Taking up where we left off in the previous post, we are going to explore
the spatial and temporal resolution. I am going to summarize key points from Watts, 2010 .
2.
Spatial resolution
Models can cover different spatial scales from few centimeters to
hundreds of kilometers. Hence, depending on the hydrological problem of interest,
the spatial resolution can be chosen according to the details that are to be described.
For instance, global hydrological cycles
are of hundreds of kilometers in spatial resolution. Such models require to be
broken down into smaller geographical units. This is where distributed
models come into the picture. Spatially distributed models process the
geographical units and produce results for each of them. So the whole catchment
is divided into sub areas.
On the opposite side there are lumped models, where the
whole hydrological system is considered as a single unit. For example a lumped
model of a catchment might consider only one point flow into the system and
give an averaged values for parameters such as soil moisture content.
Semi-distributed models sit between the lumped and distributed
models. These models are usually a series of lumped models.
3.
Temporal resolution
Hydrological models can be considered with different time-steps that
could range from few seconds to years. Some other models give averaged values
over a long time period. Model formulation has an impact on the chosen time
resolution. For instance, complex numerical models may behave unstable when
considering long time-steps. So it is up
to the modellers to choose the appropriate time resolution for the problem being
addressed.
So as we saw from this post and the previous post, there are
different factors and scales that modellers should consider when making the
decision of choosing a modelling approach. Next we are going to see what the
tools of decision making are when choosing a modelling approach.
Till next time,
Let's say you're trying to speed up a distributed hydrological model by reducing spatial or temporal resolution. What do you sacrifice? Are there any processes which can't be captured below a certain level of complexity?
ReplyDeleteHi Chad!
ReplyDeleteI think (and I'm not a professional hydrological modeller-bummer I know)You should see why one would choose distributed in the first place. If the spatial resolution can be reduced for the sake of speed ,and does not play a key role then why not a semi-distributed or even a lumped model? And I guess depending on what you are after, you can adjust the time-steps . Some problems may be interested in daily/monthly data, in which case annual time-steps ,even though faster might not be helpful.
About the second part of your questions, I am not sure about which processes as far as hydrology is concerned. But again I think it depends on what the model is looking to address. For instance, if there is a catchment that is under influence of more than one river or is influenced by tides, then perhaps more complexity should be incorporated. hope that answers the question to some extend :)