Geoscience Reference
In-Depth Information
developed for the NCAR Community Land Model (CLM) (Branstetter and Erickson III
2003
, and references therein). A river transport model is also useful because it can be used
to evaluate the performance of an LSM against gauge station data.
Most LSMs are soil-vegetation-atmosphere transfer (SVAT) models, where the veg-
etation is not a truly dynamic component. Recently, coupling of hydrological or SVAT
models with vegetation models has received some attention, to serve more specific eco-
logical, biochemical or agricultural purposes. Dynamic vegetation models are used to
simulate the evolution of vegetation cover, photosynthesis, carbon and nutrient inventories
and the fluxes of water, CO
2
,CH
4
,N
2
O, volatile organic carbon and fire-related emissions
between the land surface and atmosphere. Illustrative examples of vegetation models are
the Land biosphere Process and eXchange (LPX) model (Wania
2007
; Spahni et al.
2010
)
and the CoupModel (Gustafsson et al.
2004
; Jansson and Karlberg
2004
; Jansson et al.
2005
,
2008
; Karlberg et al.
2006
,
2007
; Klemedtsson et al.
2008
; Norman et al.
2008
;
Svensson et al.
2008
).
There are a number of potential problems with LSMs that can cause errors in the
forecast. These include components that cause ''model error'' and components that cause
''predictability error''. Components that cause ''model error'' are as follows: incomplete
description of physical processes perhaps done for computational efficiency, perhaps a
reflection of incomplete knowledge; inaccurate parameters; and inaccurate forcings.
Components that cause ''predictability error'' are inaccurate initial states and boundaries.
All these problems are the subject of research in the land surface modelling and assimi-
lation community.
4 Data Assimilation of the Hydrological Cycle
4.1 Introduction
The only practical way to observe the land surface on continental to global scales is by
satellite remote sensing. However, this cannot provide information on the entire system,
and measurements only represent a snapshot in time. Land surface models can predict
spatial/temporal land system variations, but these predictions are often poor, due to model
initialization, parameter and forcing errors and inadequate model physics and/or resolution.
A way forward is to merge the observational and model information through data
assimilation (Kalnay
2003
).
Mathematics provides rules for combining information objectively, based on principles
which aim to maximize (or minimize) a quantity (e.g., a ''penalty function'') or on
established statistical concepts (e.g., Bayesian methods) that relate prior information
(understanding, which comes from prior combination of observations and models), with
new information (e.g., an extra observation). The merged product, termed the posterior
estimate or an analysis, adds value to both observational and model information. The data
assimilation methodology takes account of the different nature (e.g., spatio-temporal res-
olution) of the observational and model information, using an observation operator (see,
e.g., Talagrand
2010a
).
Assimilation of land surface observations is at an earlier stage than, for example,
assimilation of atmospheric observations (see various chapters in Lahoz et al.
2010a
).
However, during the past decade, land data assimilation has been a very active field of
research. Land data assimilation considers both ground-based in situ data and satellite data.
Often, satellite land surface data are assimilated and the process validated using in situ
