Inland Flood Hazard Mapping for Antigua and Barbuda:
Summary Report

OAS/USAID Post-Georges Disaster Mitigation Project

Implemented by the Organization of American States,
Unit for Sustainable Development and Environment
for USAID-Jamaica/Caribbean Regional Program

This report was prepared by Vincent Cooper, University of the West Indies, Trinidad.

Note: The final maps that accompany this report are available online: Antigua | Barbuda

1. Introduction

This report describes the approach to mapping the natural hazard of inland floods on Antigua and Barbuda. It first presents a summary of the main cause of flooding on the islands. This is due not only to extreme rainfall, which provides the potential water for inundation, but also depends on the condition of the watersheds at the time of the event. The detrimental effects of flooding are then presented, which on the islands are mainly health hazards in terms of drowning and the contraction of diseases, and damage to economic systems including public infrastructure and private property. The data needs and adequacy for flood quantification are then discussed. For its quantification three sets of data are required, and they include rainfall data for characterizing the rainfall with potential to cause floods, the physical properties of the watershed which determine the volume and time-distribution of runoff onto low-lying areas, and the geometric features of the drainage channel which may or may not have sufficient capacity to accommodate the runoff from the watersheds. The two approaches used for flood mapping are described, followed by the presentation of the methodology implemented for producing the maps. A general summary of the findings for each island concludes this report. Among other things, it suggests what are necessary to improve the accuracy of the maps, which given the limited data, must be regarded as only preliminary maps.

The nature of inland flooding being investigated on Antigua and Barbuda is described by the following definition: Flooding is a general and temporary condition of partial or complete inundation of normally dry land areas from overflow of inland waters or from the usual and rapid runoff of surface waters from rainfall. In these islands, the major trigger of this inundation is extreme rainfall on the watersheds. These extreme events are frequently associated with hurricanes that produce significant quantities of rainfall during their passage over the islands, and the slower the movement over the islands, the greater the quantity of rainfall. The most recent example the relation between rate of movement with rainfall amount was Hurricane Lenny in 1999, which hovered around these islands for about three days and dumped a phenomenal amount of rainfall on them. Widespread flooding may also occur from rainfall associated with the Intertropical Convergence Zone (ITCZ) weather system.

Whether or not floods occur within a watershed, on the occurrence of an extreme rainfall event, depends on the characteristics of the watershed. Most watersheds in Antigua and Barbuda have the following properties:

1. Soils that have generally low permeability or/and shallow depths, and therefore there is no significant reduction in the amount of rainfall available for runoff;
2. Dense drainage networks (mainly on Antigua) that permit rapid flow within the channel banks and rapid concentration of runoff onto the floodplains;
3. Expanding urbanization, which replaces pervious areas with impervious surfaces such as roofs and roadways; and
4. widespread grazing by goats and cattle (mainly on Antigua) that compacts soils and reduces infiltration rates, and that also removes vegetative cover leaving the soil surface exposed to full impact of the rainfall drops and vulnerable to soil erosion.

The properties therefore provide little attenuation of the rainfall amount or little retardation of the travel of the flood volume toward the low-lying areas. Thus watersheds in Antigua and Barbuda are susceptible to flooding. Fortunately, watersheds are small, in the order of tens of square kilometers, and so the volume of runoff would be relatively small even for very extreme rainfall.

Floods may have detrimental effects on three aspects of the human condition—health, agriculture, and economic systems. In terms of human health, floods can cause death by drowning, impact or exposure. In Antigua and, to a lesser extent, Barbuda, rising floodwaters from overbank flows poses hazard of drowning of human beings as the floodplains are densely populated. Numerous physical injuries are possible as well, and a common effect of flooding is disease and disease transmission. Diseases commonly are associated with disruption of fresh water supply; contact with floodwaters contaminated by septic tank and wastewater treatment plant overflows; and the creation of appropriate habitat conditions for certain rodents, insects and organisms that transmit diseases. Detrimental effects on agriculture are mainly due to loss of livestock—goats, sheep and cattle—due to drowning or becoming dispersed. Destruction of crops, seeds, and stored food stocks during inundation are limited as the country has a small agriculture base. Public infrastructure and private property may be damaged with inundation. Initially, ground floors of buildings and residences may be affected with rising flood stage. But with increasing stage and flow velocities, buildings and loose property can be swept away. The housing developments within Cook Creek, as indeed on other floodplains, are particularly vulnerable as ground floors are generally not elevated more than about one metre (about three feet), above ground level, nor are these buildings properly anchored to their foundation. Some houses floated away due to the uplift pressure of water flooding the housing developments during Hurricane Lenny. Floating debris such as cars and logs can increase damage as they impact structures downstream. Properties on riverbanks may be in danger if the toe of these banks begins to erode and trigger bank failure. Major disruptions of the road network may affect productivity as the workforce is delayed in reaching their workplaces.

