Modeling rainfall and flooding


by Xnn
Tags: flooding, modeling, rainfall
Evo
Evo is offline
#19
Nov18-09, 09:27 PM
Mentor
Evo's Avatar
P: 25,934
Quote Quote by sylas View Post
Ah. We have a new thread! Thanks. There seems to be a possible point of confusion: some readers may have thought that the paper is about predictions.

It is actually nearly all about an empirical observation from the twentieth century, of an increase in the rate of large floods. There is a fair bit of statistical work to try and sort out whether the increase is significant and suggestive of a change in conditions over the last century. There are also some model experiments to test a hypothesis about the cause of the observed trend.

The focus of the models in the latter half of the paper is to test a hypothesis for the cause of observed increases in great flood risk.

Significance of the increased great flood risk

The largest part of the paper is an evaluation of the significance of the observed increase in frequency of great floods. This is done in two ways:
  1. A simple statistical method assuming independence of events and a stationary process obtained a 1.3% probability of obtaining the observations by chance. A brief extract referring to this argument appears in msg #12, as the first extract in blue.
  2. Then a much more subtle method was used to deal with the independence assumption, which we should not actually expect to hold in general, even with a stationary process. The method used was to use output from a "control" experiment with coupled climate models and with constant radiative forcing. That is, the models here are not a prediction; but a basis for obtained a comparison with observations and estimating how unlikely they are under the stationary process null hypothesis.

The result with the model experiment indicate that the observed increase in great flood frequency in real life occurs about 3.5% of times in the models of a stationary climate. As the paper says:
Thus, the model-based significance analysis, which implicitly uses the spaceħtime correlation structure of floods in the model, essentially confirms and reinforces the simpler binomial analysis.
Cause of the increase in great flood frequency

On page 516 there follows a smaller part of the paper, which considers a possible hypothesis for explaining the observed increase in great floods. This section of the paper starts with the sentence:
The apparent increase in flood risk might be associated with radiatively forced climate change...
followed by three paragraphs detailing some further experiments with models and non-constant radiative forcing.

The conclusion of this part of the paper appears on page 517:
Thus, the recent history of the observed trend index is generally consistent with the range of results from the scenario experiments.
The experiments on scenarios here are therefore being used to test a hypothesis for the cause of the empirical observations described in the first part of the paper. This paper is not really a modeling paper; models are used simply as a tool in the paper to try and reveal details of the empirical observation.

Summary

This paper is focused on an empirical observation of floods in the past, and considers both the significance of the observation (could it be simply a random bit of bad luck?) and possible causes (could the cause of the observed increase be related to non-constant radiative forcing over the last century?).

The "increasing risk" described in the paper title is an observation of an increased risk over the twentieth century. The paper considers the significance and a possible cause, all in the context of observed changes in climate over the twentieth century.

The brief mention in the abstract that the trend is likely to continue is a fairly usual sort of comment in a paper like this, making explicit the relevance or importance of the work.

Cheers -- sylas
Does the paper take into consideration agriculture, animal farming, housing and business developments, roads, changes in drainage, redirection of streams and rivers, articial and natural, dams - artificial and natural, silting, etc... and that these things need to be studied all the way the down from the source of the water? This is a HUGE undertaking, but obviously very important in understanding and predciting where and why an area might flood. This is really complicted stuff.

Am I wrong that this paper falied to take these critical flood causing conditions into consideration?
mheslep
mheslep is online now
#20
Nov18-09, 10:11 PM
PF Gold
P: 3,021
Quote Quote by Evo View Post
Does the paper take into consideration agriculture, animal farming, housing and business developments, roads, changes in drainage, redirection of streams and rivers, articial and natural, dams - artificial and natural, silting, etc... and that these things need to be studied all the way the down from the source of the water? This is a HUGE undertaking, but obviously very important in understanding and predciting where and why an area might flood. This is really complicted stuff.

