WO2016198873A1 - Flood risk mapping and warning system and method - Google Patents

Abstract

A system and method are disclosed which can provide pluvial hazard predictions over large scale areas and which can be regularly updated to take into account topographical changes over time as well as planned changes. The system uses a plurality of different data inputs for analysing and extracting data relating to elements which may contribute to water flow changes including for instance curbs, hedges, walls, buildings and so on. The system and method can be used for a variety of other applications other than pluvial hazard detection and prediction.

Description

FLOOD RISK MAPPING AND WARNING SYSTEM AND METHOD Field of the Invention The present invention relates to a system and method for mapping flood risk and for warning of flood risk in one or more zones. The system and method disclosed herein is not restricted to flood prediction as they can be used for a variety of other applications including, for example, revenue assurance and leakage detection by utility companies, specifically water supply and removal companies; as well as for other non-water relates uses some of which are indicated below.

Background of the Invention Flooding is a phenomenon which is experienced in many countries and has devastating consequences on inhabitants and businesses. Environmental changes are also bringing flooding to areas not previously affected and increasing flooding to flood prone zones.

It is historically difficult to determine the likelihood of flooding in a zone, such as in towns or streets, and various studies have sought to address this. Of importance are the reliefs created by structures within a zone, for example a road or network of roads, buildings and so on, which may affect the flow of surface water within the zone. Examples of studies in this field include: Overland Flow and Pathway Analysis for Modelling of Urban Pluvial Flooding, Professor Cedo Maksimovic et al, in Journal of Hydraulic Research Vol. 47, No. 4 (2009), pp. 512- 523 of the International Association of Hydraulic Engineering and Research; and "Urban Drainage Models for Flood Forecasting: 1/D/1 D, 1 D/2D and Hybrid

Models", N Simoes et al, at the 12^th International Conference on Urban Drainage, Porto Alegre/Brazil, 1 1 -16 September 201 1. While these studies discuss methodologies for seeking to determine risk of flooding in a built-up zone, difficulties arise with scaling these theories beyond confined areas and as a consequence have to date failed to find widespread application. These trials also do not and are not able to address many of the topographical elements which contribute to flood risk.

Summary of the Present Invention

The present invention seeks to provide a system and method for determining the risk of flooding in a zone and to a method of warning of flooding risk. The present invention also seeks to provide a system and method for predicting other climactic, water based and topographical events.

According to an aspect of the present invention, there is provided a method of determining risk of an environmental event in a zone caused by an

environmental element, including the steps of:

obtaining zone data from a plurality of sources including at least one cartographical map and at least one plan view image of the zone;

superimposing the data from the plurality of sources to form a combined data map;

determining from the combined data map the nature and location of obstacles to the environmental element;

determining from the determined obstacles a likely path of movement of the environmental element in the zone; and

determining from the determined path one or more areas in the zone at risk of an environmental event.

Typically, the environmental effect may be flooding.

The use of an image provides data relating to possible obstacles to water which are not readily determined by a cartographical map, such as fences, walls, kerbsides and so on, which may not appear at all on such a map. The combined data map can therefore provide a much more accurate indication of obstacles to the path of ground surface water than cartographical maps alone.

In one embodiment, the method includes the step of obtaining at least an aerial image, which in practice is preferably a visual image of the zone. In other embodiments, the image includes either or in addition a perspective or elevational image of at least a part of the zone. An image in plan can be superimposed directly, often after normalisation of the image data, with a cartographical map, with the image being used to identify features indicative of obstacles not on the cartographical map. A perspective or elevational view can also be used in conjunction with a cartographical map to identify such obstacles and in the case where both a plan view image and an elevational or perspective image are used, the elevational or perspective view can be used to verify the existence and nature of obstacles, such as walls, kerbs, steps and so on, which might not be apparent from a plan view image.

The plan view image in the preferred embodiment is obtained from LIDAR data. The cartographical map is preferably a high quality map, in the United Kingdom advantageously an Ordinance Survey map or similar map.

In an embodiment, the method also includes the step of obtaining land registry data indicative of planned or completed building works in the zone and combining the land registry data with the cartographical map and visual image data into the combined data map. Use of such data can provide an additional indication and verification of obstacles as well as potential obstacles. This data can also be used in a predictive assessment to provide advance warning of the effect of intended or planned obstacles on flood risk, for example by planned roads, buildings, walls and so on.

The land registry data can also provide, for instance by reference to planning data relating to intended works, information useful in determining which area or areas of a zone are likely to change and as a result which parts of the combined data map may need updating over time. This can significantly reduce the amount of data processing required to keep the combined data map up-to-date.

Advantageously, the method includes the step of obtaining street sewer location data, the combined data map including said sewer location data. The method may include the step of determining topographical data of the zone, the combined data map including said topographical data.

The method preferably includes the step of generating a warning indicator relating to the determined risk of flooding, advantageously a visual flood risk indicator. This may be a coloured indicator, for instance providing a plurality of different colours, each indicative of a level of determined risk. In an example, the different colours may include a red indication for a determined high risk of flooding, green for a determined low risk of flooding and amber for a determined

intermediate risk of flooding.

A system to implement the method includes a processor unit including a plurality of input/output elements able to be coupled to respective sources of data, and an output unit operable to indicate flood risk in a zone. The output unit may include a visual warning, in the preferred embodiment a light colour warning, such as red for high flood risk, amber for medium flood risk and green for low colour risk.

According to an aspect of the present invention, there is provided a system for determining risk of an environmental event in a zone caused by an

environmental element, including:

an input unit configured to obtain zone data from a plurality of sources including at least one cartographical map and at least one plan view image of the zone;

a processing unit for superimposing the data from the plurality of sources to form a combined data map, for determining from the combined data map the nature and location of obstacles to the environmental element, for determining from the determined obstacles a likely path of movement of the environmental element in the zone, and for determining from the determined path one or more areas in the zone at risk of an environmental event.