Areas prone to flooding have the following characteristics:

• very gentle slopes that are perhaps poorly drained because their streams, creeks and drainage channels also have very gentle slopes;
• constricted natural drainage owing to, for example, heavily vegetated areas such as mangrove forests, meandering streams, or tidal influences. If communities are located around the outlet, then the hydraulic structures, such as bridges and culverts, may constrict flows owing to poor maintenance or because they become blocked with debris, and thus may further limit the natural carrying capacity of the rivers;
• large watershed areas (relative to their area) contributing runoff from rainfall;
• so positioned within the watershed that runoff from the surrounding lands concentrate at the floodplain;
• contributing areas having large runoff coefficients due to impervious soils, impervious ground cover, or/and limited abstraction losses by vegetative cover.

Data needs for flood quantification can be grouped into three categories: data for describing the meteorological event which may trigger the flood; data about the watershed which are used for determining the fraction of the rainfall that would runoff the land and how it would be distributed in time; and data for describing the geometry of the drainage channels which may or may not be able to convey all of the runoff within its banks, depending on their water carrying capacity.

The Critical Storm: The definition of a critical storm is somewhat subjective and its description may vary depending on the purpose of the investigation. In all cases it is defined by four features:

1. rainfall depth;
2. duration;
3. the time-distribution of the rainfall depth over the duration of the storm; and
4. the spatial distribution or the area over which the features in (i) to (iii) remain unchanged.

Historical daily rainfall records collected within the watersheds are vital to obtaining the rainfall depth. From the application of appropriate statistical analyses on these records, the depth of rainfall of certain return periods (see definition below) can be found. Daily rainfall records are normally available in plenty supply but more often than not there is little information to describe how that depth of rainfall occurs over the 24-hour period. The resulting runoff is very much dependent on this time distribution, and it is not sufficient to average the rainfall over the 24 hour period. Standard time distribution curves have been developed for the United States and some of these curves have been successfully used in the Caribbean (and indeed in other parts of the world). In the absence of time-varying rainfall data collected from within the watersheds, then these curves can be used. The storm duration is related to a characteristic of the watershed known as the time of concentration. The time of concentration, T_c, depends on the physical characteristics of the watershed, and it represents the time required for water from the furthest point in the watershed to reach the place where the hydrograph is being measured. For maximum runoff generation, the critical storm must have a duration at least equal to the T_c of the watershed. The variation in space also must be considered in developing the critical storm. The amount and timing of rainfall vary considerably from one place to the other for the same storm. Therefore, the capturing of such variability requires finding a weighted average of rainfall measured at several raingauges set up throughout the watershed. The use of one raingauge may not be truly representative of the rainfall over an entire area, but in many cases, only one gauge is available.

For Antigua, six years of daily rainfall from a single station at Jolly Hill was used to produce the critical storm. At this station the rainfall recorded from Hurricane Lenny was 460 mm (about 18 inches) over three days, 242 mm (about 9 inches) occurring within a 24-hour period. This daily rainfall was chosen as the critical storm and the Type III time distribution curve—a designation used to describe one of the standard curves in which most of the rainfall occurs within a small fraction of the 24 hour period—was used to distribute the rainfall in time. It was also assumed that this critical storm was spatially uniform and would have had the same rainfall depth and time distribution everywhere on the island.

Runoff: Having defined the critical storms, it would then be possible to estimate runoff properties from the surrounding watersheds. The amount of water available as runoff from the rainfall and the distribution of this runoff in time are controlled by the features of the watershed. The important features are:

• Size of the watershed;
• Soil infiltration rates and depth of storage;
• Land cover;
• The total length of all the rivers and their tributaries;
• Shape, slope and ruggedness of the watersheds.

The first three features control the amount of water available as runoff onto the plains. The other features control the timing of the hydrograph peak. Most of the required data are available on standard maps. Most are natural conditions, but the land cover and, to some extent, the length and characteristics of the rivers and tributaries can be altered depending on the human activities occurring within the watersheds.

All required data can be extracted from standard maps and land capability reports, and these were all readily available, some in digital format.

Geometric Features of the Drainage Channels The geometric properties of the drainage channels and the nature of their lining determine the flow rate that can be accommodated within their banks. Meanders on channels reduce their capacity, so too would reduced slopes. Capacity is also reduced with increasing roughness of the channels. Carrying capacity is affected by man-made hydraulic structures, such as bridges, weirs, sluice gates culverts, that may be constructed within the channels.

