Fitting Ecological Knowledge to Remotely Sensed Long-term Monitoring Data: A framework from semi-arid grasslands of the Pilbara region of north-western Australia

<p>
Remotely sensed data are fast being integrated into long-term ecological monitoring. In this context remote sensed data are used to detect change in processes such as net primary productivity. However, can remote sensed data say more about ecosystem functioning and enable us to predict future system responses to events such as fire and flood? This paper describes methodology we are developing to fit ecological models of space-time vegetation dynamics to a time series of remotely sensed data. Our models produce an image of the spatial arrangement of vegetation for a point in time, dependent on ecosystem events. Output from our model is then compared to &#39;real&#39; remotely sensed data by applying image metrics. This approach allows for: (1) testing of competing models of ecosystem processes; (2) exploration of possible linkages and interactions between factors such as unpredictable fire and rainfall events; and, (3) estimation of hidden parameters in existing autecological models that are not easily measurable in the field. Out methodology has the potential to allow increasingly detailed models of space-time vegetation dynamics to be applied, and can be used to relate model predictions to real observed data.</p>