The system and method disclosed herein can provide a combine data map closely representative of the obstacles to ground surface water flow in a zone, better able to determine flood risk. The system is able to be scaled up to cover very large areas, including whole countries and continents, by selective use of data and also by integrating within the data indications of changes or planned in obstacles in a zone, thereby allowing focusing of data processing to those areas of change. In particular, the combined data map may be updated by using planning applications, information of byways and changes to satellite or aerial photographs over time and determining where the changes are. Brief Description of the Drawings

Embodiments of the present invention are described below, by way of example only, with reference to the accompanying drawings, in which:

Figure 1 is a flow chart of a preferred method of determining flood risk in a zone;

Figure 2 is a schematic diagram of the basic elements of a system as taught herein;

Figure 3 is a schematic map showing an example of flood prediction in a built-up area;

Figure 4 is a map showing the generation of an enhanced DTM;

Figure 5 is a bird's eye view of a zone for use in use in identifying areas of a tarmac in the zone;

Figure 6 shows the view of Figure 5, filtered to show enhanced detain of the zone's topography;

Figure 7 is a satellite image showing a bridge;

Figures 8 to 10 show the processing of LIDAR data related to the zone of Figure 7 in the processing of bridge identification;

Figures 1 1 to 13 show a practical example of bridge detection;

Figures 14 to 16 show how hedges may be identified;

Figures 17 to 19 show how grass/trees may be identified;

Figures 20 to 24 show how curbs may be identified;

Figures 25 to 29 show how walls may be identified;

Figures 30 to 35 show how buildings may be identified;

Figures 36 to 41 show how calibration can be used to ensure accurate data generation;

Figure 42 is an example of a pluvial hazard map;

Figure 43 is a 1 D-1 D model simulating street network flow and underground drainage system;

Figure 44 is a map showing possible flood hazards;

Figures 45 to 48 show the combined effect of flood hazard modelling on various areas of a city; Figure 49 is a schematic block diagram showing the primary elements of the preferred embodiment of system and method taught herein;

Figure 50 to 52 are colour charts showing colour filtering and processing for extracting data form images.

Description of the Preferred Embodiments

For ease of understanding the concepts and methodologies taught herein there is first provided a description of the preferred and potential sources of topographical data. Data processing methods and algorithms, needed to generate the combined data set for predicting events are also described. Specific examples of the preferred algorithms are then set out in a series of specific examples. It is to be understood that any number of the data sources may be used together, though in all embodiments it is preferred that there is at least one cartographical map and at least one plan view image of the zone depicted by the map.

Data Sources/Intermediary Products LIDAR Data

The preferred embodiments use Lidar Data, which is vertical scaled data providing a 3D geometric description of the ground. It comes in various formats from various sources so the system and method taught herein normalises this data to make sure the data has a common scale and resolution with other zone data.

LIDAR data is preferably handled as a Digital Elevation Model (DEM) format.

Ordinance Survey Map Data

The preferred embodiments use Ordinance Survey (OS) Map data provided by Ordnance Survey in the UK. For other countries, other sources of map data may be used. Ordinance Survey data provides land use data (which can also be obtained from Satellite Imagery) and also boundary, road and building

intersections. The method and system use OS data for calculation of intersections such as bridges and tunnels which can have a significant impact on the topology of land for flooding. Road network

In the preferred embodiments, data relating to the road network is processed by the following routines. a) Filters

The first step in processing the road network is to apply filters to remove areas of little or no interest. This includes filtering out green areas via a colour filter and then running a blurring filter to remove small details (where only large- scale detail is of interest). Any areas which are not standard road tarmac coloured (or connected to them) are filtered out at this stage. b) Edges

The system and method then run a Sobel edge detection routine on the image, which provides a greyscale output image that shows where all the edges are detected. This removes all extraneous detail from the image and the output is a grey scale image showing large feature outlines. c) Small Features

The output of the edge detection routine is put thorough a second filter to remove small details. The primary features of interest are lines of a certain thickness or above, anything else is likely to be of no concern.

These preliminary filters should provide an image which is almost entirely road network based. d) Skeletonise

The next algorithm to run is a skeletonising algorithm. This reduces a thick network of lines down to individual lines that are a single pixel wide. From these thin lines the process then picks up all the intersection and termination points using morphological transformations and stores both the orientation and position of those intersections.

The output is a list of intersection and termination points, with both a geographical location (expressed in local coordinate system) and an approximate orientation. That list can then be compared with the previous list from the last iteration and determine if any points have significantly moved or been added. New points or moved points indicate new road construction or road extension.

The comparisons which are run preferably include checking if the number of points has changed for a defined region and using fuzzy logic to examine each individual point to see if there is one on the list that roughly maps.

Land Registry Data

Land Registry Data is provided in the United Kingdom by the Land Registry. This data comprises a set of polygon lines indicating the boundaries of properties within an area.

The method and system taught herein overlays this data over other map data and use it for two primary purposes: (a) to define where boundaries are for intersection review (in the boundary crossing algorithms) and also to seek out potential boundaries (such as walls) on satellite imagery.

Satellite Imagery

The preferred method and system use satellite imagery for two purposes. First, to verify land usage in the area (roads, green pastures and to some extent man-made structures such as buildings) and, secondly, to investigate if it is possible to determine readily whether there are obstructions to assumed water paths from walls and other such structures.

In a preferred embodiment, the method and system pre-process a satellite image to prepare it for other subroutines for later use. The process is broken down into, preferably, five separate steps as follows: a) Satellite Image Retrieval

In this process the method and system check for the availability of satellite data for the date period of interest grabs the latest image. Depending on the API services used, the process may only have access to slightly older data due to cloud filtering.

The output from this process is image files (advantageously in GeoTIFF format). b) Cloud Processing

The obtained images are cloud filtered as standard, or come with an overlay that describes where clouds have been detected in the images. Some filters (such as UV images) work through clouds but still experience some distortion. This allows the process to stitch images together to get a more up-to- date image than the previous step. The system is preferably able to extract elements from newer images that are cloud free and 'stitch' those into older images to get a more up to date image for an area. The output of this step is the same as the previous step, but the images may be more recent in parts. c) Shadows

All images which are acquire can be expected to have ground shadows (not cloud shadows). The process performs two routines - the first to calculate where the shadows are (which can be done with a variety of existing algorithms) and remove them, but also by using the time of day and geo-location the process can reverse engineer height information from those shadows to help build up a 3D image of the landscape (or at least ascertain building heights).

The output of this step is a shadow-filtered image, alongside some vector information about the shadows which have been removed. d) Vehicles

Once the process has filtered out shadows it investigates the potential for removing vehicles from the roads where appropriate. This will depend on the resolution of the images. The output is the same GeoTIFF image filtered for vehicles (to a degree). e) Filters

Finally the system pre-processes the images with a variety of low-level filters to prepare it for later stages.

Street view Data

Street view data provides in the preferred method and apparatus a side or perspective view of the relief of land and structures in a zone, allowing, for example, review of a particular intersection or feature from the perspective at ground level to verify if a possible obstruction actually exists. Various method steps may be to determine if there is an obstruction visible on the image.

Buildings

Buildings can be identified, in the preferred embodiments by the following routines. a) Filters

The first step in processing buildings is to apply filters to remove areas of no interest. This includes filtering out green areas via a colour filter and then running a blurring filter to remove small details (only medium-scale detail is of interest). The process also alpha-masks the entire road network out of the image so that it can be ignored in this processing. b) Edges

Preferably, a Sobel edge detection routine is then run on the image, which provides a greyscale output image that shows where all the edges are detected. This will outline all the buildings and other small features that are left outside the road network. This also removes all extraneous detail from the image and the output is a grey scale image showing feature outlines. c) Small Features

The output of the edge detection routine is then put through a second filter to remove small details - this should remove anything else that is small (e.g.

sheds, cars that have not been filtered out, small geographical features).

