In an environment where numerous errors can occur which will impact on the model outputs, it is naive to think that only one source of error will exist. This example shows the combined effect of errors in slope and aspect algorithms as well as errors in the fuel type classification.
The fire risk model developed by CNIG usesseveral input data sets. these include slope, aspect and fuel type. Errors associated with algorithms used for deriving slope from DEMS are discussed elsewhere within this demonstrator. Aspect can also be derived from a DEM in numerous ways, often resulting in slightly different results
Within the STORMS project, work has been undertaken to develop methods of assessing fuel type from satellite images. Fuel models 4, 5 and 6 (which are all within the classifications for pine trees) are easily confused with each other when interpreting satellite imagery, nonetheless, ground truth measurements suggest that 75% of these classes are correctly identified. This still leaves a quarter of the area occupied by these fuel types which may be incorrect.
In the present example, two fuel maps are presented for an area in portugal. One is considered to be the "best guess", the other has had 25% of the area reported as consisting of these fuel classes 4, 5 and 6 randomly reclassified into an alternative possibility. The modified fuel map can be viewed by selecting the "confused" fuel map option.
The influence of slope and aspect algorithm are also included, with examples for Ritter's (the IDRISI default) and Horn's (the ARC/INFO default) methods also being available.
The effects of using these different datasets as input for FIREMAP results in two scenarios of fire-risk, both of which could be argued as being "correct", highlighting the uncertainty associated with modelling.