
The river also influences where its freshwater and constituents travel. The established Louisiana shelf problem is connected to the Mississippi–Atchafalaya system, coastal transport, shelf depth and seasonal mixing. It is not a uniform oxygen deficit across every Gulf waterbody. The long-term synthesis by Rabalais and Turner describes evidence from shelf cruises, repeated transects, moored oxygen instruments and sediment history. Those different records support the mechanism over much longer periods than the six annual pairs assembled here. Primary synthesis, Gulf of Mexico Hypoxia: Past, Present, and Future.
Our freshwater-pulse and salinity case addresses freshwater distribution at measured locations. This study adds a separate consequence that depends on the water column’s density structure and oxygen demand. A salinity change can help explain the physical setting without establishing a particular bottom oxygen concentration or a redfish catch response.
What 160 measured oxygen records add to the story
To move beyond annual area headlines, we analyzed NOAA Hypoxia Watch’s bottom-oxygen records from Oregon II cruise 332, June 9–July 18, 2019. The CSV and public observation service each contain 160 records, with matching station identities and oxygen values. All 160 have usable oxygen, position and water-depth fields under our stated checks. Twenty-three records meet the conventional bottom-hypoxia criterion of 2 mg/L or less.
The observations span a large Gulf sampling footprint, not just Louisiana. We therefore show all points and separately summarize a declared analysis box from 94°W to 89°W and 28°N to 30°N. That box captures 48 observations from June 16 through July 1. It is a geographic research window, not Louisiana’s jurisdictional boundary. Six of those 48 observations are at or below 2 mg/L.
These denominators are essential. Six out of 48 is 12.5% of the sampled observations inside that box. It is not an estimate that 12.5% of Louisiana’s seafloor was hypoxic. Survey stations have a sampling design, the observations were taken on different days, and the surrounding seabed is not divided into 48 equal areas. We have not interpolated an area or reconstructed the annual shelfwide map from these points.
Inside the box, the eight observations in water shallower than 20 metres include four hypoxic records and have a median near-bottom oxygen concentration of 1.27 mg/L. The 18 observations at 20–50 metres include two hypoxic records, with a median of 3.23 mg/L. The 22 observations at 50–200 metres include none, with a median of 4.78 mg/L. The full regional range is 0.00–5.80 mg/L; a single regional average would hide much of that habitat contrast.
The contrast is substantial within this cruise, but depth alone cannot explain it. The shallow and deep points occupy different locations and were sampled at different times. Bottom characteristics, freshwater influence and mixing can vary with those locations. This is an observational description, not an experiment in which only depth changed. “Go deeper” would turn a bounded result into an unsupported universal rule.
The depth fields deserve a closer look too. NOAA reports both total water depth and sample depth. In 95 of the 160 records the reported sample depth is slightly greater than the reported water depth; differences across the complete file range from −4.9 to +3.0 metres. We retain both original fields rather than subtracting them to invent a precise probe clearance above the seabed. Grouping the regional observations by sample depth instead of total water depth produces the same three bin counts and hypoxic counts in this case. That sensitivity check supports the reported grouping, while leaving the measurement-reference discrepancy unresolved.
This historical cruise is separate from the annual LSU/LUMCON survey used for the forecast comparison. It supplies spatially located, near-bottom oxygen observations, not paired surface-and-bottom profiles, fish catches or conditions at a reef today. Showing the entire footprint makes that distinction inspectable.
Why the newest station map is not necessarily the newest usable dataset
We also checked the current public Hypoxia Watch station service for summer 2026. It returned 189 station-location records. Every one had a null value in the dissolved-oxygen field at the time of retrieval on October 21, 2026. The accompanying downloadable station shapefiles provide locations and depth information; they do not supply the missing oxygen readings.
A null is an unknown value, not 0 mg/L. Treating these 189 nulls as zeros would manufacture an entire cruise of anoxic observations. Plotting their locations as though they supplied oxygen measurements would be misleading for a different reason. We retain the response and show the usable-data count in the methods so this choice can be checked.
This result describes the public endpoint we inspected, not an assertion that NOAA never collected oxygen during 2026. The agency’s current map has migrated to a different hosted service; an older API link returned “Service not found.” Some historical full-profile archive links also returned error pages during acquisition. We used the working, source-advertised historical oxygen CSV and checked it against the replacement public service. We do not label unavailable profiles as downloaded or turn station locations into a vertical analysis.
The episode illustrates why a research page needs acquisition checks as well as a list of citations. A date on a map and an apparently successful HTTP download do not establish that the file contains the measurements needed to answer the question. Our public data include the actual saved responses, while the repository retains the unsuccessful access evidence.
The fish evidence measures a response, not just an oxygen map
Environmental measurements show where habitat may be constrained. To ask what fish actually do, we examined Glaspie and colleagues’ 2019 study of fish catches and diets in the northern Gulf. Its research cruises covered August 4–13, 2006; July 30–October 14, 2007; and August 1–11, 2008. Fish came from bottom and midwater trawls, with oxygen measured by a CTD instrument. These were research catches, rather than recreational catch reports collected around a fishing rumor.