Normally, detailed surveys are required to gather the geometric data for describing the drainage channels. Some surveys were carried out, but this was limited, and only within Cook Creek.

Two approaches to flood mapping, differing mainly in the degree of accuracy in mapping floodwater levels, were applied. The first approach was done at the floodplain scale and the results, though still approximate for lack of sufficient data, provide information on flood levels within the plain. This traditional approach required hydrological inputs and detailed hydraulic data. It is concerned with bank overflow as it considers the carrying capacity of the drainage facilities on the plains and the extent to which water from surrounding watersheds remains as "within-bank" flow. For the results to be meaningful careful analysis on the hydraulics of the floodplain is needed and the analyst must take into account all the significant features within the plain that could affect water levels.

If all required information is available, and most importantly if floodwater levels have been collected from actual flood events, then the hydraulic analysis can be fine-tuned by making adjustments to uncertain inputs until the water levels from the analysis match the levels from the recorded flood events. Typically, the accuracy is within 150 mm (6 inches).

The second approach was applied because of lack of detailed information about the channel geometry. Whereas the same runoff information as above is needed to quantify the amount of water within the floodplain, this approach does not attempt any hydraulic analyses. Instead, the method estimates a flood depth over the entire floodplain, which is essentially an average value derived from a simple division of the total volume of water from all runoff flowing into the plain over the surficial area of the plain. This depth does not provide any information about how water depths vary within the plain as it assumes that the plain is flat throughout. Furthermore, this value may overestimate actual flood levels in some places, and underestimate them in other places. Such approaches may be necessary when the type of information required in the first approach is missing or unavailable, but the approach is very useful for providing basic ranking of the extent to which floodplains over the island is susceptible. Its approximations must be well understood so that it can be properly used. Most of the requirements can be extracted from standard topographical maps.

Islandwide Scale Mapping:

1. Identify areas, called floodplains, having mild slopes. This can be done either via manual inspection of topographical maps or with the aid of digital elevation models (DEMs) and computer software;
2. for each plain, demarcate the watershed draining into the plain and then determine the ratio between the watershed area and the floodplain area (from (i) above);
3. for a particular 24-hour return period rainfall, generate runoff hydrographs discharging into the plain, based on the shape, the drainage network, the soil properties and the landuse;
4. from the hydrographs, determine the volume flowing into the plain and estimate an average discharge rate over the hydrograph period;
5. estimate a discharge rate from the plains and use it to determine the maximum volume of water (expressed as depth of water) that is likely to be stored on the plain;
6. determine the hazard category of each plain according to water depth within the plain from the 100 year return period storm (roughly equivalent to the maximum 24-hour rainfall period within Hurricane Lenny), using the following classification:

The categorization of the hazard in this way provides a measure of the magnitude of flooding to be expected from such an event. It also provides a measure, certainly by rank if not in absolute terms, of the severity of flooding among the identified zones.

• Perform statistical analyses on the daily rainfall data and obtain amounts for various return periods.
• Derive the critical storm for a 24-hour duration, using standard time curves, and assuming that the rainfall was the same throughout the watershed;
• Obtain contour maps of Antigua and demarcate the watersheds draining into the floodplain, and conduct field trips to confirm the boundaries;
• Determine the existing land use within the watersheds from topographical maps and field visits;
• Obtain soils maps of Antigua and Barbuda;
• Digitize information from the map and field visit
• Conduct interviews with residents within the floodplains to determine specific information about recent floods—the highest water level reached, the direction from which the flood waters came, the time it reached its highest level, the length of time it stayed at that level, the time it took to subside to some remembered level.
• Survey the river or drain from which the reported flooding originated, paying attention to (i) any natural constrictions such as a narrowing of the river cross-section, or manmade constrictions due to hydraulic structures such as culverts and bridges, or to construction of property within the waterway; (ii) the free clearance of hydraulic structures and bends; (iii) the lining of the channels; (iv) the level of maintenance of the channels; (v) the nature of the floodplain from the river bank in terms of its coverage—pastures, wooded, urbanized;
• Identify critical river sections based on their impact on flooding with potential for significant losses;
• Conduct detailed surveys of these areas at critical cross-sections, hydraulic structures and any other important features.
• Examine the discharge point of the section to determine outlet control conditions—whether, for example flow was unhindered or under the influence of tides.
• Perform detailed hydraulic analyses using appropriate computer software.
• Produce a flood map showing variation of water depths within the floodplain.

