Spread phenomena usually show uneven movement over space. Such unevenness is due to two reasons:
The anisotropy of spread occurs when any of the determining factors have directional components. For example, wind and topography cause anisotropic spread of wildfires.
One of the simplest spatial heterogeneous and anisotropic spread is elliptical spread, in which, each local spread shape can be thought as an ellipse. In a raster setting, cell centers are foci of the spread ellipses, and the spread phenomenon moves fastest toward apogees and slowest to perigees. The sizes and shapes of spread ellipses may vary cell by cell. So the overall spread shape is commonly not an ellipse.
r.spreadsimulates elliptically anisotropic spread phenomena, given three raster map layers about ROS (base ROS, maximum ROS and direction of the maximum ROS) plus a raster map layer showing the starting sources. These ROS layers define unique ellipses for all cell locations in the current computational region as if each cell center was a potential spread origin. For some wildfire spread, these ROS layers can be generated by another GRASS raster program r.ros. The actual locations reached by a spread event are constrained by the actual spread origins and the elapsed spread time.
r.spreadoptionally produces raster maps to contain backlink UTM coordinates for each raster cell of the spread time map. The spread paths can be accurately traced based on the backlink information by r.spreadpath module.
Part of the spotting function in r.spread is based on Chase (1984) and Rothermel (1983). More information on r.spread, r.ros and r.spreadpath can be found in Xu (1994).
Options spot_dist, w_speed and f_mois must all be given if the -s (spotting) flag is used.
r.spread -s max=my_ros.max dir=my_ros.maxdir base=my_ros.base \
start=fire_origin spot_dist=my_ros.spotdist w_speed=wind_speed \
f_mois=1hour_moisture output=my_spread \
x_output=my_spread.x y_output=my_spread.y
2. Before running r.spread, the user should prepare the ROS (base, max and direction) maps using appropriate models. For some wildfire spread, the r.ros module based on Rothermel's fire equation does such work. The combination of the two forms a simulation of wildfire spread.
3. The relationship of the start map and ROS maps should be logically correct, i.e. a starting source (a positive value in the start map) should not be located in a spread barrier (zero value in the ROS maps). Otherwise the program refuses to run.
4. r.spread uses the current computational region settings. The output map layer will not go outside the boundaries set in the region, and will not be influenced by starting sources outside. So any change of the current region may influence the output. The recommendation is to use slightly larger region than needed. Refer to g.region to set an appropriate computational region.
5. The user should be sure that the inputs to r.spread are in proper units.
6. r.spread is a computationally intensive program. The user may need to choose appropriate size of the computational region and resolution.
7. A low and medium (i.e. <= 0.5) sampling density can improve accuracy for elliptical simulation significantly, without adding significantly extra running time. Further increasing the sample density will not gain much accuracy while requiring greatly additional running time.
Last changed: $Date$