The authors express catch per unit effort as fish per minute at the targeted trawling depth. Their catch analysis distinguishes 91 trawls totaling 1,707 minutes above the fitted 1.15 mg/L breakpoint from 46 trawls totaling 943 minutes at or below it. The uneven effort explains why raw totals alone would be a poor comparison. Even effort-standardized catches remain affected by the gear and by where fish are available to it.
The study reports lower catch per unit effort in hypoxic areas for several species, including red snapper, Atlantic croaker, Gulf butterfish and gray triggerfish. That finding is directly relevant to the claim that oxygen has no fishing consequence because someone caught fish nearby. A fishery can remain active while the location, species mix and availability of fish change. Glaspie et al., primary manuscript and citation.
An important qualification comes from the sampling gear itself. The bottom net did not close during deployment and retrieval, so some fish could have been caught above the bottom layer. A fish collected over hypoxic bottom water is therefore not proof that it spent the tow living at that bottom oxygen concentration. This helps explain why “fish present” and “bottom habitat impaired” are not contradictory observations.
The fitted fish-community breakpoint was 1.15 mg/L, with a 95% bootstrap interval of 0.99–3.24. The diet breakpoint was 1.71 mg/L, with an interval of 0.98–3.33. The latter point estimate is higher, suggesting that feeding changes may be detectable before a sharper catch-community change. But the intervals overlap widely. They do not establish two precisely ordered physiological switches, and they are not survival thresholds for an individual snapper.
The common 2 mg/L threshold serves a different purpose: consistent classification and mapping of low-oxygen bottom habitat. A fitted community breakpoint below it does not prove that all concentrations above 1.15 mg/L are harmless. The study’s labels depend on its particular response variable and analysis; the number defining a map, the number marking a statistical change and the oxygen requirement of a specific fish are different quantities.
The diet work adds a mechanism that a simple presence-or-absence map misses. The researchers measured prey in stomachs, expressed prey biomass relative to fish dry mass, and averaged same-species stomach results within a trawl. That last choice helps avoid treating every fish in one tow as an independent environmental replicate. Their multi-factor analysis considered diet group, species, year, day/night and bottom temperature, with additional checks for location and sampling-day structure.
Fourteen fish species had enough processed stomachs in both oxygen categories for the diet-group analysis. Oxygen-related diet differences were detected for zooplankton feeders and bottom-prey feeders; the fish/mobile-prey group did not show a detected group-level difference in that test. The reported p-values are 0.001, 0.03 and 0.20, respectively. These measure evidence for a diet-composition association under the authors’ model. They do not quantify the percentage loss of prey, fish growth or fishery production.
Some small zooplankton, including Oithona and Paracalanus, contributed more mass to the diets of zooplankton-feeding fish caught in hypoxic areas. Several bottom-associated prey groups contributed more mass in oxygenated areas. The authors discuss a plausible redistribution mechanism: mobile animals can move away from low oxygen, changing where predators and prey meet, while access to bottom prey becomes more difficult.
This is why the effect cannot be summarized as uniformly “more feeding at the edge.” Compression can make some prey easier to encounter at the same time that it removes access to others. The identity and size of the predator, its diet and its ability to move all affect the result. A lack of a detected overall diet effect for one group is also not proof that the two oxygen conditions are biologically equivalent.
Red snapper provide a useful caution about applying categories. Table 1 lists 144 red snapper caught, and the analyzed snapper fall into the study’s zooplankton-feeding cluster based on their sampled stomach contents. That is a classification of these research samples. It does not redefine the diet of all adult snapper at Louisiana reefs. The trawl evidence gives a measured Gulf response and a reason to investigate local oxygen conditions; it does not supply a catch-per-trip forecast for a charter boat.
Why a productive reef does not disprove a habitat problem
Catch rate depends on both the fish available and their probability of being caught. Habitat compression can increase encounters along an oxygen boundary or in a remaining usable layer. Under that mechanism, a boat may find fish concentrated in a smaller space even though the total amount of suitable habitat has decreased. An improved local catch rate would not, by itself, demonstrate more fish production across the shelf.
The distinction mirrors our artificial-reef case study: concentration, growth, recruitment and net production need different evidence. A reef’s structure is one habitat feature. Oxygen, temperature, prey access and the usable depth range are others. A structure that produces marks on sonar does not independently establish oxygen conditions at the seabed or show that all those marks are the target species.
For a snapper angler, the practical implication is to treat where the fish hold as a three-dimensional habitat question. The relevant comparison may involve fish suspended above bottom rather than a choice between two reef coordinates. Our red-snapper guide and offshore-methods lesson explain depth and presentation. The study adds evidence for why bottom depth and usable fish habitat can diverge, without prescribing a particular depth from an old cruise.
It also sharpens an important release question. The descending-versus-venting study evaluates release evidence; this hypoxia case does not overturn it or establish a new device setting. Release depth, oxygen habitat and the depth at which a fish was caught are related considerations, but no dataset here compares survival after releasing snapper into measured oxygen layers. That would require a dedicated study with actual exposure and fate observations. We should not derive that missing experiment from an oxygen map.
