On the morning of 1 September 2026, the sirens sounded along Nullah Leh in Rawalpindi, and people living in the low-lying streets were instructed to evacuate, with the gauge at Kattarian Bridge reaching 29 feet. Since Dawn has stated that the nullah overflows at 30 feet, the flood in Nullah Leh must have been just about a foot below the point of overflowing. Approximately 200 millimetres of rain had fallen in the catchment area.
It is difficult to judge that kind of margin just from the gauge reading. The real issues are questions concerning planning. Once the water has left the channel, where does it go? What lies in its way? And which solution provides the greatest relief for the money?
We created a sample simulation of the flood in the Nullah Leh corridor using ArcGIS Pro in order to demonstrate that such questions can be investigated in just a few hours rather than having to wait for the monsoon season. The simulation is shown in our YouTube walkthrough of the Nullah Leh flood simulation. In this article, we explain what took place on 1 September, what the sample simulation shows, the point at which a quick tool like this ceases to be reliable, and how planners can make use of the method before the next flood occurs.
What was the outcome of the Nullah Leh flood on 1 September 2026?
A record rainfall for Rawalpindi
Heavy rain began in the early hours of Tuesday and continued for several hours. As reported by The Friday Times, the PMD stations recorded about 200 mm at Shamsabad (the reports indicate a figure between 197 and 201 mm), approximately 185 mm at New Kattarian and 157 mm at Golra in Islamabad. The Express Tribune described the rainfall as setting a record for the city over a 16-year period, since similar rain had last occurred in 2010. Dawn stated that there had been evacuations in Kattarian and Gawalmandi, and that the Rawal Dam had exceeded its prescribed limit and its spillways were due to be opened that afternoon.
What the gauges read
In Rawalpindi, it is the two gauges that give the main indication of the situation. The Kattarian Bridge is the point where the nullah enters the city, while the Gawalmandi Bridge is located further downstream. On 1 September, the level was 29 feet at Kattarian Bridge and 23.5 feet at Gawalmandi Bridge. According to the Tribune, local authorities considered 20 feet to be the evacuation level, and Dawn has stated that the danger level begins at 22 feet. Later that day, the water level dropped as the rain lessened.
The water often arrives from upstream
Dawn stated in July 2025 that more than 8,000 cusecs flowed down from Islamabad, with Kattarian reaching 14 feet and Gawalmandi rising to 9 feet, although the WASA chief said that almost no rain had fallen over Rawalpindi city itself.
The explanation lies in the area's geographical features. The three tributaries of Islamabad—Saidpur, Tenawali and Bedarawali—converge on the main channel just above the Kattarian Bridge, at the point where the two cities meet. Moreover, Nikki Lai, Pir Wadhai Kas and Dhok Ratta Nullah join the river still further on, as a case study organised by the WMO on the Lai Nullah basin states. Hence, the five water-source points in our sample run serve as a representation of the flow coming from the Islamabad side.
What causes the Nullah Leh basin to flood so rapidly?
A steep, urban catchment
Most sources state that the basin has an area of about 235 km²; the WMO case study specifies 239.8 km², of which 161.2 km² is in Islamabad, and 73.6 km² is in Rawalpindi. Over a distance of about 30 km, the elevation falls from 1,240 m in the Margalla Hills to 420 m at the Soan River. The hill tributaries are steep, with slopes in some places reaching nearly 10 per cent, so when it rains on the hills and on the built-up area of Islamabad, the water reaches the nullah quickly. A recent expert evaluation estimated that the time between heavy rain in the Margalla Hills and the flood peak arriving in Rawalpindi is one to three hours.
2001 set the benchmark
On 23 July 2001, the Islamabad Met station recorded 620 millimetres of rain over a period of about ten hours. The WMO case study states that 74 people died, that about 400,000 people were affected, 1,087 houses were completely damaged, and 2,448 houses were partly damaged in Rawalpindi, and that the losses exceeded Rs 15 billion. Since other reports provide different figures for the number of deaths, the death toll should be regarded as a range. The same document also mentions 19 flood years from 1944 to 2002.
