RESUME / ABSTRACT

Land surface models (LSM's) are necessary to provide the lower boundary conditions for weather and climate numerical models. To do this, LSM's typically model the heat and moisture processes from 2 meters deep into the soil to the plant canopy. Of fundamental importance in these models is the soil moisture for it is known that large errors for this variable can negatively affect the LSM and impact a numerical weather prediction model coupled to it. The initialization of soil moisture in these models has been regarded as a challenge for more than a decade, given, among other issues, the unavailability of reliable soil moisture observations on a continental scale.

The above problem often leads to the need for using other observations to infer the necessary corrections to the soil moisture state. My past work at NCEP involved the improvement of several aspects ofthe NCEP Noah land surface model and its numerical properties (reliability, efficiency, updates and differentiability) and the inference of soil moisture from skin temperature via variational methods in a single point off-line mode.

My current work with CaLDAS (Canadian Land Data Assimilation System) involves the use of screen level air temperature and humidity to infer soil moisture states on a continental scale.