What the history can tell us about the next season
A useful outlook starts with mechanisms and conditions that would change the conclusion. A season with substantial nutrient delivery, strong surface-to-bottom density layering and an extended calm period provides a credible setting for oxygen demand to outrun bottom replenishment. Strong mixing close to survey time can yield a smaller observed footprint than that seasonal setting suggested. If layering returns while oxygen demand remains, the process that supported hypoxia can operate again. These are conditional expectations supported by the physical system and agency interpretation, not a new numerical forecast.
The 2026 event shows why both the spring inputs and the intervening summer history matter. An annual forecast can help describe regional seasonal risk. Repeated oxygen observations can reveal development, movement or disruption over the cruise period. A suitable local depth profile can resolve vertical habitat at the place and time it is measured. Each layer of evidence answers a different question; combining them improves interpretation without making any one of them a guarantee of tomorrow’s fishing.
Wind needs the same care. Our wind-versus-tide analysis concerns water-level departures at a coastal gauge. A wind-driven water-level change is not a measurement of oxygen replenishment. Mixing depends on wind and wave energy acting on the actual water column, including its depth and density structure. No wind-speed cutoff for “the dead zone is gone” is estimated here.
For future updates, the most useful addition would be repeated, quality-documented oxygen profiles with temperature and salinity at comparable shelf locations, linked to weather and fish observations taken at the same places and times. That would let us distinguish a thinning low-oxygen layer from a horizontal boundary shift and test whether fish changed depth, left the station or experienced different prey conditions. A longer, consistently defined forecast archive would also support a stronger evaluation than the six-year window used here.
A small year and a long-term recovery are different findings
The policy target concerns a five-year average below roughly 1,900 square miles, not a single annual survey crossing that line. NOAA’s 2026 release reports a five-year average of 3,745 square miles—still almost twice the target—even though that year’s survey measurement is smaller than 1,900.
Our arithmetic from the rounded annual news-release values produces a 2022–2026 mean of 3,754.4 square miles, compared with 4,754.8 for 2021–2025. The latest mean therefore differs from the agency’s reported mean by 9.4 square miles. We show both rather than alter one annual value to make them agree. Revisions or a different underlying series could explain the discrepancy, but we did not establish its source. Neither number changes the conclusion that the five-year mean remains well above the target.
The rolling comparison also explains why an apparent improvement needs context. The 1,000.4-square-mile decline in our five-year mean occurs when the 6,334-square-mile observation from 2021 leaves the window and the 1,332-square-mile observation from 2026 enters it. The intervening four observations are the same in both averages. That arithmetic does not mean every one of five successive summers improved, and it cannot assign the decline to nutrient-management actions alone.
There is a real reason to track the long-term series: sustained reductions in nutrient-driven oxygen demand are a different achievement from a storm temporarily shrinking the survey footprint. The 2026 measurement is useful evidence about conditions during that cruise. Establishing durable recovery requires the broader nutrient, oxygen and multi-year record.
Final conclusion
The Louisiana shelf dead zone is best understood as a changing loss of usable oxygen habitat, with consequences for where fish and prey can occur and how they interact. The published fish study documents catch-community and diet associations; the historical oxygen records show a patchy measured footprint; the annual forecast comparison demonstrates how strongly the realized survey result can diverge from a seasonal expectation.
Together, the evidence rejects two shortcuts: a catch does not demonstrate that hypoxia has no effect, and one small annual footprint does not establish a solved nutrient problem. The positive conclusion is more useful. Seasonal inputs identify the setting in which hypoxia can develop, mixing and stratification help explain its evolution, and properly located oxygen observations narrow the habitat question. That framework supports a reasoned outlook while leaving the exact future reef conditions and catch rate to evidence we do not yet have.
Data, methods and review status
This is an original descriptive analysis of published NOAA annual values and public oxygen observations, combined with an attributed synthesis of measured fish research. It is not an experiment conducted by this site. The annual comparison uses 2021–2026 NOAA ensemble forecast sheets and subsequent releases. The oxygen analysis uses the complete 2019 cruise-332 bottom CSV, independently checked against the replacement public service; it does not join those records to the 2006–2008 fish trawls.
We checked the fish paper’s methods, results, discussion and relevant numerical tables, including visual inspection of effort, breakpoint, diet-test and sample-count passages. The accepted manuscript contains an unresolved data-archive placeholder. We did not acquire its raw fish histories or refit its statistical models. The public bottom-oxygen service provides no per-record sensor-quality flag in this extract; our checks verify completeness, valid values and consistency between two published representations, rather than independently calibrating the instrument.
The analysis protocol, source ledger, original observations, derived tables and offline reproduction code are downloadable below. Missing values remain missing, class boundaries use unrounded measurements, and the source-reported forecast intervals remain attached to their years. Independent fisheries/statistical review is pending. Recheck this dated case when NOAA revises annual areas, the historical observation archive, model products or the current data service.
