Geoscience Reference
In-Depth Information
5.1 Assimilated Observations
•
To be useful for model development and assimilation, the dominant modes (in space
and time) of the land system must be sampled;
•
To be efficient for state updating, observations need to be available at a reasonable time
interval to capture short-term dynamical variations (cf. the importance of satellite
overpass frequency; Walker and Houser
2004
; Pan and Wood
2010
);
•
Observations must be collected in long enough historical records to identify long-term,
climatological,
statistics for bias mitigation (Reichle and Koster
2004
) or trend
identification;
•
Observations need to be sampled at different spatial scales to capture both local and
global processes;
•
There is a need to have a reasonable signal-to-noise ratio (e.g., SMAP's target of
brightness temperature uncertainty is 1.3 K; Entekhabi et al.
2010a
), and an uncertainty
in the error description appropriate for scientific studies;
•
There is a need to relate observations to key system state variables, that is, there needs
to be system observability.
5.2 Forward and Retrieval Models, with Particular Reference to Radiances
and Backscatter Processes
•
To achieve appropriate retrieval accuracy, there is a need to use advanced methods to
describe physical processes in radiative transfer models (RTMs);
•
When assimilating radiances at large scales (e.g., from microwave sensors), there is a
need for calibration of RTMs (De Lannoy et al.
2013
; Forman et al.
2013
).
5.3 Land Surface Models
•
There is a need to use advanced methods to describe physical processes (this limits
structural uncertainty) and couple land surface models with models describing more
specialized processes such as run-off routing, dynamic vegetation or snow (Pauwels
et al.
2006
);
•
There is a need for consistent global parameter datasets to limit predictive uncertainty
due to parameter uncertainty;
•
There is a need for high-quality forcing data (this limits input uncertainty), mainly for
precipitation (Maggioni et al.
2011
; Reichle et al.
2011
).
5.4 Data Assimilation Challenges
•
There is a need to fill in the spatial and temporal gaps in observations (Reichle and
Koster
2003
; De Lannoy et al.
2012
);
•
There is a need to disaggregate data in space and time and into their individual
components (Forman et al.
2012
; Reichle et al.
2013
);
•
There is a need to ingest directly radiances or backscatter information (as opposed to
retrievals) to avoid inconsistencies between auxiliary information in retrievals and land
surface models (Crow and Wood
2003
; Durand et al.
2009
; Flores et al.
2012
);