The record of mangrove restoration is uneven. Lewis (2005) argued that most failures stem from planting without first asking why mangroves are absent, and in particular from placing seedlings at elevations flooded too long for them to survive. Primavera and Esteban (2008) documented the same pattern in the Philippines, where planting was repeatedly directed to low seafront flats that were convenient and uncontested but unsuitable. An assessment in Sri Lanka found that a large share of planting projects there had left few or no surviving trees (Kodikara et al., 2017). On eroding mud coasts the difficulty is physical as well: where the sediment balance is negative, seedlings are undermined before they establish, and Winterwerp et al. (2013) set out the conditions of sediment supply and wave energy that must be restored first. Any assessment of shoreline effect must therefore begin by establishing whether a stand exists at all.

A shoreline is not a single line, and Boak and Turner (2005) reviewed the many indicators in use, from the instantaneous waterline to the vegetation edge and the high-water mark. On a muddy, low-gradient coast with a large tidal range the waterline can shift hundreds of metres between low and high tide, so that the choice of indicator and the control of tidal stage dominate the error. Water can be separated from land in multispectral imagery with a normalised difference water index (McFeeters, 1996) or its modified form based on a shortwave infrared band (Xu, 2006). Cloud platforms now hold the full Landsat and Sentinel-2 archives and allow them to be processed without local computing power (Gorelick et al., 2017), and open toolkits automate shoreline extraction from those archives (Vos et al., 2019). Rates of change are conventionally computed along shore-normal transects with the Digital Shoreline Analysis System (Thieler et al., 2009).

The village frontage is divided into segments of similar exposure and the planted segments are marked from the records of the planting organisation and the memory of those who planted: when, with which species, at what spacing, and with what replanting since. Reference segments are chosen on the same stretch of coast, facing the same direction, with similar foreshore width before planting and no hard defence built during the study period. If no such segment exists, the protocol can still describe change at the planted site but cannot attribute it, and this should be recorded at the outset.

For each year of the archive, cloud-free images of the dry season are selected, since the wet season brings cloud and turbid floodwater. Two indicators are extracted. The waterline is obtained by thresholding the modified water index, using only images acquired near the same tidal stage as judged from tide tables for the nearest port. The seaward vegetation edge is obtained from a vegetation index and is far less sensitive to the tide; where a mangrove fringe or grassed bank exists it is the preferred indicator. The pixel size of the imagery sets a floor on what can be detected, thirty metres for Landsat and ten for Sentinel-2, so that only movements of several pixels can be treated as real.

Transects are cast at regular spacing perpendicular to a baseline drawn landward of all shorelines. Along each, the net movement between the earliest and latest shoreline and the rate from a linear regression through all dates are calculated, separately for the period before planting and the period since. The quantity of interest is not the rate after planting but the change in rate at planted transects relative to the change at reference transects over the same years.

Fixed plots are laid out across the planted zone from its landward to its seaward edge and marked with posts. In each, trained village monitors count living and dead stems and measure the height of a fixed number of tagged trees twice a year. Natural recruits are counted separately from planted stems, because colonisation by wild seedlings is a sign that the site suits mangroves. Signs of grazing, cutting and barnacle fouling are noted.

Whether sediment is building or being stripped is measured directly. Cross-shore profiles are levelled from a fixed benchmark on the embankment along permanent lines through planted and reference segments, after the monsoon and again late in the dry season. Sediment stakes within the plots give a simple local record of surface change; where a partner can install them, rod surface elevation tables provide the precise standard (Cahoon et al., 2002), and their use in mangroves is reviewed by Krauss et al. (2014). Where the shore ends in an erosion scarp, the distance from marker posts to the scarp edge is taped each season.

Residents hold a record no satellite has. In facilitated sessions, groups of older fishers, women who collect on the foreshore and embankment labourers each draw the position of the bank at remembered dates, such as a named cyclone or the building of a sluice, on a printed image of the village. They also list events that bear on interpretation: breaches, repairs, dredging, the opening of shrimp ponds. The maps are compared with the satellite shorelines, each serving as a check on the other.

The evidence is brought together in a meeting of all parties. The report states, in plain language and with maps, how the shoreline moved at planted and reference frontages before and after planting, whether a stand has established, whether the bed in front of it is rising or falling, and what other events could explain the pattern. It grades its own conclusion as supported, possible or not shown.