Data Assimilation Meetings at Reading

Date Meeting type Speakers
6 March 2013 Invited Speaker Jose Gomez-Dans (UCL)
The edge preservation society: Monitoring land surface disturbance using data assimilation.

Disturbance is omnipresent in the land surface: deforestation, fires, insect outbreaks or storms cause important changes to vegetation, and affect the ecology, climate and carbon balance of vegetation. Earth Observation data is the only practical way of studying and monitoring disturbance on a global scale. Typically, change detection techniques are used, but these are often heavily empirical and hard to generalise. More advanced techniques rely on the use of radiative transfer modelling to understand the physics of radiation scattering and absorption, but these often result in ill-posed problems. Regularisation (in the Tikhonov sense, so an expectation of smoothness in the evolution of the land surface) has been successful in these problems, but for monitoring disturbance it tends to smooth over the typically abrupt changes of the state that occur when for example, a fire burns through a landscape. Edge-preserving data assimilation is an extension to typical 4DVAR systems that allows for abrupt changes of the state. I will introduce the ideas behind edge-preserving techniques, and use them to illustrate a burned area detection algorithm using MODIS data.

Page navigation