8. Summary Results

For Antigua, extensive flooding is confined to about four regions mainly within the north where the majority of the population lives. Two of these regions are of critical importance, the floodplain of Cook Creek Watershed, which is densely populated with basic housing, and the Fitches Creek floodplain, also densely populated and containing important road links to the international airport. The other two areas at Potter Village and Parham Village also have significant population and are vital to the island road network. Many of the problems are owed to the physical characteristics of the watersheds and their floodplains and the inadequacy of the natural drainage channels. Poor infrastructure design, construction and infrequent maintenance serve to exacerbate the flooding. Additionally, there are pressures for developing the watersheds for housing that, if unchecked, have potential for increased runoff and more severe flooding in these areas.

Apparently there is a well-developed raingauge network collecting daily rainfall over the island. The database, however, is not in a form that makes for ready retrieval of the data and was not available for the project. The control points for establishing coordinates on the national grid are not easy to locate. The negatives of several maps, including the 1 in 5000 series, suffered water damage during a hurricane thus precluding extraction of fairly detailed topographical mapping of the floodplains. Information on the irrigation network, including drainage cross-sections and details of the hydraulic structures may be available, but could not be readily accessed.

The map was prepared using baseline information on other topographical features prepared for the PGDM project. The difficulties in incorporating surveyed data into the project base DEM are mentioned in the technical report.

The implementation of the methodologies, described above, was constrained by time and therefore, at best, the resulting maps should not be considered more than preliminary. Still, with sufficient time, the methodologies can be followed during subsequent map improvement exercises.

The acquisition of a longer dataset of rainfall records and detailed surveys within the floodplains is required for improving the accuracy of the map. The models used for conducting the study are readily available and there is adequate technical support for their proper use.

For Barbuda, runoff is generally rapid owing to its very thin soil depths and outcrop rock surfaces that provide very little infiltration and relatively low flow retardation. Its watersheds are primarily undeveloped with scrubby vegetation covering most of the area. No rivers, and not many gullies exist and therefore runoff is mainly as sheet flow. Runoff only accumulates in depressions; otherwise it may flow through Codrington, the main town, on its way to be discharged in the salt ponds and the sea. The roads appear to function as drains and just north of Codrington where several roads merge, runoff is concentrated there from a catchment area of about 15 km². This was the area noted to have suffered worst from Hurricane Lenny. It contains several abandoned quarry pits whose limestone lining has since become blocked with a thin impermeable soil layer. Water collected in these pits from runoff would remain impounded for several days to weeks owing to slow infiltration rates of the lining. Drainage infrastructure is minimal consisting of small box drains and small circular culverts through Codrington.

One rainfall gauge exists on the island, thanks mainly to the interest of the current school principal. There are seven years of daily rainfall, the time since the raingauge has been established, and records are readily available in electronic format. The database should further be developed so that data can be provided in universal ASCII format. Detailed topographical maps at resolutions better than 1 in 10000 do not seem to be available, or if they exist, they are not readily accessible.

The implementation of the methodology, described above, also was constrained by time and therefore, at best, the resulting maps should not be considered more than preliminary. Detailed flood mapping done for Antigua may not be appropriate for Barbuda as it has no pronounced rivers.

The map was prepared using baseline information on other topographical features prepared for the PGDM project. Field surveys within the demarcated flood zones are required to more accurately define the flood zone and perhaps to qualify flooding according to the desired guidelines of the project.

The proper maintenance of the raingauge and the continued collection, processing and analysis of rainfall records are necessary for upgrading the map. If not done as yet, then the arrangement of rainfall data collection should be formalized with the Meteorological Office in Antigua, or with the Barbuda Council.

Definitions

Flood: A general and temporary condition of partial or complete inundation of normally dry land areas from overflow of inland waters or from the usual and rapid runoff of surface waters from rainfall.

Hydrograph: A flood wave passing a particular place in the river. It represents how the stage (water elevation) or the flow in a river changes with time.

Return Period: This is an expression of the probability of a particular rainfall depth of a given duration occurring. The rainfall depth corresponding to a specified return period will be equaled or exceeded at a probability equal to the inverse of the return period, that is the relation can be expressed as follows: Return Period = 1 / (Probability of occurrence). For a 10 year return period: So the probability that the rainfall depth for a 10-year return period will be equaled or exceeded is 0.1 (1/10)

Critical Storm: The rainfall input used for producing the hydrograph. It represents what is viewed by the analyst to be most important in predicting the effects of the rainfall on runoff. The rainfall depth, the duration of the storm, the type of distribution of the rainfall depth in time and the uniformity of the distribution of the storm over the watershed must all be specified for complete specification of the critical storm.

Time of Concentration: The time elapsed between the start of the rain and the runoff from the most remote point in the watershed to reach the place where the hydrograph is being measured.