These preliminary filters should provide an image which is almost entirely buildings and small feature based. d) Transforms

The preferred process is only interested at this stage in the zones, or blobs, that are left, which are all a certain size or above. For each zone the preferred routine simply stores the approximate centroid of that zone and its size. It is not concerned with vertices or anything else - buildings will either grow in size (which means the size of the zone or the centroid will change significantly (i.e. more than 10%), or appear (i.e. a new zone exists).

The comparisons run here will check the number of zones in a region

(which should stay constant if there is no change) and the approximate location and size of each zone (small changes are ignored, again we are using fuzzy logic to determine them). 'Normalised' Map data

Normalised data is produced by the preferred method and system by intermixing LIDAR data together at a single resolution and overlaying that with boundary and land use data. This combination of zone data provides an optimum model for determining ground surface water flow, which may for example use a 'rolling ball' model to describe such paths and pooling of water.

Sewer Layers Data

Sewers layer data may be determined from water companies and describes the underground model including relevant flow metrics. This is used in the preferred method and system to calculate the capacity of the sewer network and to check for any backflows and so on. Paths and Pools model

The paths and pools model used in the preferred embodiment describes how water flows on the ground according to land topography, land types and any barriers found from satellite imagery/street view.

Final pluvial risk analysis

Once combined map data for a zone has been generated and flood risk determined, the method and system of the preferred embodiment provide a green/amber/red matrix showing the probability of flooding in each area within that map.

Error ratio

In the preferred embodiment, calculated through each step above is an error coefficient. Each step determines how much error gets added in (as a 2D value across the map). This is also reflected in the pluvial risk calculation at the end (so an area could be red with a small error ratio, which indicates a higher risk potential).

Use of Process in the Detection of Leaks

In addition to being useful in the prediction of flooding, the system and method taught herein can also be used for other application. One is in identifying leaks in pipework, for instance.

To look at water leaks in pipelines the preferred system makes two assumptions. First, leaking water in a pipe will cause a natural cooling in the ground area around that leak. Secondly, that cooling is detectable by a thermal satellite image. a) Source data

The routine uses two sources of data - the pipe network (expressed as vectors) and a thermal satellite image.

The two sets of data are overlaid over one another and then each pipe link is followed along its route and store a thermal gradient for each unit of distance of the pipe in the database. This gives us a baseline for that pipe to see how it might change in the future. The thermal gradient for a pipe should be fairly constant; subject to ground conditions (different types of ground cover would have different thermal properties).

For each pipe vector a sequence of numbers is stored for each unit of measurement of the image obtained (depending on the satellite data resolution). b) Update data

Then, on a periodic basis, the system would run the same algorithm again giving a new thermal gradient for that pipe length. Comparing that to the original pipe gradient can highlight where there was a peak or trough different from the original pipe. If the overall gradient was the same for that pipe (irrespective of being slightly warmer or cooler due to transient conditions) then it can be ignored. This is done by normalising the data across the two readings (taking the average of each and subtracting the difference from the higher reading to bring the two charts into line). If there is then significant deviation this is flagged up for review.

Use of Process in the Detection of Vegetation encroachment

on power lines/railways

In a similar vein to the pipeline network above it is also possible to look at near-infrared and green imagery. Again we the system and method follow the network of vectors for the underlying network, only this time extrapolating out each side perpendicularly for a pre-set distance and storing those numbers as well. The routine is in effect looking for significant change in either the green spectrum or certain parts of the near-infrared spectrum that indicate significant changes in vegetation.

This will also be tied into a growth model suitable for the type of vegetation in that area (determined from geographic location) and rainfall/weather estimates to ensure that future prediction is dealt with as well.

The process preferably includes the following routines. a) Initial image

Using standard satellite imagery and near-infrared imagery, overlay the associated network as vectors.

Following each vector along its path way, calculate the broad path by extending out perpendicularly for a pre-set distance each way.

Store the green colour details (hue and saturation) and near-infrared details for each pixel. b) Subsequent Images

Use the same subroutine as above and compare the relative data.

Highlight significant changes in colouration (hue or saturation) and points where new pixels are encroaching onto the underlying network (that is getting closer to the vector in question). c) Growth Models

The preferred embodiment also runs a growth model simulation on that section of track to see the likely changes based on forecast rainfall/weather for the region and the usual vegetation type. This is an estimated calculation to see if there is likely to be significant growth over the next 3 to 6 months.

Processes

Topographical Generation

This phase or the method or element of the apparatus combines the data sets from LIDAR data at various resolutions with land registry data and Ordinance Survey data to determine an accurate topography of the ground in a zone under consideration. This works in the following steps:

1. Use of linear interpolation and extrapolation to intersect LIDAR data sets if they are at different resolutions, adjusting error ratio across the map accordingly (upwards in the case of worse granularity of data). This provides a single granularity of data (at the highest resolution available, which may be interpolated for some areas at a higher error ratio). 2. Overlay of this data with building information and land use information from the OS maps to allocate each LIDAR data point with a 'land use' type to assist with the rolling ball model.

a. Optionally, incorporating satellite imagery at this point to further enhance this data with more accurate information from the ground. This could be done by using colour filters (such as dropping out blues) and tuning the contrast of the image to produce more accurate land use descriptors (the method and system may pick up different tones of colour and potentially patterns to determine the type of ground, for instance grass, overgrown areas, rural land, concrete and asphalt which all have different run off characteristics for water flow).

3. Examining any bridges and tunnels on the OS map and adjusting the relevant LIDAR data points to the lower values (that is, within the bridge/tunnel) to ensure accuracy and to prevent the creation of a false 'wall' between areas that might be key to creating a flood pathway.

4. Overlay of boundary lines from Land Use data to determine where likely boundaries are for boundary crossing checks.

Rolling Ball Hydroloqical Model

This is an existing, proven system, preferably enhanced and/or adapted to perform better in a multi-use environment and with higher resolution data.

Intersection Checking

Once a pools and pathway model has been determined from the above, the method and system check for any intersection points in each flood pathway with the boundary information from the Normalised map data to see if they are crossing a boundary location. This is done with simple matrix functions.

Intersection Verification

For each boundary intersection identified above, the system and method verifies the corresponding satellite imagery and examines the intersection on that imagery to seek to determine if there is an obstruction on the ground. Again the method and system use image filters (in this example, dropping out different colours and adjusting tonality of the image) to seek to pick out the line of a wall or building along that intersection, for example. The method and system use the boundary information to determine where to focus, thereby avoiding the need to process the whole image. If the method or system determines there is a wall (preferably with a low error ratio feedback), the method and system update the topographical, combined, image with the existence of a wall (in one example by raising LIDAR data points along the wall where it can be seen) and re-run the hydrological model. This is likely to result in a change to the route of water flow in the area under consideration. The system and method then flag this area to indicate that the area has been adjusted, thereby to avoid a repeat check.

