Hurricane response emergency modeling with InfoWorks WS


Courtesy of Innovyze

The island of Jamaica is located in the Caribbean Archipelago, 90 miles south of Cuba, 600 miles south of Florida and 100 miles south-west of Haiti. It has a population of 2.6 million, 800,000 of whom live in the metropolitan area of the capital, Kingston. The island is mountainous, with a distinct east-to-west ridge and a dramatic relief change in the east.

Like the rest of the region, Jamaica suffered the impact of tropical cyclones in 2005. Warmer waters in the Gulf of Mexico have created a series of ‘super hurricanes’ this year, including the devastating Katrina and Rita, which inflicted enormous damage across Louisiana and neighboring states. The Caribbean islands are also in the direct path of the seasonal hurricanes and have, over the years, experienced considerable damage.

Rainfall associated with these events can be as severe as 1ft in 24 hours. Landslides, flooding, wind damage and storm surges are the main threats, and in Jamaica most are major risks.

Hurricane Dennis

Infrastructure – and in particular road and water supply systems – are particularly susceptible to such damage. This possibility was realized when Category Three Hurricane Dennis hit on 7 July 2004, and the raw water pipeline to the Sea View water treatment plant was destroyed. It was estimated that repairs to the supply for some 35,000 people in the high-value area that depended on this source would take up to four weeks. Since this line was the only supply line to the treatment plant, an interim supply was urgently needed.

Alternative solutions that were available included diverting supplies from adjacent supply areas, establishing temporary pumping facilities to augment supplies from the main network, and trucking water to the areas that could not be reached in any other way.

It was agreed that a network model of the area should be rapidly deployed to assist in the operational steps and decision-making. Unfortunately, the network in question did not have an existing hydraulic model, which considerably exacerbated an already difficult situation.

Urgent task

Fluid Systems Engineering in Jamaica was commissioned by the National Water Commission to create the emergency model using Wallingford Software’s InfoWorks WS solution, which it had recently purchased. Model designer Maurice Jones emphasized the challenge: “There was no data and the results were needed yesterday.” In fact, the company had just four days in which to develop a working model before a key meeting.

The base mapping was undertaken using a digital aerial photo image of 1 ft pixel resolution and ‘x-y’ RMS error of 6ft, auto-rectified to the national grid. Scanned images of a 1:50.000 topographic map referenced to the national grid were made available.

The data required compression by image management solution GeoExpress with MrSID®, which enables fast distribution while preserving the fine details of the original imagery. It also links seamlessly with InfoWorks WS.

The digital surface model collected from side-scanning Lidar imagery (topographic data collected by special ground-scanning radar attached to an aircraft) on a 15ft grid with a ‘Z’ RMS (vertical accuracy) of 6ft was utilized. This data was also integrated with the InfoWorks WS model after preprocessing to an ESRI Grid format with ERDAS Imagine and final data extraction with ArcView.

Paper trail

System records were also available – paper maps, though not very well maintained, showed pipe sizes, valves, control valves, reservoirs, and pressure zones. In addition, a small model with some pipe sizes and node locations from a previous modeling exercise was available. The modelers also relied on communicating with people on the ground, gathering personal knowledge of the network through interviews, as well as water quality data.

The information about consumption was derived from a 2001 population census, house counts, details of special areas and enumeration districts including NWC-CIS by special areas and street addresses. Production data from the Sea View treatment plant was used to determine the typical daily demand profile.

The basic data elements and GIS data were processed in ESRI ArcView and included initial alignment of pipes, node locations and street names, which were required to facilitate operational outputs from the model. The pre-processing in ArcView allowed work to progress on two fronts at the same time.

A team of data entry personnel and modelers worked around the clock in ESRI ArcView and InfoWorks WS to determine the initial alignment of pipes, node locations and street names. The street name data was very important for coordination with the emergency crews that would use the operational outputs from the model. This pre-processing approach was also beneficial in that it enabled a GIS technician with no modeling skills to undertake a lot of the time-consuming data entry and validation for the model.

Data was prepared starting with the limited information available in the ESRI shape files, which was extended by heads-up digitizing on aerial photo base maps from system records. InfoWorks WS data import and validation tools were crucial to streamline this process.

Pre-processing precautions

Careful planning and organization of the GIS data pre-processing allowed for the incremental creation of the InfoWorks WS model. All new sub-sets had to have a unique numbering sequence or importation could lead to the corruption of previously-validated work. Although several InfoWorks extensions facilitate node and link construction, particular care had to be taken when using the extension ‘snap’ tools to ensure connectivity. In addition, careful planning of naming standards for the GIS database ensured seamless integration of data with InfoWorks WS.

Model analysis started as soon as node data was imported. Since DTM was available, node elevations could be accurately determined. The initial analysis consisted of a simple plot of the node elevation contour, which gave a general picture of the supply problems. The supply area was on a ridge, with lower areas to the north and south. Water supply was available from adjacent supply areas up to elevation 726ft in the south section of the model and 1580ft in the west section of the model. Both gravity and pumped augmentation options were possible.

In the model construction phase, automated InfoWorks WS tools for model construction, data validation and inference were irreplaceable time savers. InfoWorks tools were used for refining pipe alignments, node locations, adding pipe and node data, estimating missing data utilizing the inference tool, as well as extracting the information on reservoirs, pumps, and control valves from background images.

Initial validation included use of the proximity and connectivity trace tools, which Maurice Jones says was “essential”. Next, any nodes showing zero elevation were checked and their elevations were estimated from DTM using inference tools. He added: “The profile tool is also an important validation tool. The network can be checked for quirks that might have been overlooked or missed with other tools.”

After basic physical validation, trial demands were assigned to nodes for basic testing of the model. The initial analysis found that to divert supply from the west required system adjustments to PRV (pressure reducing valve) settings and placing some control valves on the bypass for reverse flow operation.

The model also revealed that small pipe sizes along the route would result in high head loss and air lock problems along the leveled, elevated sections of the profile. The flow was eventually diverted as recommended although, as predicted, the areas below the saddle at elevation 1230ft and higher received only a limited supply.

Supply in the elevated western area also had to be restricted. An additional proposal, to divert supply from a high lift pumping facility at the Constant Spring water treatment works to the Stilwell Road reservoir for further pumping into the higher elevations, could not be readily achieved.

It proved very difficult to get water to the reservoir, despite the indications from the model - existing demand in the area was likely to be well in excess of that allocated as a trial demand. A more realistic allocation of demands was needed to give an accurate input for the emergency model. Polygon demand data was fortunately available to model demands more accurately.


The model was developed, validated and used to find an alternative supply option within the tight four-day window thanks to some incredibly hard work on the part of the modelers and the tools available in InfoWorks WS. The achievement was particularly remarkable considering there was no existing model and only limited data was available.

Key among the solution’s tools in this project was its ability to integrate diverse data sources under the same platform including aerial photos, a GIS database, scanned drawings and digital models. InfoWorks WS data import tools also streamlined integration with existing data sources and allowed a team of people to work on the project at the same time.

Finally, when the data was imported, validation tools quickly helped pinpoint any deficiencies while inference tools helped to automatically estimate missing data.

As to the question of whether the huge effort to create the model in just four days was worth it, Maurice Jones says: “Yes. Why? If for nothing else every network needs a model.

“The network had many problems before the emergency and the creation of the model now gives a great opportunity to address them. The model has in the past month grown considerably to include other adjacent sub-systems as formalization of alternate supplies becomes pressing and solutions are being sought for other supply issues.”

Customer comments

No comments were found for Hurricane response emergency modeling with InfoWorks WS. Be the first to comment!