Fire forecasting
Methods, systems, and computer programs are presented for tools for fire monitoring. One method includes an operation for receiving, by a fire forecasting program, fire-related inputs including vegetation data, topography data, weather data, and fire-monitoring information. The fire-monitoring information includes the shape of fire burning in a region. Additionally, the method includes an operation for generating a fire forecast for the region based on the fire-related inputs. The fire forecast describes a state of the fire in the region at multiple times in the future, the state of the fire comprising a fire perimeter, a fire line intensity, and a flame height. Additionally, the method includes operations for receiving updated fire-monitoring information regarding a current state of the fire in the region, for modifying the fire forecast based on the updated fire-monitoring information, and for causing presentation of the fire forecast in a user interface.
1. A method comprising:
training a machine-learning program with training data to obtain a fire spread model, the training data comprising values for static features that do not change over time and dynamic features that change over time, the fire spread model estimating if a pixel on a map corresponds to an area that is on fire for a time in the future;
receiving, via a computer network and by a fire forecasting system, fire-related inputs including vegetation data, topography data, weather data, and fire-monitoring information including data identifying a physical shape of fire burning in a region;
generating, utilizing the fire spread model, a first fire forecast data for the region based on the fire-related inputs, the first fire forecast data describing a plurality of states of the fire in the region for respective multiple times in the future, each state of the plurality of fire states comprising a fire perimeter, a fire line intensity, and a flame height for respective times of the multiple times in the future;
receiving, via the computer network, updated fire-monitoring information regarding a current state of the fire in the region;
modifying, by utilizing the fire spread model, the first fire forecast data based on the updated fire-monitoring information to generate modified fire forecast data; and
causing presentation in a user interface of a fire forecast based on the modified fire forecast data.
2. The method of claim 1 , wherein generating the first fire forecast data comprises:
utilising a physical model to determine first fire spread data based on the fire-related inputs, the physical model being a simulation model based on physics and interactions between the fire and the atmosphere;
utilising the fire spread model to determine second fire spread data; and
combining the first fire spread data with the second fire spread data to determine a combined fire spread data, included in the first fire forecast data, describing the physical shape of the fire burning in the region for the multiple times in the future.
3. The method of claim 1 , wherein the region is divided into cells, wherein the first fire forecast data includes an indication of whether each cell is on fire.
4. The method as recited in claim 1 , wherein generating the first fire forecast data includes determining an estimated point of ignition of the fire, wherein modifying the first fire forecast data comprises adjusting the estimated point of ignition of the fire based on the updated fire-monitoring information.
5. The method of claim 1 , wherein generating the first fire forecast data comprises:
determining a future rate of spread of the fire based on the fire perimeters for the plurality of states.
6. The method of claim 1 , wherein generating the first fire forecast data comprises:
determining a scar area in the region for each state from the plurality of states, the scar area identifying parts of the region that have already burnt.
7. The method of claim 1 , wherein generating the first fire forecast data comprises:
identifying geographical blocks within the region that include residential dwellings; and
determining a risk level from fire for the geographical blocks.
8. The method of claim 1 , wherein the region is divided into a plurality of cells, and the first fire forecast data includes, for each cell of the plurality of cells, one or more: of slope of the terrain, fuel-model, average canopy-height in the cell, vegetation density in the cell, percentage of canopy cover in the cell, wind speed, wind direction, percentage of fuel moisture of vegetation in the cell, foliar moisture, and a flag indicating if the cell is an initial ignition site.
9. The method of claim 1 , wherein generating the first fire forecast data comprises:
accessing information from physical models for fire spread, each physical model being a simulation model based on physics and interactions between the fire and the atmosphere; and
combining the information from the physical models for fire spread with estimates from the fire spread model.
10. The method of claim 1 , wherein features of the training data are selected from a group consisting of satellite images, vegetation information for the region, weather data, land use, slope, aspect, elevation, and soil moisture.
11. The method of claim 1 , wherein the vegetation data includes information fire fuel in the region, wherein the topography data includes information regarding changes on the landscape caused by management activities and natural disturbance.