If the method or system determines with confidence that there is no a wall, the method and system moves onto the next intersection point.

If, on the other hand, the method or system is unable to verify the presence of a wall, the method or system checks to see if street-view information for that data point is available from a Street View provider (or angled aerial photography) and a view that section on the ground is determined if available.

The system and method again use image filters to enhance the image to better process the data, in this case seeking to overlay a 2D image into 3D world space to illustrate where the wall would be on the ground in 3D. This is preferably done using simple trigonometric functions and geographical data (map

coordinates).

The method and system preferably then examine the area under

consideration to see if there is a pattern resembling a wall at the point of interest (preferably using image analysis functionality such as pattern recognition and feature extraction, for instance using Gabor filters to find brick-work like patterns). Again, if the method or system find a clear wall the LIDAR data points are updated to create that wall and the rolling ball model is rerun. Uncertainty in this case is illustrated by extending the error value "down the pathway from the wall" as well as in the locality of the wall.

If street view data is not available the method and system preferably raise the error value in the area concerned, and extend that error ratio outwards around the area concerned based on how likely the wall is to exist. 1 d-1 d Hydroloqical model

This creates a pluvial risk factor at each point based on the intersection of water flow from the paths and pools model created by the Rolling Ball model (adjusted for obstructions) and the sewer network. This creates a close-to-realistic model of how actual rainfall is likely to progress through the sewer system, based on high capacity rainfall over an extended period (data supplied in this case by the Meteorological Office). Each LIDAR data point may be given a flood risk indicator from 1 to 100 and this is then translated into Red/Amber/Green status based on pre-determined rules.

Error calculation

An error calculation is an overlaid value across the LIDAR data points in the map. Each LIDAR node has a start error value of zero which increased by each level of uncertainty about that node (for instance from interpolated LIDAR data, uncertainty around path flow data and perhaps uncertainty around land use).

Curb Identification

Curbs can be identified by the following algorithm:

1. Overlay byway information onto satellite imagery (byway is freely available from OS on their vector product map)

2. Run edge analysis on the satellite imagery to enhance edges

3. Follow roads lines (with deviations and error margin) to find boundary edges. Extrapolate curbs alongside roads to factor around bridges, trees and any vertical obstructions. Use ground colouration on original map data to determine which are curbs (they have a fairly distinct and uniform colour as a rule).

4. Describe curbs using shape information from the above and store as an overlay for LIDAR DTM

5. To check joe guts of curbs use street view and compare against street furniture, cars, car tyres in places along the street.

6. To see where curbs dip into the road look for colour changes in photo and/or shading at intersections and driveways. Tracking Changes in Zone Data and Maps

The following is a preferred embodiment of process for tracking changes in satellite imagery over time.

The preferred embodiment does not store original satellite imagery (which may require very large storage commitment). The preferred system, by contrast, stores a series of hashes for satellite imagery (both computational hashes using standard algorithms such as MD5 and image processing hashes such as Gabor filters that describe the detail of the image in numerical terms) and lower resolution black and white imagery representative of the edge enhanced version (which has much lower storage requirements than a full colour image). Those hashes are computed from two areas:

Both involve losing the fine-grain detail between images but provide an effective overview of the diagrammatic representation. This means the system would pick up large scale change but not the small details that would change (e.g. cars moving, seasonal changes). The system may still pick up some false negatives (for instance a large car park will provide a totally different image between two satellite images), but it is better to have false negatives than false positives (that is where a change has occurred but is not detected).

From an algorithmic perspective the preferred embodiment performs the following:

1. For each map square (for instance, 1 km² resolution square):

a. Scale the satellite imagery to a standard resolution (to ensure scaling does not adjust the representation)

b. Calculate Gabor values and MD5 values for the satellite imagery

c. Create an edge enhanced version using Sobel edge enhancement and refine to 16 colour b&w at a low resolution (this produces a 'maze' like

representation of the image that outlines key features but loses detail)

d. Store both aspects in a database and link the edge imagery

2. When new map data becomes available, carry out steps a, b and c and compare to the original as follows: a. MD5 hash will determine if there is the exact same image or not instantly without any further refinement

b. Gabor filters will determine if there is no underlying change to the image on a visual basis, but the image is different (that is, it is another image taken by a different aircraft/satellite but nothing has really changed)

c. Edge enhanced version will then allow to verify which areas have changed (the system may also shift the edge enhanced version around to marry up with the original image in case the imagery is out of sync with GIS references slightly) if any.

Data Processing

The flow chart of Figure 1 shows the data processing steps of the preferred method. As will be apparent to the skilled person, the method shown provides an iterative data formation process for producing the combined data map and therefrom the determination of flood risk within the zone. The combined data map provides a much more complete indication of the topography within the zone in question, much better able to predict with better accuracy possible flood risk within the zone.

The plurality of data sources enables the method and system to identify elements which are not shown in cartographic maps. For instance, not only may walls not appear on a cartographic map but even if they are depicted, other separation structures, such as fences and the like, interact with water differently than a wall. The same applies for a hedge. The use of a plurality of data sources and in particular image data enables the method and system to distinguish between the various types of structures, and specifically can provide clarity as to how water passes along or through such structures.

The use of 3D views can identify alleyways, by using analogous methods as those for identifying walls. Hedges and trees can be identified by colour filters on satellite views and using boundary information as a starting point for assumption of the existence of possible hedges and trees. Walls and fences are identified by colour filters on satellite views and using boundary information as well. The type of structure can be identified by 3D views over laid with 2D images to give GIS DATA. The overlaid data views can then be converted to greyscale and the contrast increased, which provides a platform to use pattern recognition to identify the type of structure being analysed.

The preferred method and system advantageously learns. For example, each week/biweekly the system may review old satellite imagery and compare it to more recent to identify any changes to structures, road layouts and topographies. The system may use planning applications and road highways data to know when changes might occur and when to look at areas more frequently.

Figure 2 shows in very schematic form the basic components of a system for implementing the methodologies taught herein. The system includes in this embodiment a central processing system 100, although in other embodiments there could be a multiplicity of individual systems, for example used at customer or utility locations. The system 100 includes a data collection and processing system or unit 102 which receives data from a plurality of external data sources 100a- 100n of any of the types disclosed herein, and possibly also from a database 108, which may be internal to the system 100. The data sources 110 may provide stored or real time data while the database 108 may store more long term data, an example being OS map data.

The system 100 also includes a predictive engine 104 which generates one or more predictions, in the preferred example of flood risk in an area being surveyed. There is also provided an output unit 106, which may provide one or more warning indicators, such as a "traffic light" warning system as described above, acoustic and/or visual indicators, as well as communication to third parties by any suitable telecommunications system including telephony, via the internet, via a cellular or satellite network and so on. Some examples are set out elsewhere in this specification.