Am I wrong that this paper falied to take these critical flood causing conditions into consideration?
As shown in the last section Sylas referenced in post #12, the authors specifically point out they consider none of those factors Evo. But nor do they make any claim what so ever about what's causing the floods. They simply did the work to collect global flood data (100 yr) and run trend studies on it. They report that.

Unfortunately Xnn's post (#4) in the thread that introduced the 100 yr flood paper makes the statement "another paper that has found an increase in floods during the 20th century due to global warming", which is completely unsupportable from that paper.
sylas
sylas is offline
#21
Nov18-09, 10:12 PM
Sci Advisor
sylas's Avatar
P: 1,750
Quote Quote by Evo View Post
Does the paper take into consideration agriculture, animal farming, housing and business developments, roads, changes in drainage, redirection of streams and rivers, articial and natural, dams - artificial and natural, silting, etc... and that these things need to be studied all the way the down from the source of the water? This is a HUGE undertaking, but obviously very important in understanding and predciting where and why an area might flood. This is really complicted stuff.
The paper does not do any of that. It is not really a paper about predictions, and it most certainly does not attempt to predict where floods might occur. The paper is clear that conclusions are tentative, and of course you are right that this is a very complex area; no one paper is going to wrap up all of that!

The contribution of this paper is to quantify the stochastic significance of an observed twentieth century increase in great flood frequency (which suggests that there is some cause involved other than just chance) and also to present evidence that increased radiative forcing is a credible cause for the observed trend.

It does not rule out other factors, or attempt an exhaustive consideration of them. It does not attempt to predict where floods occur. It is rather testing a hypothesis which is already on the table in prior scientific work -- that large floods should be expected to increase given radiative effects of anthropogenic changes in atmospheric composition, such as intensification of the global water cycle. The paper concludes with suggestions for further work, which explicitly includes some of the aspects you mention.

The evidence from the paper does suggest that other factors work as well as, rather instead of, the factor of increased radiative forcing.

It's normal in a complex world for many factors to be at work. It's good science to give evidence that a particular factor has a role, without giving a complete wrap up of all causes for a phenomenon or a complete description that would let you make firm predictions. This paper is only about a general frequency world wide. There's no attempt to predict when, or where. There's still plenty of work to do here, of course!

Cheers -- sylas

PS. Actually, mheslep, IMO it would be a reasonable informal description to say that this paper gives evidence that the 20th century increase was due to global warming. The phrase used in the paper is "Radiative effects of anthropogenic changes in atmospheric composition"; and for a general audience it is quite reasonable to translate that as "global warming".
mheslep
mheslep is online now
#22
Nov18-09, 10:34 PM
PF Gold
P: 3,021
Quote Quote by sylas View Post
[...]
The contribution of this paper is to [...] and also to present evidence that increased radiative forcing is a credible cause for the observed trend.

[...]

PS. Actually, mheslep, IMO it would be a reasonable informal description to say that this paper gives evidence that the 20th century increase was due to global warming. The phrase used in the paper is "Radiative effects of anthropogenic changes in atmospheric composition"; and for a general audience it is quite reasonable to translate that as "global warming".
I agree with everything you wrote that I omitted in the quote above , and disagree with what's left! The authors provide no evidence for the cause of the observed data. The full statement in the abstract: "Radiative effects of anthropogenic changes in atmospheric composition are expected to cause climate change, in particular an intensification of the global water cycle1 with a consequent flood risk2" is background fluff. The only angle for radiative forcing in this paper is in the modeling. Frankly that statement should have gone into the introduction, and would have been fairly placed there, but not in the abstract which should be a tightly worded summary of the original work done by the authors, and not others (my view).
lisab
lisab is offline
#23
Nov18-09, 10:50 PM
Mentor
lisab's Avatar
P: 2,915
Quote Quote by sylas View Post
The paper does not do any of that. It is not really a paper about predictions, and it most certainly does not attempt to predict where floods might occur. The paper is clear that conclusions are tentative, and of course you are right that this is a very complex area; no one paper is going to wrap up all of that!

