Discussion with WG1 Raster based approaches more difficult at larger - - PowerPoint PPT Presentation

Discussion with WG1

“Raster‐based approaches more difficult at larger scales”

• Best answer as finite element? – allows continuity of different scales
• But issues with data? Can they just be interpolated? What is an acceptable level
• f sampling?
• Nested modelling approaches? Including temporal nesting of processes as well as

spatial nesting

• All approaches are piecewise continuous – need to think about how to define

(dis‐)continuities

• Use of data to inform appropriate scales
• Other disciplines tend not to use connectivity as inputs
• But counterexamples of flow in fractured media
• Need to use connectivity measures to move between scales / simplify models

Discussion with WG1

“Emergence”

• Gully heads
• Vegetation patterns
• Is emergence just pattern recognition?
• Use of network tools to evaluate emergent patterns?
• Percolation theory as example of a way to inform expectations of

system behaviour / structure

Discussion with WG1

“Structure vs function”

• Costs if make separation – worthwhile in geomorphology because of

different rates of change in processes. No explicit guidance about what the threshold might be

Discussion with WG1

“characteristic scale”

• Part of answer to first question – see above!
• Key challenges across different disciplinary areas – no necessary

common solutions

• But often use network or graph theory

Discussion with WG2

• If supercomputers are the way forward with raster‐based models, how do we provide sufficient

input data at appropriate resolutions?

• From the data group people are mostly thinking of better and more dynamic land surface data (DEMs) while

there seemed to be little efforts to get better other spatially distributed model inputs

• Some of the colleagues would be interested in a better monitoring of vegetation change
• Utility of benchmarking exercise
• Seemed to be not in the focus of the data group
• “Voodoo connectivity” based on data uncertainty and epistemic uncertainty
• Not discussed
• Can we measure between the plot and catchment scales to assess models?
• Most discussed question:
• There are increasingly data available regarding sediment movement in catchment (due to regular UAV observations; However

quality of data might be not sufficient if short term event based is necessary)

• Moreover, the new (tracer) techniques were discussed which could be used for model validation
• Discussion moved to the possibilities of measuring surface runoff or other water related features within the catchment. Most

interesting seemed to be the surface runoff monitoring as shown in the talk of Rens Masselink.

• Lastly the use of measuring runoff and sediment delivery in streams was controversly discussed. If looking in catchment

connectivity (surface runoff and sediment fluxes) most agree that it might be better to measure at the toe slope and not in

• streams. However, for comparision of different sources of sediment, P etc. we probably need both.

Discussion with WG4

• To what extent are we using indices to interpret models or do we need to develop

indices to build/improve models?

• Not the most interesting index use for WG4… Probably existing indices are enough
• Models can be used to calculate indices for which there is no (spatially‐distributed) data: e.g.

Sediment Delivery Ratio

• High‐resolution indices can create parameters/function for coarser‐resolution models
• Can models be used to evaluate indices?
• Models are not necessarily better than indices
• Models can be used to interpret/develop indices: focus on where they disagree
• Consistency is expected due to similar spatial input data
• Interesting: use models to track sediment sources‐to‐sink, for index development
• Structure or function?
• Indices look at structural connectivity; models can help them become functional
• ‐> find appropriate relation between rainfall & connectivity index value which indicate “activation”
• e.g. in the connectivity exercise: plot IC vs SDR, according to rainfall, see if rainfall activates higher SDR

Discussion with WG4

• What are appropriate scales?
• Evaluation: both spatially distributed results, and lumped (e.g. Sediment Delivery Ratio)
• The larger the analysis scale, the less important are high‐resolution processes (maybe in tilled

landscapes microtopography is more important)

• Catchment scale can be too large for aggregated connectivity indices to be meaningful: internal

connectivity differences will be smoothed

• Same spatial resolution issues between indices and models
• Resolution can be too low or too high
• e.g. microtopography might complicate landscape analysis at larger scales
• Link between index and types of process that work in the landscape
• Too few indices from which to choose; but indices e.g. Borselli have been applied successfully

in different landscapes

• Being based on topographical principles would make indices applicable in several landscapes
• Indices might be more location‐independent for hydrology and erosion, than for landslides
• Indices might not work in very flat areas, as they are very dependent on topography

Questions from WG3 to WG5

• Caution : Only three people in the panel, of whom one true end‐user.
• Q1. Is there a preference for spatially distributed models vs. lumped models ? Do stakeholders need maps ?
• Likely depends on the stakeholder (politician, national‐regional‐local administration, practicioner, …)
• May depend on level of scientific education. Not all end‐users may be at ease with the use of mapped

information.

• Stakeholders often like maps (“map‐happy”), but may not be aware of the underlying assumptions. Maps

may be taken as the truth. The scientist / modeler may not be questioned regarding the model ouptuts.

• Other stakeholders on the contrary have extreme awareness of model assumption, and do not want black

box models.

• The preference between spatially distributed models vs. lumped models also depends on model output
• reliability. Well accepted, well‐documented models are preferred over mode ‘experimental models’, even if

the former are less spatially‐explicit.

Questions from WG3 to WG5

• Q2. Are raster models essential for stakeholders ?
• See above
• Q3. Lumped, network, vector models : how to communicate such

representations of landscapes to end‐users?

• Independent of the type of model, continuous feedback between modeler

and end‐user is essential.

• In addition, end‐users need to be provided with a level of confidence

regarding the model outputs. Indeed, the model outputs are seldom an end to itself, but will be integrated (mental models) with additional contextual information (social, political, ….) to arrive at decisions.

• The model needs to be transparent, whatever the type.

Questions from WG3 to WG5

• Q4. Trade‐off between realism of representation of connectivity

features/processes vs. ease of use / parameterization of models.

• Again, depends on the type of end user.
• Many end‐users are “single‐goal’ oriented and would prefer simple

tools, especiually practicioners or local actors.

• Complex models for multiple‐goal problems.

Questions from WG3 to WG5

• Q5. Is the answer to the above scale‐dependent ?
• Most likely. It also depends on the goal / purpose of the study.
• For some studies, it may be sufficient to know the overall effect of

measures (e.g., on flood control, erosion reduction) and less important to know what is happening where.

• At the very local scale, spatial models may be less effective than the

practicioners’ expert knowledge. E.g., practitioner may know where to do what based on his own experience and landscape analysis. But a ‘lumped model’ may help decide the type of actions and their extent (e.g., total length of erosion control structures needed to reduce erosion by X%).

Questions from WG3 to WG5

• Q6. Do stakeholders have preferred models that should adapt to

connectivity ?

• /
• Q7. Which connectivity‐related questions would stakeholders want

addressed ?

• (one example : on‐site vs. off‐site impact of erosion control measures

after forest fires. Is full control (disconnectivity) of erosion desirable

• r may it have long term negative impacts downstream by depriving

downstream areas of needed nutrients ?)