Some specific examples of data processing carried out by the system and method taught herein are now described. The examples are optimised for determining pluvial hazards, by enhanced DTM, changes in DTM, including pathways and ponds; for capacity prediction, that is water supply/usage prediction; and also for identifying network anomalies, such as pipework leaks and metering anomalies, in which water usage can be identified and compared to associated metering records. The preferred system creates an enhanced DTM that automatically identifies detailed features in the real topographical environment. It can automatically identify curbs, hedges, fences, walls and bridges, for example, and create a map of urban features within an existing DTM. This can be achieved with no costly human intervention, no human error, and is able to provide very accurate mapping from existing, low resolution DTMs.

The preferred system includes an Image Differential Engine which identifies changes over time in a given environment using automatic comparative analytics. This automatically compares historical and contemporary records, identifies where changes occur, can send alerts for updates. As only selective updates of the DTM are required, the system is very cost-effective in terms of processing requirements. Referring to Figure 3, this shows in schematic form an example of a map having superimposed an indication of fluvial hazard, achieved by accurately modelling pathways and ponding of pluvial surface water and its interaction with storm drain and sewer networks.

Figure 4 shows a map overview for building an enhanced DTM. Preferably, raw LiDAR data is converted to a raster and then used 'as is'. Features such as bridges (figure 4) form anomalies that may skew flow path calculations. Highway curbs not shown in normal DTMs can greatly influence flow paths. The enhanced DTM automatically picks these and other features out so they can be corrected to form a highly accurate model of the topographic environment.

Tarmac Detection

Figure 5 is a satellite image used for identifying areas of tarmac in a zone.

Satellite images are used and colour filters and hue/saturation filters then applied. Shadows affect tarmac images very strongly so are pre-processed to create an approximate alpha-mask for tarmac / asphalt. With reference to Figure 6, this shows the preferred alpha mask output, in which combined street centre-lines creates a solid road network. There is a low error coefficient due to combined data. Bridge Detection

Bridges and other overhead structures can be detected as follows. Figure 7 is a satellite of a zone with a bridge. Main water flow paths are often along roads. Bridges present a solid obstacle on LiDAR data as it is only a 2D raster image. The preferred method and system processes out bridges automatically using tarmac identification. That is, the street algorithm overlays tarmac to find bridges. Figure 8 shows the LIDAR data prior to processing. This is compared to a street centre line database (of main thoroughfares) and satellite data (for tarmac detection/anomaly detection). With reference also to Figures 9 and 10, short sections of streets are processed at a time. Street width is determined from boundary data or tarmac detection and also average height across the street width is calculated, if there is a distinct spike in street height this is deemed to be a bridge, or other overhanging structure. Average street height is calculated using linear regression and overlaid on bridge sections.

Referring to Figures 1 1 to 13, these show an example of processing of map data and street view data to identify a bridge. Figure 1 1 shows a map with a possible bridge highlighted from the map lines. A street view image (Figure 12) is obtained from one of the known sources, of the structure identified in the map. For this routine, all other objects apart from the bridge are ignored for the purposes of processing. Linear regression Figure 13) is used to align bridge height to street level. Seen right, elevated features now correspond to street level and will not falsely influence indicated flow pathways.

Hedge Detection

With reference now to Figures 14 to 16, these show the processing of image data to identify hedges and fences. Fences and hedges obstruct (and to a degree absorb) water flow. Fences are often too thin to be detected on LiDAR data. The method and system find fences and hedges along property boundaries and check for their presence on LiDAR data, which produces more accurate flow paths for water. With reference to Figure 15, this shows LiDAR data (for verification information), property boundary data (APN parcels) and satellite data (greenery detection / edges). With reference to Figure 6, filters are applied to the satellite image to pre-process the data as follows: colour detection is used for hedges, edge detection is used for fences. Boundary lines are overlaid and checked for presence of a fence/hedge. Colour detection confirms the type.

Finally, a cross-reference is preferably made to LiDAR data for relevant elevation changes.

Tree and Grass Detection

Referring now to Figures 17 to 19, these show how grass and trees may be detected. Grass and tree lined areas affect water absorption and flow rate differently to asphalt / concrete. Areas of grassland are not always accurately mapped. Individual trees can be detected if isolated, otherwise areas of trees can be isolated from grassland. Referring to Figure 18, LiDAR data is obtained, as is satellite data (multiband). Hue/saturation/luminance filters may be used to highlight the elements of interest. With reference to Figure 19, filters are applied to satellite image to pre-process the data. For instance, colour detection is used for green areas, texture analysis/comparison is used for differentiation between grass and tree, blob analysis is used for individual tree isolation (where possible). This data is cross-referenced to LiDAR data for tree verification (height

information).

Curb Detection

Figures 20 to 24 depict the identification of curbs in a zone of interest.

Curbs do not show up on LiDAR data, although they can significantly affect flow of flood water. Curbs have less impact when floodwater rises. With reference to Figure 21 , street centreline data is obtained as well as LiDAR data (to adjust heights - also output), satellite data (road width / asphalt x - 60cm) and a street view (curb height z-20cm). With reference to Figure 22, filters are applied to a satellite image to pre-process the following parameters: colour detection/radar for asphalt areas. The street centre line is then overlaid, for the determined road widths. This can be cross-referenced to a street view for curb height processing. There is lower LiDAR data in the mid-street section, to produce better street profile. On a digitised map (Figure 23), a numerical list of identified and measured roadside curbs can be produced for later efficient data harvesting. This can also be stored in a database, as depicted in Figure 24.

Wall Detection

Figures 25 to 29 show how walls can be identified and also some issues with buildings. With reference first to Figure 25, small boundary walls do not always get picked up in LiDAR data, yet they can have a significant impact on both flow pathways and rising floodwater (constraining or directly flow thereof). Walls are more likely to change rapidly on local landscape. The method/system cross- reference various data sources to produce estimates of walls and adjusts LiDAR data accordingly. With reference to Figure 26, the following data is utilised: LiDAR data (to verify / adjust data), satellite data (for edge detection) and boundary information (APN Parcels) - best place to search. Referring now to Figure 27, filters are applied to the satellite image to pre-process as follows: an edge detection routine (e.g. Sobel) is used to find edges; boundary information is overlaid onto the processed image to find matching lines; a cross-reference to colour profile to original satellite image is used to find best match and eliminate false positives; a cross-reference is made LiDAR data to find if already contained in data (or false positive, e.g. building edge), if not adjusted appropriately.

Figure 28 depicts some data issues which may arise from such processing.

There may be too many extraneous plots in the data and roads may be included in plots, creating false positives. Figure 29 shows further issues. Buildings are highlighted as 'plots', some of which may be overlapping. Plot data for land parcels may be too inaccurate, included building outlines in some cases and random sections of roadway in other places. Wall detection is usually more suitable for residential areas rather than commercial areas.