The contribution of this paper is to quantify the stochastic significance of an observed twentieth century increase in great flood frequency (which suggests that there is some cause involved other than just chance) and also to present evidence that increased radiative forcing is a credible cause for the observed trend.

It does not rule out other factors, or attempt an exhaustive consideration of them. It does not attempt to predict where floods occur. It is rather testing a hypothesis which is already on the table in prior scientific work -- that large floods should be expected to increase given radiative effects of anthropogenic changes in atmospheric composition, such as intensification of the global water cycle. The paper concludes with suggestions for further work, which explicitly includes some of the aspects you mention.

The evidence from the paper does suggest that other factors work as well as, rather instead of, the factor of increased radiative forcing.

It's normal in a complex world for many factors to be at work. It's good science to give evidence that a particular factor has a role, without giving a complete wrap up of all causes for a phenomenon or a complete description that would let you make firm predictions. This paper is only about a general frequency world wide. There's no attempt to predict when, or where. There's still plenty of work to do here, of course!

Cheers -- sylas

PS. Actually, mheslep, IMO it would be a reasonable informal description to say that this paper gives evidence that the 20th century increase was due to global warming. The phrase used in the paper is "Radiative effects of anthropogenic changes in atmospheric composition"; and for a general audience it is quite reasonable to translate that as "global warming".
But how can you say that you can't rule factors due to chance, when the modern landscape has changed drastically due to improved land moving machinery and infrastructure? Ruling out chance assumes that nothing else has changed in the system. But lots of things have changed in the twentieth century...levies, human built flood plains, acres and acres of concrete and asphalt. There are too many moving parts in the system to make a definitive statement, IM very HO, that any observations are, or are not, due to chance.

And while I agree that perhaps an "informal" conclusion may be reached from these observations, I certainly don't want to tether manufacturing with regulations based on such weak evidence.
sylas
sylas is offline
#24
Nov18-09, 10:52 PM
Sci Advisor
sylas's Avatar
P: 1,750
Quote Quote by mheslep View Post
I agree with everything you wrote that I omitted in the quote above , and disagree with what's left!
Ain't that always the way? Disagreement is so much more entertaining and motivating.

I guess it comes down to the question of whether model based experiments are evidence or not. I tend to think yes; it's no different in principle from how numerical simulations are used in all kinds of fields. For an example in one of my fields of interest, the best evidence for a certain group of asteroids formed in a past collision is a numerical simulation of how orbit parameters change over time. See: Nesvorny et al (2002) The recent breakup of an asteroid in the main-belt region, in Nature Vol 417, June 13 2002, pp 720-722. Poupular account Detective work identifies "baby" asteroids in New Scientist. Extract: "A computer model that calculates orbital changes indicates that the 13 best-known orbits of asteroids in the group coincided 5.8 million years ago - a sign they were then part of the same body.". Note that the model doesn't actually locate asteroids at a point in space, but simply figures out that they were all in about the same orbit at that time.

The abstract looks fine to me; it's quite normal for an abstract to give directly relevant context like that. The initial sentence is a hypothesis being tested in the paper using model based evidence -- or whatever you prefer to call it. The paper is giving support for the sentence in the abstract.

Cheers -- sylas
sylas
sylas is offline
#25
Nov18-09, 11:16 PM
Sci Advisor
sylas's Avatar
P: 1,750
Quote Quote by lisab View Post
But how can you say that you can't rule factors due to chance, when the modern landscape has changed drastically due to improved land moving machinery and infrastructure? Ruling out chance assumes that nothing else has changed in the system. But lots of things have changed in the twentieth century...levies, human built flood plains, acres and acres of concrete and asphalt. There are too many moving parts in the system to make a definitive statement, IM very HO, that any observations are, or are not, due to chance.
Science does not deal in formal proof. The paper does give good evidence that the increase was indeed significant (i.e., very unlikely to be just random chance); it also supports the hypothesis that a stronger water cycle in a warmer world should be expected to give a trend similar to what is observed.

