A territory-wide survey of BVI salt ponds was conducted in 1995, and a subset of ten ponds, distributed across 5 islands, was selected for monitoring. Salinity and depth were recorded monthly from January through December of 1995. Logistical difficulties prevented two sampling visits to Anegada (June & December; four ponds) and one visit to Guana (December; one pond). Additional salinity measurements were recorded intermittently between 1991 and 2001. Mean pond salinities compared in this paper were calculated from the 1995 data exclusively, because they reflect a full annual cycle of salinity fluctuations in each pond.
Salinity changes during and after hurricanes were monitored intensively (two to three samples weekly) in one pond in 1995 and in four ponds in 1998. Subsequent sampling, performed between 1998 and 2001, focussed on describing hydrological variation among ponds.
Physical characteristics, including pond area, surface connection with the sea, composition of bottom sediments, period of inundation and watershed area were described in 17 ponds (Table 1; Figure 2). Tidal water movements were measured in all but four ponds, three of which were dry for the majority of each year.
Each pond was mapped and measured (area, perimeter and nearest distance to the sea) using the British Virgin Islands National GIS database (2001–2002) provided by the BVI Conservation and Fisheries Department. These dimensions included only the regularly inundated, non-vegetated portion of each pond.
Watersheds were measured by tracing over a GIS contour map the area of hillside that drained into each pond. Pond areas were included in watershed dimensions. Contours for Anegada and Norman Island were not available. The watershed areas for four ponds on Anegada, which is flat (maximum elevation 9 m), were approximated as 1 1/2 times the area of each pond, an estimate based on the overall geography of the island. The watershed area of NOR was not estimated. Watershed area was used as a relative measure of the potential effect of rainwater entering each pond.
Observations of geomorphology, surface connection with the sea, berm structure, sediments, and surrounding land use were noted while walking the perimeter of each pond.
Pond inundation (presence or absence of standing water) was noted during site visits and at any other time when such observations were possible (e.g. travelling within view of a pond). Inundation periods are reported as the maximum number of months per year that each pond was continuously inundated. Water depth was measured using a weighted measuring tape at permanent sampling stations at each of 10 ponds during 1995 only.
Tidal influence on water level in ponds was determined by tracking the position of the water's edge and converting this lateral movement to vertical movement by triangulation from the shore slope (the shoreline tracking method, detailed in ). Because pond shores were nearly flat, this method was highly sensitive to water movements. A water level drop of only 1 cm, for example, would expose half a meter of shoreline. Wind and resulting small waves introduced a standard error of ± 0.026 cm depth, based on measurements from 10 different days in a non-tidal pond (JOS). The maximum change caused by a moderate wind (approximately 20 knots) blowing towards the sampling area corresponded with a 0.14 cm rise in water level. To avoid classifying a pond as tidally influenced in error, we used a 0.2 cm increase in water level as the minimum threshold for positive determination of tide-driven water movement in ponds (tidal influence). Water level measurements were conducted on clear days with wind speeds of 20 knots or less. Measurements immediately following rain showers were excluded from this analysis. Declining water levels were not used to ascertain the presence of tidal influence because ebbing water was not distinguishable from evaporative water loss.
For ponds that did not initially show tidal fluctuations, water level measurements were repeated at least three times, including during at least one tidal cycle in which the sea rose higher than 25 cm above mean low water (m.l.w). The mean level of high tide during a typical year (1998) was 20 cm above m.l.w .
The timing and amplitude of sea tides were determined by Walker's (1992) DOS program for worldwide tide predictions , using St. Thomas, U.S. Virgin Islands, as the closest reference point. No time correction was used, as the reference point was within 50 km of all ponds except the Anegada ponds, which were within 90 km of St. Thomas. Sea level was monitored during February, 1998, using a graduated stake in shallow water at Bluff Bay, Beef Island. The observed timing of high and low tides was approximately equivalent to those predicted by Walker's program. Mean sea levels shown in this paper were calculated as consecutive 5-day means of the difference between m.l.w. and the daily high and low tides.
Water salinity was measured in situ using a hand-held refractometer with automatic temperature compensation (Vista, model A366ATC). Samples having salinities beyond the range of the instrument (100 ppt) were proportionally diluted with distilled water before the final measurement. Instrument resolution was 1 ppt and error was ± 2 ppt; however, when dilution was necessary, error increased to a maximum of 5% of water salinity.
Salinity stratification was assessed in BAN on 13 August (16:30 and 20:45) and 14 August (7:00 and 11:50), during dry weather. It was also assessed in GUA after rainfall between 29 July and 1 August, 2000, (6 samples) and between 19 August and 20 August (5 samples).
Groundwater salinity was monitored monthly from January through March of 1998 in 4 constructed groundwater wells (1.0 to 1.3 m deep) approximately 30 cm above the vegetation line around the perimeter of BLU. Another 4 wells (0.7 to 0.9 m deep, 0.5 to 10 m landward of vegetation line) were sampled at FLA on 9 August, 1998.
Precipitation data were supplied by local residents at two locations. Other long-term rainfall data were not available in the BVI.
A drought period between 14 March and 6 May, 1995, during which no rainfall was recorded anywhere in the BVI, provided an opportunity for evaluating long-term evaporation rates in salt ponds. The drought-period change in pond depth and salinity in BAN was used to calculate long term evaporation. BAN was chosen for this analysis because it was completely isolated from seawater influence (including tidal effects) and it retained water throughout the dry season. The resulting evaporation model assumed that the quantity of water lost or gained through the ground was insignificant when compared with evaporative water loss. In fact, pond levels monitored by the shoreline tracking method (described above for detection of tidal influence) were constant at night, indicating that through-ground drainage was minimal at best. Evaporation was expressed as the volume of water lost per day per hectare of pond, as these units allowed for the comparison of volume losses between ponds of different sizes.
A second method for measuring evaporation assessed small changes in water level over a number of hours using the shoreline tracking method. Only data from ponds in which no tidal influence was detected, BAN and JOS, were used for these calculations of short-term evaporation.