Building Detection

With reference now to Figures 30 to 35, these show how buildings may be identified. Referring first to Figure 30, buildings are often highlighted in LiDAR and other data, thought they change over time (development and demolition). Building roof structure contributes to ponding, so more accurate building maps can provide more accurate flow models and ponding. Building data can be used to improve LiDAR quality. With reference to Figure 31 , the following data may be used:

satellite Imagery (2D multi band); vector information from Open Maps and the Government, and LiDAR Data. Referring to Figures 32 to 35, the system employs colour filtering, roof-line ridge detection (illustrated), comparisons with LiDAR Data, as well as feature extraction from other data sources and the generation of an error coefficient from data source comparisons.

Error Coefficients

Error coefficients are likely to be very useful in practical implementations of the method and system, for the following reasons: (i) input data is not always accurate, (ii) automated image analysis is not a perfect science, (iii) there is always a chance of false positive and false negative.

For this purpose, each data artefact that is calculated comes with an error score. Finding the same data artefact in multiple sources increases the error score (higher = more valid). Only high value artefacts are used in final calculations For example, for bridges the presence on LIDAR data + presence on map + presence on satellite image is deemed to represent a perfect score. When present on 2/is it deemed highly probably, likely to be included for processing. When present on 1/3, the low error coefficient excluded from processing.

Some artefacts (e.g. buildings, trees and so on) can still have lower scores even though present as detection routines aren't 100% sure. Multiple processes / sources of detection can still increase that score. Calibration of the Data

Urban building characteristics vary from area to area, depending on social, cultural, commercial and demographic considerations.

Referring to figure 36, this shoes a demonstration of an area with buildings which are predominantly flat roofed. This creates challenges in writing algorithms to normalise heights. With reference to Figure 37, seen in contrast, residential buildings in perspective arch-roofed and show up with a greater signature in raw LiDAR data. A street view system can be used to examine walls from close perspective. Figure 39 shows building detection, in which tall buildings can be imaged in perspective, which can identify complex shapes, and building layering (multiple tiers), as can very high levels of shadows. Building height can be detected as indicated in Figure 40, by use of historic LiDAR data, OS survey mapping and stereoscopic satellite imagery combined. The height of buildings can be extrapolated from this combined data and this height data used to differentiate between buildings and street level. Satellite stereoscopic data may be

approximated DEM with 1 -2m resolution. Accurate building layout is overlaid onto satellite data. These measures eliminate errors and increase data quality. They can be used for slums, any building and any area. With reference to Figure 41 , it can be seen that the processed image has had bridges removed by levelling to street level, roadways dropped where curbs are detected, refined wall and fence structures where these have been discovered. The system and method can add/delete enhancements, create error coefficients and ground truth all features.

The Pluvial Hazard Map

Figure 42 shows an example of a pluvial hazard map. This is generated by 1 D/1 D watershed modelling, calculating rainfall, assessing watershed drainage capacities, calculating sewer network capacity and factoring storm drain capacity.

1 D/1 D Modelling

Recent developments in flood modelling have led to the concept of coupled (sewer/surface) hydraulic models. Figure 43 shows a 1 Dimensional sewer model coupled with a 1 Dimensional surface flow model (1 D/1 D) and can be used for near 'real time' modelling.

The system and method automate the pathways and ponds modelling, almost instantaneously. They automatically model interactions of surface and sub-surface water. The algorithms within 1 D/1 D model make the map very accurate whilst streamlining processed data. 2D Pathway Modelling

With reference to Figure 44, this shows a 2D map depicting possible flood hazards. Ideally, at least three models are created north, south and east of the test area (depending on the topography of the land and direction of possible flood waters). In this example, flow paths for 20, 50, and 100 year flood events are modelled. The modelling preferably shows how surface water interacts with storm drains, sewer networks buildings and roads This can be used to assess current efficiency of infrastructure, water systems and existing flood mitigation measures and therefore to plan future works to alleviate the issues of flooding in the locality.

Rainfall stations can be used to provide precipitation estimates which can be used in creating a pluvial hazard map. The data typically obtainable from rainfall stations includes station depth, rainfall duration and amount, and frequency curves. A hazard map, as depicted in Figures 45 to 48 can then be generated, with ponds and paths beijng depicted in blue and ditches being depicted in red. Figures 46 to 48 are zoom views 1 to 3 respectively shown in Figure 45.

Processing - Sobel Edge Detection

With reference to Figure 49, the preferred image processing routine extracts edges from obtained images. The following sub-routines are performed.

Zone creation (from Sobel): takes Sobel edges and produces zones

(described by polygons) that approximately match the source image.

Blocking algorithm: turns 'lumpy' LIDAR data into smoother blocks of data, assuming that the data is from an urban landscape (i.e. buildings / streets).

Creates smoother building outlines from the LIDAR data.

Zone creation (from LIDAR data): produces polygons from the outlines of suspected buildings within the LIDAR data.

Blurring/processing filters: standard image processing routines that ensure that data is suitable for our feature analysis engine. Varies depending on the source image type but usually consist of Gaussian filters to remove fine-grain pixilation and also shadow smoothing filters to remove source shadows to the highest possible degree. Feature analysis engine: a core part of the system. Each classification of feature (building, road, tree, water etc. etc.) has a defined set of characteristics stored in a database. These characteristics will vary from region to region, but consist of a set of boundary conditions within each of the source data types.

These characteristics preferably consist of the following data types:

Value range (e.g. Red, Green, Blue, Hue, Saturation, Luminance, height range, infrared, ultra violet, radar reflectivity levels and other characteristics)

Calculated characteristics (angled nature of the zone outline, height characteristics of the zone with reference to surrounding area (i.e. step up/step down), uniformity of height across profiled area)

The system the uses a comparison engine which defines each classification of feature by boundary conditions for each of the above characteristics. For each boundary condition that is satisfied by each feature the probability that that feature is of the correct classification. Each boundary condition is weighted as well, with tighter band conditions indicating more likely features.

For instance, areas of water will have a high radar reflectivity, will have higher blue values in the RGB spectrum, will have a fairly smooth uniformity of height and will be flat. These are expressed in terms of boundary conditions.

Boundary conditions will also vary by geographical area. For instance areas of grassland in humid countries will be more green due to better hydrological conditions, whereas they will be more yellow in drier conditions. This is preferably handled in two ways:

Boundary conditions can vary for each feature type across different areas. Even the angle of inbound sunlight can affect boundary conditions, so the geographic area has an impact on the boundary conditions.

Secondly the system 'trains' the boundary conditions by pre-processing a series of training images ourselves and identifying features manually (e.g. areas of water, buildings, road networks, grassland and trees). The system then uses an artificial intelligence engine to create boundary conditions for the peak areas for each of those images.

