7th Workshop on Forest Fire Management - EARSeL, European ...
7th Workshop on Forest Fire Management - EARSeL, European ...
7th Workshop on Forest Fire Management - EARSeL, European ...
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
156<br />
II - VALIDATION OF RS PRODUCTS FOR FIRE MANAGEMENT<br />
MODIS reflectances sensed within a temporal window of a fixed number of<br />
days are used to predict the reflectance <strong>on</strong> a subsequent day. Rather than<br />
attempting to minimize the directi<strong>on</strong>al informati<strong>on</strong> present in wide fieldof-view<br />
satellite data by compositing, or by the use of spectral indices, this<br />
informati<strong>on</strong> is used to model the directi<strong>on</strong>al dependence of reflectance.<br />
This provides a semi-physically based method to predict change in<br />
reflectance from the previous state. A statistical measure is used to determine<br />
if the difference between the predicted and observed reflectance is a<br />
significant change of interest. The algorithm is repeated independently for<br />
each pixel, moving through the reflectance time series in daily steps. A<br />
temporal c<strong>on</strong>straint is used to differentiate between temporary changes,<br />
such as shadows, that are spectrally similar to more persistent fire induced<br />
changes. The identificati<strong>on</strong> of the date of burning is c<strong>on</strong>strained by the frequency<br />
and occurrence of missing observati<strong>on</strong>s and to reflect this, the algorithm<br />
is run to report the burn date with an 8 day precisi<strong>on</strong>. Further algorithm<br />
details are provided in Roy et al. (2005) and Roy et al. (2008), and<br />
the product is available to the user community (WWW1).<br />
2 - The burned area validati<strong>on</strong> protocol<br />
The potential research, policy and management applicati<strong>on</strong>s of satellite<br />
products place a high priority <strong>on</strong> providing statements about their accuracy<br />
(Morisette et al., 2006). Inter-comparis<strong>on</strong> of products made with different<br />
satellite data and/or algorithms provide an indicati<strong>on</strong> of gross differences<br />
and possibly insights into the reas<strong>on</strong>s for the differences. However<br />
validati<strong>on</strong> with independent reference data is needed to determine accuracy<br />
(Justice et al., 2000). Validati<strong>on</strong> is the term used here, and more generally,<br />
to refer to the process of assessing satellite product accuracy by comparis<strong>on</strong><br />
with independent reference data (Strahler et al., 2006).<br />
An validati<strong>on</strong> protocol for the validati<strong>on</strong> of moderate resoluti<strong>on</strong> burned area<br />
products (Boschetti et al., 2009) has been developed as a joint initiative<br />
of the Committee <strong>on</strong> Earth Observati<strong>on</strong> Satellites (CEOS) Land Product<br />
Validati<strong>on</strong> (LPV) Subgroup (WWW2) and GOFC GOLD (Global Observati<strong>on</strong> of<br />
<strong>Forest</strong> and Land Cover Dynamics) <strong>Fire</strong> (WWW3).<br />
2.1 - Temporal requirements for validati<strong>on</strong> datasets<br />
Given that burned areas are a n<strong>on</strong>-permanent land cover change, it is necessary<br />
to define the temporal interval described by the validati<strong>on</strong> reference<br />
data. For example, in areas where forests burn, fire affected areas may<br />
remain observable in satellite data for years, while in grass/shrubland systems<br />
burned areas may disappear within a single fire seas<strong>on</strong>. The length of<br />
time that the spectral signature of burned areas is detectable in satellite<br />
data after a fire depends <strong>on</strong> the physical evoluti<strong>on</strong> of the post-burn sur-