![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
The impact of the soil surface properties in water erosion seen through LandSoil model sensitivity analysis |
VerfasserIn |
Rossano Ciampalini, Stéphane Follain, Bruno Cheviron, Yves Le Bissonnais, Alain Couturier, Christian Walter |
Konferenz |
EGU General Assembly 2014
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 16 (2014) |
Datensatznummer |
250098895
|
Publikation (Nr.) |
EGU/EGU2014-14614.pdf |
|
|
|
Zusammenfassung |
Quantitative models of soil redistribution at the landscape scale are the current tools for
understanding space-time processes in soil and landscape evolution. But models use larger
and larger numbers of variables and sometimes it becomes difficult to understand their
relative importance and model behaviours in critical conditions.
Sensitivity analysis (SA) is widely used to clarify models behaviours, their structure
giving fundamental information to ameliorate models their selves. We tested the LandSoil
model (LANDscape design for SOIL conservation under soil use and climate change) a
model designed for the analysis of agricultural landscape evolution at a fine spatial resolution
scale [1-10 meters] and a mid-term temporal scale [10-100 years]. LandSoil is suitable for
simulations from parcel to catchment scale. It is spatially distributed, event-based, and
considers water and tillage erosion processes that use a dynamic representation of the
agricultural landscape through parameters such as a monthly representation of soil surface
properties.
Our aim was to identify most significant parameters driving the model and to
highlight potential particular/singular behaviours of parameter combinations and
relationships.
The approach was to use local sensitivity analysis, also termed “one-factor-at-time”
(OAT) which consists of a deterministic, derivative method, inquiring the local response O to
a particular input factor Pi at a specified point P0 within the full input parameter space of the
model expressed as:
-O/-P = (O2-O1) / (P2-P1)
The local sensitivity represents the partial derivatives of O with respect to Pi at the point
P0.
In the SA procedure the topographical entity is represented by a virtual hillslope on which
soil loss and sensitivity are calculated. Virtual hillslope is inspired from the virtual catchment
framework proposed by Cheviron at al. (2011): a fixed topology consisting of a 3X3 square
pixel structure having 150 m length allowing to test different spatial configurations of the
properties within the hillslope.
To test the model we identified different parameters. A three-category (P,R,p) sensitivity
analysis procedure was therefore found possible and appropriate to control the effects of
hydrological factors (P,R) and soil-terrain parameters (p). All the analysis were done with
the use of the integration of the ArcGis software structure, on which the LandSoil model is
based, and the PEST model (Doherty, 2004). PESTR is an iterative, non-linear parameter
analysis software platform based on the Gauss–Marquardt–Levenberg algorithm (Marquardt,
1963).
The results show the relevance of the rainfall amount in simulation and some interesting
interactions between parameters such soil roughness – soil crusting and soil cover. |
|
|
|
|
|