The 39-Hectare Answer

Roughly 1,200 hectares of abandoned fishponds went into a feasibility assessment. Thirty-nine hectares came out of it as ecologically viable for mangrove restoration. About three per cent.

Presented badly, that reads as a failed assessment. It is the opposite. Four screens were applied — hydrology, species and habitat matching, soil and salinity, and a historical land-cover analysis of what the site had been before it was diked. Thirty-nine hectares is what survived all of them.

What this piece publishes is the order those tests should run in, and the reason the order matters. The claim is deliberately narrow: not that mangrove restoration delivers — that is a question for monitoring reports, not an article — but that failure in mangrove restoration is predictable, that nearly every documented failure mode is observable from the ground before capital commits, and that the observation is cheap relative to the failure. What almost no pre-investment process budgets for is the walk.

The desk answer is always bigger than the ground answer

Remote sensing is very good at extent and land cover and structurally weak at the two things that decide a mangrove restoration site: elevation and connectivity.

Take elevation. The window in which mangroves establish and grow is narrow. Lewis (2005) reports typical surveyed elevations for mangrove species in Florida in the range of +30 to +60 cm above mean sea level — a sloped platform a few decimetres deep in the vertical. SRTM, still the backbone of a great many desk screens, was specified to an absolute vertical accuracy of 16 metres at 90 per cent confidence (USGS EROS) — more than fifty times the width of the band the trees occupy. Newer global models are better and still not close. No freely available DEM tells you whether a given pond empties on a spring low water.

Connectivity is worse, because it turns on a feature — a channel, a sluice, a collapsed gate — often a metre wide, seasonally vegetated, and only diagnostic at the bottom of the tide. Imagery is acquired at whatever tidal stage the orbit delivers, and at high water almost every pond looks connected.

Historical land-cover analysis is genuinely useful and answers a different question. It tells you a pond was mangrove. It does not tell you the pond can be mangrove again. Between those two statements sit the dikes, a metre and a half of excavated soil, and twenty years of altered drainage.

None of which is an argument against desk work. Global restoration-potential mapping — Worthington and Spalding (2018) being the reference case — is the right instrument for portfolio prioritisation and the wrong one for a decision on a single tenure block. The failure is in treating the first as the second.

What happens when that step is skipped is one of the better-documented patterns in restoration ecology. Reviewing eight Philippine mangrove rehabilitation initiatives, Primavera and Esteban (2008) put long-term survival at 10–20 per cent and traced it to wrong species and wrong site: Rhizophora into sandy substrates of exposed coastlines, and planting pushed into lower intertidal and subtidal zones because the middle and upper intertidal, where mangroves actually grow, had already been converted to ponds — and ponds have owners, whereas the seafront is open access. The seafront was planted because it was available, not because it was suitable.

A field re-survey of 23 Sri Lankan sites by Kodikara and colleagues (2017) found no living mangroves at all in nine of them five or more years after planting, and tidal influence negligible at eight; of 1,000–1,200 hectares planted since 2004, 200–220 hectares remained forested eight to ten years later. Sites where technical guidance was used to direct planting averaged 46 per cent survival. Sites where it was not averaged zero.

Those are not random outcomes. They are the predictable consequence of one omission, and they are what a hydrology-first screen exists to catch.

The screen, as an ordered test

Order is the whole design. Each test costs more than the one before it, so each is placed where it can stop the spend: a site that fails Test 1 never incurs the cost of Test 5. Run the same seven criteria as a parallel checklist and you pay for all of them on every hectare, which is exactly why pre-investment screening gets cut.