Narrow points make it worse
According to an analysis by JICA, as reported by The Friday Times, development in Islamabad increases the peak runoff while encroachments and the dumping of waste reduce the channel's capacity in Rawalpindi. In an opinion article in that same paper, a civil engineer states that several of the bottlenecks between Kattarian and Chaklala were less than 20 metres wide before the improvements, the design top widths having been about 47 to 70 metres. That should be regarded as just one knowledgeable perspective. The important point is that both rainfall and channel capacity are relevant, and a simulation can be used to examine each.
What does the simulation of the Nullah Leh flood show?
The example given here is intended to demonstrate the method; the scenario it presents is not a prediction and has not been adjusted to match actual flood data.
How the sample is set up
The flood simulation tool employed in ArcGIS Pro—according to the Esri documentation—uses simplified shallow water equations processed on the graphics card to simulate the movement of water. The setup involves a digital elevation model of 50 cm (the presenter states that this comes from a Vantor and TomTom product), simulation cells of 1.115 m in size, and rainfall amounting to 200 mm per hour over the entire area for a one-hour period with no evaporation; there are five water-source points located along the nullah indicating the flow coming from Islamabad. On top of this is a layer consisting of 40 flood sites derived from newspaper reports, the records of which date back to 2001, for comparative purposes.
Where the water gathers
Deep water gathers round Kattarian and Dhok Hassu, at the point where the flow comes in from Islamabad, and then flows along the nullah past Ganj Mandi, Gawalmandi and Dhoke Ratta; the darkest blue indicates the deepest water on the 0 to 2 depth scale as shown in the legend.
The names that appear repeatedly in the news reports are also found among those listed by the WMO case study as areas that are frequently flooded—that is, the area between Ganj Mandi Bridge and the railway bridge, Arya Nullah, Dhok Ratta Nullah and Dhok Charaghdin. The same document states that flooding occurs in these areas when the Gawalmandi gauge reaches 18 feet; this is a visual match and not the result of a statistical test. Yet it is the kind of agreement you would expect to see before placing any trust in a model for other predictions.
Why ponds appear away from the channel
You'll see some ponds located on slopes and in certain depressions some distance from the nullah; that is exactly the purpose for which the tool was designed. According to Esri's technical paper, rainfall is distributed evenly across each cell, and water collects in the depressions. Our sample does not include storm sewers; since some of these depressions would drain via actual pipes, treat them as areas where the terrain holds water rather than as intended depths.
Could a rapid flood simulation be used in place of an engineering model?
Not at all, since Esri's own paper states the same thing.
What it is good at
Esri presents the tool as a rapid and approximate method for exploring a scenario, for testing measures such as barriers, and for checking the quality of the data; it is appropriate for the situation involving heavy rain falling on urban areas where the water is shallow and pools in low spots. This does indeed describe a great deal of the Nullah Leh problem.
What it does not model
It cannot be used for determining when water will arrive at a specific location or for areas with very extensive floodplains. Rainfall and evaporation are applied in an even manner, and sewer pipes are shown only by means of simplified connections. The default "Low" time setting is appropriate for shallow water. Esri recommends increasing the setting for deeper and faster flow, since this is similar to the conditions at Nullah Leh, which reached 29 feet on the gauge. Therefore, a planning-grade run should be carried out at a higher setting and then compared.
Elevation data and structures decide the answer
The tool takes whatever the 3D scene displays; if the elevation model has rooftops or a tree canopy or is missing bridges and culverts, then the water will be directed to the wrong locations. Esri also cautions that simple extruded building blocks can cause flow to be redirected in ways that actual buildings would not. The evidence on this point was discussed in our article about DEM quality and flood-map accuracy.
What can flood simulation offer planners in order to prevent the next Nullah Leh flood?
To cut to the chase, the worth lies in comparing the different runs. You should keep the rain the same, alter one thing, and then see what happens.