Fundamentally the 'feature space' is a multidimensional cube (hypercube) with discrete zones within that space representing each type of feature the system seeks to select. The Al routine scans across each of those features and creates a bounding space for each type of feature that identifies 95%+ of the matching pixels (for images) or the closest range of features for non-pixel features. The Al routine will also zone in on tighter bands where there is a higher match percentage (so we handle wide-spectrum fits more intuitively).

For instance, looking at the following feature graphs showing some examples of the boundary conditions.

Figure 50 is an example of feature band matching in the RGB space. The vertical line indicates the median point of the colour space and the broad band underneath the graph indicates the width of the 95% band. As can be seen from these samples, this creates a nice clear colour space (this example is for back garden pools).

However looking at a feature profile for pools in a different area, that is Figure 51 , a very broad spread of colour data can be seen. However, looking at the same data within hue/saturation and luminescence as per Figure 52 one can see nice narrow bands of feature recognition. Thus, in some circumstances the RGB spectrum is of no use, however from Hue/Saturation/Luminescence we see a very different picture emerges. Some features have dramatically different profiles across the many different criteria used to differentiate, and the system/algorithm has the capacity to identify different features from different satellite spectrographs or be extended into other remote sensing features (for instance one calibration mechanic is the height variance of LIDAR data across the profiled area).

The data objects are preferably broken down into the following:

Zones: Areas of interest within the map section being looked at (created by the zoning algorithm).

Feature types: Different types of features sought to be extracted (e.g. buildings, trees, areas of grass).

Criteria: The individual elements of data looked at to determine which feature is which (e.g. red colour value, radar reflectivity, LIDAR height variance). These are all expressed as numerical values within a defined range.

Boundary condition: The numerical 'spread' of each criteria linked to a feature type. Percentage match: The probability that each zone corresponds to one feature type.

Error coefficient: An indicator of our confidence that each zone

corresponds to one and only one feature type.

The preferred matching algorithm follows the following rough pseudo-code approach:

For each identified zone

Calculate all criteria

Set percentage match to zero

For each feature type

If the zone features are within the boundary condition of that feature type

Increment the percentage match by the feature type percentage

End If

Next Feature Type

If one (and only one) feature type is above our threshold level for identification, mark it as that type

If no feature type is above our threshold, and one is significantly higher, mark it as that type with a small error coefficient

If more than one feature type is above our threshold, take the highest value match and mark with a small error coefficient (based on the difference between the matched type and the next closest match)

If multiple feature types all have low match values then take the highest match and store with a high error coefficient based on the difference

Next zone

The preferred artificial intelligence engine works on the following pseudo code algorithm: For each feature type

For each identified zone

For each criteria

Calculate the spread of criteria

Calculate the median of the criteria values

Examine the spread of the criteria - If a narrow band calculate the boundary that represents 95% of the criteria

If a broad band calculate the boundary that represents a smaller percentage but a significant portion of the criteria.

Based on the number of criteria, assign an arbitrary percentage value to the

boundary condition (based on the number of criteria and band spread)

Next criteria

Examining the boundary conditions again, adjust the percentage of each narrower band up and each broader band down to create the best possible match.

Next zone

Next feature

For each feature type

Examine boundary conditions for other zones and see if there is dramatic overlap If there is, examine tightest bands for those zone types (that are different) and adjust percentage values accordingly (upwards) and other percentage values downwards

Next feature type

These features sets can also be created manually for some parameters, for example:

Building Water Grassland Road

Colour Spectrum Varied (5%) Blues (20%) Greens (20%) Greys (20%)

Radar Varied (5%) High (20%) Varied (5%) Low (20%) Reflectivity

Height variance Fairly uniform Uniform (20%) Varied (5%) Uniform (20%)

(i.e. low) - 20%

Height Higher than As As surrounding Lower/same comparison surrounding - surrounding/lower (20%) (20%) 20% (10%)

Zone outline Regular (straight Varied (5%) Varied (5%) Straight lines lines) - 20% (20%)

These are just some samples of data types that can be used. The system/algorithm is extensible to accommodate various types of raster and vector data. For instance, proximity to boundary lines can be a criterion for detection of walls.

In cases where no LIDAR is available, DSM can be created from two satellite images, for instance at 1 m z resolution, which can provide a rough height of features.

The teachings herein can be used for a variety of other applications, a few of which are described below.

Power line and rail network vegetation encroachment

The system and methods taught herein can also be used to identify and/or predict encroachment onto power lines, the rail network and so on of vegetation, enabling authorities to take action before the occurrence of any problems by such encroachment.

A preferred embodiment of this routine includes the following features: 1 ) Satellite imagery in visible light, UV and IR spectra is collected for the area of interest on a given date;

2) Using filtration, a signature curve is generated across the plane of the power cable or rail line at 90°, extending into the tree/vegetation line either side;

3) The data in this curve will have unique characteristics formed by the difference in shading from the presence of a variety of vegetation. Dense tree formations, lighter tree formations and scrub/grassland for example will all produce distinct signatures, hence identifying vegetation type/density boundaries in the area;

4) This initial curve is used as a calibration reference for subsequent curves derived from future imagery; 5) Further satellite imagery as above is sampled at regular intervals. The curves generated from this new data are compared to the original reference curve to determine where change has occurred in the vegetation signature;

6) The comparative process is completely automated, relying on relatively low resolution imagery to simply determine differences between sets of data, sampled over time;

7) Therefore the difference derived from comparing historical and contemporary data curves shows where the vegetation profile has changed in the given area.

Water pipe leak detection

A preferred embodiment of this routine includes the following features: 1 ) Satellite imagery in visible light, UV and IR spectra is collected for the area of interest on a given date;

2) Extending into the soil either side of the pipe, a signature curve is created by determining the characteristics of the soil surrounding the pipe;

3) The data in this curve will have unique characteristics formed by the presence or lack thereof of moisture at the time of image sampling;

4) This initial curve is used as a calibration reference for subsequent curves derived from future imagery;

5) Further satellite imagery as above is sampled at regular intervals. The curves generated from this new data are compared to the original reference curve to determine where change has occurred in saturation levels in soil surrounding the pipe;

6) The comparative process is completely automated, relying on relatively low resolution imagery to simply determine differences between sets of data, sampled over time;

7) Therefore the difference derived from comparing historical and contemporary data curves shows where the saturation level has changed in the given area, and thus determines where a failure in the pipe has occurred. Rail embankment/escarpment landslip detection

A preferred embodiment of this routine includes the following features: 1 ) Satellite imagery in visible light, UV and IR spectra is collected for the area of interest on a given date;

2) Extending into the embankment or escarpment either side of the rail line, a signature curve is created by determining the characteristics of the soil surrounding the line;

3) The data in this curve will have unique characteristics formed by the presence or lack thereof of moisture at the time of image sampling;

4) This initial curve is used as a calibration reference for subsequent curves derived from future imagery;

5) Further satellite imagery as above is sampled at regular intervals. The curves generated from this new data are compared to the original reference curve to determine where change has occurred in saturation levels in the embankment or escarpment;

6) The comparative process is completely automated, relying on relatively low resolution imagery to simply determine differences between sets of data, sampled over time;

7) Therefore the difference derived from comparing historical and

contemporary data curves shows where the saturation level has changed in the given area, and thus determines where the potential for landslip is present.