Test	What it answers	Fails when
1. Tidal connectivity at low water	Does water leave the pond?	Pond is perched or the outlet is silted shut
2. Inundation frequency and duration	Does the floor sit in the same tidal band as a healthy reference stand?	Floor is below the reference band — a lagoon, not a forest site
3. Dike-breach feasibility	Can the hydrology be restored, where, with what plant, and where does the spoil go?	No viable breach location, no access, or spoil has nowhere to go
4. Distance to propagule source	Will the site recruit itself once it floods?	Nearest reproductive stand is too far, wrong species, or off the inflow path
5. Soil salinity and sulfidic layer	Will seedlings survive the substrate, and what does reflooding release?	Hypersaline profile, or oxidised acid sulfate soils near the surface
6. Upstream freshwater input	Is the salinity gradient still intact above the site?	Catchment flow is diverted, embanked or abstracted upstream
7. Competing tenure on the pond	Who can stop the work, and on what instrument?	Live lease, uncancelled record, occupancy, or a producing neighbour on the shared dike

1. Tidal connectivity, measured at low water

The first test is the cheapest and disqualifies the most ground. You go at the bottom of the tide, on a day the tables say the water will be low — in the Philippines that means the NAMRIA Tide and Current Tables and a local benchmark, not a phone app — and watch whether the pond empties.

Ponds hold water for two entirely different reasons: because the tide fills and drains them, or because they cannot drain at all. From above, at the wrong tidal stage, those are identical. On the ground at low water they are unmistakable — a perched pond has a static waterline, no exposed drainage network, and an outlet that is a vegetated depression rather than a channel.

This is also where you locate the reference stand. Lewis (2005) is unambiguous: the single most important factor in designing a successful mangrove restoration project is determining the normal hydrology — depth, duration and frequency of tidal flooding — of existing natural mangrove communities in the area, and the normal surrogate for costly tidal data gathering or modelling is a tidal benchmark and a survey of existing healthy mangroves. The corollary is that you cannot screen a pond in isolation. You need a comparator in the same tidal setting, and you have to walk to it.

2. Inundation frequency, against that reference

Frequency and duration are not the same variable and they decouple badly. The data Lewis reproduces from a single basin forest on Tampa Bay makes the point: the site was flooded just 88 times in two years, yet the total hours of flooding came to 59.4 per cent of the available time, because it trapped tidal water and rainfall together. His comment on it is that this is not the prevailing understanding of mangrove tidal hydrology. Design to frequency alone and you put the species in the wrong place.

The framework is old and still correct: Watson's 1928 inundation classes, as Lewis recounts them, run from Class 1, inundated by all high tides, to Class 5, inundated only by exceptional or equinoctial tides — sorting mangrove zones by the nature of the tide that floods them rather than by the number of inundations or their total duration. That is what turns an elevation survey into a species plan.

Practically: levels on the pond floor tied to a local benchmark and referenced to the tidal datum, a matching transect through the reference stand, and on marginal sites a pressure logger left through a full spring–neap cycle. A fortnight of real water-level data settles arguments no model will.

The output of a good screen is mostly “no”. That is precisely what it is worth paying for.

3. Dike-breach feasibility

Here ecology becomes earthworks and an opinion becomes a budget line. Which dike section, how wide, how deep, machine or hand labour inside a tide-constrained window — and where does the spoil go? Spoil pushed onto the pond floor raises the elevation you just surveyed and can invalidate Test 2.

Breach placement is not a detail. Stevenson, Lewis and Burbridge (1999) examined disused ponds at field sites in Central America and the Philippines — on the owners' condition that the exact locations not be revealed — and describe two neighbouring ponds of the same age. The smaller one, whose outboard dike breached naturally in 1987, recolonised. The larger one did not: natural recolonisation was limited to the central area of the pond, with a reduced species complement, which they attribute to reduced tidal range and limited flooding caused by a restriction in the size of the breach. Their conclusion is blunt. Simply breaching dikes seems to be one solution and is not always so easy, because the location and size of breaches determines whether tidal exchange, and therefore natural recovery, happens at all. An undersized or badly sited breach produces a pond that fills and does not drain — worse than leaving it alone.

The test also has a newer methodological edge. The draft VM0033 version 3.0, out for public consultation from 3 September to 5 October 2026, lists creating, restoring or managing hydrological conditions — removing tidal barriers, improving hydrological connectivity, restoring tidal flow to wetlands — as an eligible project activity, and in the same applicability section rules the methodology out where hydrological connectivity of the project area with adjacent areas leads to a significant increase in greenhouse gas emissions outside the project area. The breach that makes the ecology work can be the thing that makes the accounting fail. That is a determination about somebody else's land, twenty metres away, and there is no remote way to make it.