12. A system comprising:
a memory comprising instructions; and
one or more computer processors, wherein the instructions, when executed by the one or more computer processors, cause the one or more computer processors to perform operations comprising:
training a machine-learning program with training data to obtain a fire spread model, the training data comprising values for static features that do not change over time and dynamic features that change over time, the fire spread model estimating if a pixel on a map corresponds to an area that is on fire for a time in the future;
receiving, via a computer network and by a fire forecasting system, fire-related inputs including vegetation data, topography data, weather data, and fire-monitoring information including data identifying a physical shape of fire burning in a region;
generating, utilizing the fire spread model, a first fire forecast data for the region based on the fire-related inputs, the first fire forecast data describing a plurality of states of the fire in the region for respective multiple times in the future, each state of the plurality of fire states comprising a fire perimeter, a fire line intensity, and a flame height for respective times of the multiple times in the future;
receiving, via the computer network, updated fire-monitoring information regarding a current state of the fire in the region;
modifying, by utilizing the fire spread model, the first fire forecast data based on the updated fire-monitoring information to generate modified fire forecast data; and
causing presentation in a user interface of a fire forecast based on the modified fire forecast data.
13. The system of claim 12 , wherein generating the first fire forecast data comprises:
utilising a physical model to determine first fire spread data based on the fire-related inputs, the physical model being a simulation model based on physics and interactions between the fire and the atmosphere;
utilising the fire spread model to determine second fire spread data; and
combining the first fire spread data with the second fire spread data to determine a combined fire spread data, included in the first fire forecast data, describing the physical shape of the fire burning in the region for the multiple times in the future.
14. The system of claim 12 , wherein the region is divided into cells, wherein the first fire forecast data includes an indication of whether each cell is on fire.
15. The system of claim 12 , wherein generating the first fire forecast data includes determining an estimated point of ignition of the fire, wherein modifying the first fire forecast data comprises adjusting the estimated point of ignition of the fire based on the updated fire-monitoring information, wherein generating the first fire forecast data includes determining an estimated point of ignition of the fire, wherein modifying the first fire forecast data comprises adjusting the estimated point of ignition of the fire based on the updated fire-monitoring information.
16. The system of claim 12 , wherein generating the first fire forecast data comprises:
determining a future rate of spread of the fire based on the fire perimeters for the plurality of states; and
determining a scar area in the region for each state from the plurality of states, the scar area identifying parts of the region that have already burnt.
17. A non-transitory machine-readable storage medium including instructions that, when executed by a machine, cause the machine to perform operations comprising:
training a machine-learning program with training data to obtain a fire spread model, the training data comprising values for static features that do not change over time and dynamic features that change over time, the fire spread model estimating if a pixel on a map corresponds to an area that is on fire for a time in the future;
receiving, via a computer network and by a fire forecasting system, fire-related inputs including vegetation data, topography data, weather data, and fire-monitoring information including data identifying a physical shape of fire burning in a region;
generating, utilising the fire spread model, a first fire forecast data for the region based on the fire-related inputs, the first fire forecast data describing a plurality of states of the fire in the region for respective multiple times in the future, each state of the plurality of fire states comprising a fire perimeter, a fire line intensity, and a flame height for respective times of the multiple times in the future;
receiving, via the computer network, updated fire-monitoring information regarding a current state of the fire in the region;
modifying, by utilising the fire spread model, the first fire forecast data based on the updated fire-monitoring information to generate modified fire forecast data; and
causing presentation in a user interface of a fire forecast based on the modified first fire forecast data.
18. The non-transitory machine-readable storage medium of claim 17 , wherein generating the first fire forecast data comprises:
utilising a physical model to determine first fire spread data based on the fire-related inputs, the physical model being a simulation model based on physics and interactions between the fire and the atmosphere;
utilising the fire spread model to determine second fire spread data; and
combining the first fire spread data with the second fire spread data to determine a combined fire spread data, included in the first fire forecast data, describing the physical shape of the fire burning in the region for the multiple times in the future.
19. The non-transitory machine-readable storage medium of claim 17 , wherein the region is divided into cells, wherein the first fire forecast data includes an indication of whether each cell is on fire.
20. The non-transitory machine-readable storage medium of claim 17 , wherein generating the first fire forecast data includes determining an estimated point of ignition of the fire, wherein modifying the first fire forecast data comprises adjusting the estimated point of ignition of the fire based on the updated fire-monitoring information.
21. The non-transitory machine-readable storage medium of claim 17 , wherein generating the first fire forecast data comprises:
determining a future rate of spread of the fire based on the fire perimeters for the plurality of states; and
determining a scar area in the region for each state from the plurality of states, the scar area identifying parts of the region that have already burnt.