The method and system disclosed herein can be used in a variety of applications, including for instance:

Changes to flood maps over time. So the processing could be run once then again at 6 months intervals, which will show if there are any changes in the system;

Revenue assurance for utility companies

Planning applications and tax for governments, illegal planning, tree/grass maintenance, illegal cutting of trees, council tax, fluvial/coastal/groundwater flooding, creation of maps, insurance industry real understanding, disaster relief, disaster mapping, mapping of slums, mapping of battlefields/tanks, making of satellite navigation systems, making of robot intelligence, driverless cars, driverless drones, sea mapping, plane crash/boat sinking mapping, mapping in smoke or mapping on other planets.

3D Glasses - for visualising flood risk, underground systems particularly drainage and sewerage networks;

Construction workers and surveyors - for predicting possible flood risks, redesigning proposed systems and/or or making changes to structures in order to redirect flood water;

Other models of flooding

Other models for insurance and mortgages - for instance for assessing flood risks in dwellings and other buildings and in suggesting remedies;

Oil exploration

Vehicle navigations and control systems

Robots - for guidance systems

Changes to how companies create maps

Drones - for guidance systems

Navigation

In addition to the features and functionalities disclosed above, the combination of data from different sources makes it easier to identify changes within the zone and as a result to focus reprocessing effort, which can reduce significantly the overall processing volume thereby to make the method able to be scaled up to cover very large zones. The combined data map can be updated by using data from planning applications, information on byways, changes to satellite or aerial images over time and so on.

Other data inputs include: higher resolution images, drones, planes, 8 band images, multiscopic images, sonar, IR, L band radar, other bands of radar, microwaves, and so on.

The system can also be used to superimpose on existing data maps proposed works, such as planned roads, buildings, walls, water drainage systems and so on, to determine the likely change to flooding risk and therefore whether those planned changes are viable. All optional and preferred features and modifications of the described embodiments and dependent claims are usable in all aspects of the invention taught herein. Furthermore, the individual features of the dependent claims, as well as all optional and preferred features and modifications of the described embodiments are combinable and interchangeable with one another.

The disclosures in British patent application numbers 1510006.8, 1516403.1 and 1602577.7, from which this application claims priority, and in the abstract accompanying this application are incorporated herein by reference.

Claims

1. A method of determining risk of an environmental event in a zone caused by an environmental element, including the steps of:

obtaining zone data from a plurality of sources including at least one cartographical map and at least one plan view image of the zone;

superimposing the data from the plurality of sources to form a combined data map;

determining from the combined data map the nature and location of obstacles to the environmental element;

determining from the determined obstacles a likely path of movement of the environmental element in the zone; and

determining from the determined path one or more areas in the zone at risk of an environmental event.

2. A method according to claim 1 , including the step of obtaining an aerial image as the plan view image.

3. A method according to claim 2, wherein the aerial image is a visual image of the zone.

4. A method according to any preceding claim, including the step of obtaining a perspective or elevational image of at least a part of the zone, the combined map data including said perspective or elevational image.

5. A method according to any preceding claim, wherein the plan view image is obtained from LIDAR data.

6. A method according to any preceding claim, wherein the

cartographical map is an Ordinance Survey map.

7. A method according to any preceding claim, including the step of obtaining land registry data, the combined data map including said land registry data.

8. A method according to any preceding claim, including the step of obtaining street sewer location data, the combined data map including said sewer location data.

9. A method according to any preceding claim, including the step of determining topographical data of the zone, the combined data map including said topographical data.

10. A method according to any preceding claim, including the step of obtaining land slope data, the combined data map including said land slope data.

11. A method according to any preceding claim, including the step of generating a warning indicator relating to the determined risk of flooding.

12. A method according to claim 11 , wherein said warning indicator provides a visual environmental event risk indicator.

13. A method according to claim 12, wherein said risk indicator is a coloured indicator.

14. A method according to claim 13, wherein said coloured indicator includes a plurality of different colours, each indicative of a level of determined risk.

15. A method according to claim 14, wherein said different colours include a red indication for a determined high risk of the environmental event, green for a determined low risk and amber for a determined intermediate risk.

16. A method according to any preceding claim, wherein said

environmental event is flooding and the method detect for surface water.

17. A system for determining risk of an environmental event in a zone caused by an environmental element, including:

an input unit configured to obtain zone data from a plurality of sources including at least one cartographical map and at least one plan view image of the zone;

a processing unit for superimposing the data from the plurality of sources to form a combined data map, for determining from the combined data map the nature and location of obstacles to the environmental element, for determining from the determined obstacles a likely path of movement of the environmental element in the zone, and for determining from the determined path one or more areas in the zone at risk of an environmental event.

18. A system according to claim 17, including an input configured to obtain an aerial image as the plan view image.

19. A system according to claim 18, wherein the aerial image is a visual image of the zone.

20. A system according to any one of claims 17 to 19, including an input configured to obtain a perspective or elevational image of at least a part of the zone, the combined map data including said perspective or elevational image.

21. A system according to any one of claims 17 to 20, wherein the plan view image is obtained from LIDAR data.

22. A system according to any one of claims 17 to 21 , wherein the cartographical map is an Ordinance Survey map.

23. A system according to any one of claims 17 to 22, including an input configured to obtain land registry data, the combined data map including said land registry data.

24. A system according to any one of claims 17 to 23, including an input configured to obtain street sewer location data, the combined data map including said sewer location data.

25. A system according to any one of claims 17 to 24, including an input configured to determine topographical data of the zone, the combined data map including said topographical data.

26. A system according to any one of claims 17 to 25, including an input configured to obtain land slope data, the combined data map including said land slope data.

27. A system according to any one of claims 17 to 23, including a warning indicator relating to the determined risk of flooding.

28. A system according to claim 27, wherein said warning indicator provides a visual environmental event risk indicator.

29. A system according to claim 28, wherein said risk indicator is a coloured indicator.

30. A method according to claim 29, wherein said coloured indicator includes a plurality of different colours, each indicative of a level of determined risk.

31. A system according to claim 30, wherein said different colours include a red indication for a determined high risk of the environmental event, green for a determined low risk and amber for a determined intermediate risk.

32. A system according to any one of claims 17 to 31 , wherein said environmental event is flooding and the system detects for surface water.