4. Distance to propagule source

The design intent throughout is natural recruitment, not planting. That is the logic of the five-step sequence Lewis and Marshall set out in 1997 and Lewis (2005) restates: understand the species autecology, understand the normal hydrology, identify what is blocking natural succession, restore the hydrology and let volunteer propagules do the establishment — and plant only after determining through the first four steps that natural recruitment will not meet the goal. Lewis is explicit that failure in most restoration projects can be traced to going straight to the last step.

That makes propagule supply a hard screen, not a soft preference. Where a floating propagule ends up is set by the water it is in and, for propagules that float flat on the surface, by the wind: Van der Stocken and colleagues (2013), releasing propagules on outgoing tides in a Kenyan tidal creek and repeating the work in a flume, found dispersal path and distance governed by the strength and direction of the water currents, with wind's contribution depending on a propagule's buoyancy and how it sits in the water. So the test is specific. Where are the nearest reproductive stands, of which species, and does the inflow path at the proposed breach actually carry floating propagules in? A breach that is hydrologically sound but sits on the wrong side of the prevailing current is biologically dead.

There is reason to hold this line even where planting is available as a fix. From 684 planted stands worldwide across four decades, Bourgeois and colleagues (2024) found biomass carbon culminating at 71–73 per cent of intact stands about twenty years after planting, and soil carbon settling at roughly 75 per cent of intact with no notable change after an increase in the first five years. Planting has real value — on their reading it effectively prevents further carbon loss from land-use change — but it is a slower, costlier route to a lower ceiling than a site that recruits itself. That is a third-party finding about planted stands in general, not a prediction about any site.

5. Soil salinity and the sulfidic layer

Two distinct problems live in the same auger hole.

The first is salt. An impoundment that stops being flushed concentrates it by evaporation, and hypersalinity is a documented cause of dieback rather than merely a stressor — Lewis (2005) records hypersalinity arising from year-to-year variation in rainfall producing natural mangrove die-backs. A pond can pass the connectivity test on a spring tide and still be lethal through a dry season, which is why the profile matters more than a surface reading.

The second is acid sulfate soil, and it is the criterion most often found too late. Excavating and constructing a pond oxidises potential acid sulfate material into actual acid sulfate soil; while the farm is operating, the acidity is being managed; at abandonment that management stops. Stevenson, Lewis and Burbridge record the consequences: the soils become more difficult to treat as they age, recurrent acidification interferes with ecosystem recovery, and reflooding a pond with acid sulfate soils can itself increase small acid events, floc out iron that clogs aquifers and smothers streambeds, and push saline water into areas it did not previously reach. This is the one test where the answer can be that restoring the hydrology makes things worse before better, and where a laboratory is unavoidable. It also rewrites the species list — acid tolerance and salinity tolerance are not the same trait.

6. Upstream freshwater input

Mangrove structure tracks the salinity gradient, and the gradient is maintained from upstream. A road embankment, a culvert, an irrigation offtake or a new abstraction can leave a pond perfectly tidally connected and progressively hypersaline. Lewis (2005) is direct about the mechanism: disruption of the normal freshwater flows that dilute seawater can kill mangroves outright.

This is the test most reliably skipped, for a structural reason: the cause sits outside the project polygon, and nobody bills for walking a catchment above a site they are not restoring. It is still a walk — up the drainage, noting every crossing, embankment and diversion, asking residents what changed and when. Multispectral survey and hydrological overlays show the drainage pattern and the vegetation response; they will not tell you a culvert was blocked three seasons ago.

7. Competing tenure on the pond

Last on the ecological sequence, first to kill a project.

The Philippine statutory position is unusually clear on paper. Under Section 46 of the Fisheries Code (Republic Act 8550, 1998), a Fishpond Lease Agreement runs 25 years, renewable for another 25; the area must be developed and producing on a commercial scale within three years of approval; and areas not fully producing within five years of lease approval “shall automatically revert to the public domain for reforestation”. Section 49 directs the DENR, in coordination with the Department of Agriculture, local government units, other concerned agencies and FARMCs, to determine which abandoned, underdeveloped or underutilised fishponds covered by Fishpond Lease Agreements can be reverted to their original mangrove state, and then to take the steps necessary to restore them. Converting mangroves to fishponds is unlawful outright — originally Section 94, renumbered to Section 99 when Republic Act 10654 (2015) repealed and replaced the Code's prohibitions chapter.