Of course other factors may be involved. But "levies" don't change the frequency of floods; they try to limit the damage they cause. It's hard to see how flood plains could have an effect. Again, they may alter the impact of a flood, but that is not what is being measured. Most of the factors you mention are not really things that should be expected to have an effect, especially given the carefully defined account of a "great flood" described in the paper.

Unfortunately the paper is not generally accessible that I can see. But here's another extract:
Here, we consider 29 basins larger than 200,000 km2 in area for which discharge observations span at least 30 yr. We analyse annual maximum monthly-mean flows, rather than annual maximum instantaneous flows; these two are strongly correlated in large basins. In contrast with earlier studies4,5, this investigation has a global scope and focuses on extreme events; we analyse the 100-yr flood (that is, the river discharge that has a probability of 0.01 of being exceeded in any given year), which is commonly used in flood-risk assessment for river-basin planning and design of major structures. Choosing such a large-magnitude threshold probably reduces any distortion of our analysis by nonclimatic factors such as land-use changes and river development.
Note that the paper is specifically looking at river discharge to identify floods, and the 29 basins of the study are shown in figure 2. So although the paper is suitably cautious about not over-hyping the conclusion, and is explicit that there are other factors involved which could also be considered, the conclusions of the paper should not be surprising, and the investigation does give a credible level of support for its conclusion.

Note that this is far from an isolated paper. There's plenty of scientific work indicating that the water cycle has been strengthening, and that this is an expected consequence of a stronger greenhouse effect.

Cheers -- sylas
mheslep
mheslep is online now
#26
Nov18-09, 11:20 PM
PF Gold
P: 3,021
Maybe this is my imagination, but it seems to me that that papers in the hard physical sciences, e.g. physics, chemistry, rarely, say 1 in 10, put footnotes and/or background material justifying the work in the abstract, but that papers on climate do it almost de rigueur.
mheslep
mheslep is online now
#27
Nov18-09, 11:34 PM
PF Gold
P: 3,021
Quote Quote by sylas View Post
...But "levies" don't change the frequency of floods;
They apparently can change the frequency of great floods such as those discussed here. The extensive system built on the Mississippi river was credited in part with worsening the 500 year flood there some years ago. That is, that great flood might have been a one or two hundred year flood had the river been allowed to rise out of its banks naturally.
sylas
sylas is offline
#28
Nov18-09, 11:37 PM
Sci Advisor
sylas's Avatar
P: 1,750
Quote Quote by mheslep View Post
Maybe this is my imagination, but it seems to me that that papers in the hard physical sciences, e.g. physics, chemistry, rarely, say 1 in 10, put footnotes and/or background material justifying the work in the abstract, but that papers on climate do it almost de rigueur.
That's an interesting thought! it might be worth having a quick squiz at a range of journals to see if there is a difference in footnoting (easy to check objectively).

But the climate change mention here is directly relevant to the work of the paper; not merely background. The first sentence of the abstract is the factor that is explicitly taken up and tested in the paper as a factor that increases the frequency of great floods, using the model based analysis in the latter half. The last half is an explicit conclusion from the paper, also based on experiments performed by the authors and described in the paper. The bulk of the abstract describes the bulk of what the paper is doing.

Cheers -- sylas

PS. The paper explicitly describes great flood in terms of discharge from a large river basin. Levies don't impact that.
mheslep
mheslep is online now
#29
Nov19-09, 12:41 AM
PF Gold
P: 3,021
Quote Quote by sylas View Post
...
I guess it comes down to the question of whether model based experiments are evidence or not. ...
Okay, I should not have said 'no evidence', but rather 'no physical evidence', or
'only model based evidence' for the hypothesis that flooding is caused by climate change.


Register to reply

Related Discussions
Why does the temperature goes up following rainfall (then gradually cools down)? General Physics 2
High rainfall event Earth 3
Too Much Rain and Flooding! General Discussion 27
Flooding in the west General Discussion 30
New England Flooding General Discussion 9