Test the choke points before building anything
Start with a baseline scenario; next, widen a pinch point, clear a blockage, or raise the bridge deck in the scene and then run the simulation again. The tool is capable of adding channels so that water can flow past a low bridge or culvert without having to edit the elevation model, and it can also be used to introduce barriers in order to test walls or embankments. Compare the depth results downstream and the modification which has the greatest benefit will be given priority in terms of the budget.
Site and size storage upstream
The 2004 WMO case study suggested the inclusion of a flood retardation basin in Islamabad, a diversion channel directed towards the Kurang River, and a joint warning system. Although it is beyond the scope of this post to say whether or not these measures now exist, the reasoning behind the site selection remains the same. In the programme, sink areas can be used as a stand-in for pumping or for underground storage, enabling you to determine where removing a particular volume of water each second will have the greatest effect on Kattarian. You can then pass the shortlist on to a full hydraulic model for design purposes.
Tie gauge readings to streets
Thresholds have already been established: 20 feet for evacuation, 22 for danger, and 30 for spilling, as stated. A simulation can operate a series of rainfall rates and produce depth maps at regular time intervals. When these maps are overlaid on the footprints of buildings and on the roads, you end up with a street-level list showing what will flood at each stage. The WMO case study referred to a warning manual which associates locations with flood categories; this is the same concept, but based on your own simulations. With a lead time of about one to three hours, even an approximate list helps to determine who should leave first.
Inform land-use and encroachment decisions
Apply the modelled depth figures to the 40 locations that have been documented and to the 19 vulnerable spots which have recently been identified. In cases where the water is deep and the buildings are located in the channel corridor, the decision is no longer purely abstract. The WMO case study states that the Rawalpindi authorities removed more than 2,000 encroachments after 2001, compensating the affected parties at market prices and drawing up a resettlement plan, and it also recommended zoning and building regulations. Since decisions regarding the corridor affect people's homes and means of livelihood, the maps should aid in the legal acquisition, compensation, and resettlement, not take their place.
What would make this simulation planning-grade?
Measure the channel, don't assume it
The measurements for the nullah, the bridge openings, and the culvert sizes should be obtained from survey work and not from imagery. When working on a hydropower project in northern Pakistan which was funded by the World Bank, we conducted a topographic survey covering an area of 80 hectares, which involved a bathymetric survey and delivered a DTM, a DEM and contour maps, all of which were accepted by the planning consultant's joint venture. The riverbed was measured rather than estimated from the banks. The same method would apply in the case of Nullah Leh, with the terrain model being updated after each desilting season.
Add real rainfall and drainage, then check against an event
Substitute the flat rate of 200 mm per hour with one constructed from PMD records, include infiltration and surface roughness, and take account of the drains; afterwards, check the results against 1 September 2026, using the station rainfall, the 29 and 23.5 foot gauge readings, and the streets that actually flooded. Satellite radar from remote sensing can be used to obtain the areas that were flooded on the event days, the limits being those which we outlined in the previous post.
Keep the model alive
Following each monsoon, add the new flood locations, replace the newspaper's locality points with the surveyed high-water marks, and then run the analysis again. Since it is stored in a GIS database, the sample becomes a tool which planners can use each season.
Final thoughts
On 1 September, the margin was approximately one foot with respect to the reported spill level. Three points should be noted here. The first is that the floods in Nullah Leh are caused by rain falling upstream and then encountering a narrow downstream channel, and both of these factors can be examined. The second is that a rapid simulation such as the one we have used is a quick method of addressing 'what if' questions, but it should be used in conjunction with engineering models, and the accuracy of its results depends on the quality of the terrain data, the details of the structures, and the validation available. The third point is that the value of the planning exercise lies in comparing different scenarios, connecting gauge levels to streets, and making updates following each monsoon.
Whenever you are involved in drainage, urban planning or flood preparedness in Rawalpindi or Islamabad, watch the walkthrough and let us know which scenario you would choose to run. We would be pleased to discuss how the sample can be turned into a survey-backed model.