Read literally, an abandoned lease pond has already reverted. In practice the walk finds something else: leases never formally cancelled, lessees holding receipts and expectations, heirs with preemptive rights to the unexpired term under the same section, occupants fishing the pond, an assessor still carrying it on the tax map, and — the recurring one — a producing neighbour whose operation depends on the dike you proposed to breach. Primavera and Esteban made the underlying point years ago: the right elevations for mangroves are where the ponds are, and ponds have owners.

None of this is a carbon-rights determination, which is separate and harder. It is the prior question of whether the physical intervention can lawfully and peacefully happen. What you collect is documentary and social — the lease record and its cancellation status, the tax map, the occupancy picture from being present, and the names of whoever will object. None of it is in a raster.

What the ground catches and the desk cannot

The disqualification patterns repeat, and they are worth naming because a desk screen cannot see any of them.

• A pond that historical land cover correctly identifies as former mangrove, that current imagery shows as open water inside intact dikes, and that turns out to be perched — holding water because it cannot drain, not because the tide fills it.
• An outlet channel legible in imagery that is, on the ground, a silted and vegetated depression with no tidal prism behind it.
• A block where the hydrology is workable and the nearest reproductive stand is too distant, or off the inflow path, for the site to recruit itself.
• A block where the substrate, not the water, is the constraint — and where reflooding would be the trigger rather than the remedy.
• A block where every ecological test clears and a neighbour's producing pond depends on the dike you proposed to breach.

The arithmetic of the assessment at the top of this piece is the general shape of the problem: roughly thirty times as much area went into it as came out of it. The screens applied were hydrology, species and habitat matching, soil and salinity, and historical land-cover analysis. Tenure appears in the table above because it is what stops projects once the ecology clears; it was not one of that assessment's ecological screens and did not set the 39-hectare figure.

The gap between candidate area and confirmed area is not a modelling error to be tuned away. It is the structural difference between asking “was this mangrove?” and asking “will this flood, drain, recruit, and be left alone?”

What the screen costs, and why nobody budgets for it

We will not publish invented invoice figures, but the cost structure explains the omission better than any argument about rigour. The inputs are field days, not capital: two people, a vehicle or a boat, a level and a benchmark, an auger and a field pH kit, a laboratory where sulfidic soils are indicated, a pressure logger left behind on marginal sites. Against a planting programme that later fails, or a validation that stalls on an applicability condition, it is a rounding error.

The binding constraint is the calendar. The informative low waters cluster around spring tides, a few days a fortnight, and if weather closes them you wait for the next set; a spring–neap cycle cannot be compressed. Anyone promising a hydrological screen inside a week is promising a desk study with a site photograph attached.

The commercial reason it gets cut is simpler. The deliverable is mostly a list of what does not work: no hectarage to announce, no planting-day photograph. A process whose main output is “no” is hard to fund and is exactly the process that protects the capital behind it.

The claim, restated

Failure here is not bad luck. Wrong elevation, wrong species, no recruitment path, acid substrate, an undersized breach, a lease nobody cancelled — every one is observable from the ground, in a fixed order, before money is committed. The literature has described these failure modes for decades. What has been missing is the discipline of testing for them in an order that lets the cheap test kill the expensive one.

The direction of methodological travel makes that discipline more valuable, not less. VM0033 remains at version 2.1, active since 4 September 2023; the version 3.0 draft replaces the single consolidated methodology with a modular framework and introduces a new dynamic performance benchmark for afforestation, reforestation and revegetation combined with restoration of wetland ecosystems in tidal wetlands. Whatever survives consultation, the trajectory is toward more auditable justification of what a project changed hydrologically and of the counterfactual it changed it against. A screen documented in order, at the time, with the failures recorded, is evidence for that. One reconstructed afterwards from an imagery archive is not.

The walk is the cheapest part of a restoration project, and the only part that can still say no.