16 February 2011

Why More Precipitation Does Not Necessarily Mean More Flood Damage

It is common to see studies that find an increase in precipitation (whatever the cause) quickly linked to claims of increasing floods. Making such a link, however comfortable and intuitive, is not so direct in practice.

In a peer-reviewed essay in the Bulletin of the American Meteorological Society in 1999 we explained the "apparent paradox" between observations of increasing precipitation but at the same time, a lack of trends found in the same regions in peak streamflow (floods).  Here is an excerpt from that piece:
Recently, Lins and Slack (1999) published a paper showing that in the United States in the twentieth century, there have not been significant trends up or down in the highest levels of streamflow. This follows a series of papers showing that over the same period "extreme" precipitation in the United States has increased (e.g., Karl and Knight 1998a; Karl et al. 1995). The differences in the two sets of findings have led some to suggest the existence of an apparent paradox: How can it be that on a national scale extreme rainfall is increasing while peak streamflow is not? Resolving the paradox is important for policy debate because the impacts of an enhanced hydrological cycle are an area of speculation under the Intergovernmental Panel on Climate Change (Houghton et al. 1996).

There does exist some question as to whether comparing the two sets of findings is appropriate. Karl and Knight (1998b) note that
As yet, there does not appear to be a good physical explanation as to how peak flows could show no change (other than a sampling bias), given that there has been an across-the-board increase in extreme precipitation for 1- to 7-day extreme and heavy precipitation events, mean streamflows, and total and annual precipitation.
Karl's reference to a sampling bias arises because of the differences in the areal coverage of the Lins and Slack study and those led by Karl. Lins and Slack focus on streamflow in basis that are "climate sensitive" (Slack and Landwehr 1992). Karl suggests that these basis are not uniformly distributed over the United States, leading to questions of the validity of the Lins and Slack findings on a national scale (T. Karl 1999, personal communication). While further research is clearly needed to understand the connections of precipitation and streamflow, in this letter we report the results of a recent study on the relationship of precipitation and flood damages. This letter seeks to address the apparent paradox from the perspective of societal impacts. We suggest that an analysis of the relationship of precipitation and flood damages provides information that is useful in developing relevant hypotheses and placing the precipitation/streamflow debate into a broader policy context (cf. Changnon 1998).

A recent study (Pielke and Downton 2000) offers an analysis that helps to address the apparent paradox. Pielke and Downton relate trends in various measures of precipitation with trends in flood damage in the United States. The study finds that the increase in precipitation (however measured) is insufficient to explain increasing flood damages or variability in flood damages. The study strongly suggests that societal factors – growth in population and wealth – are partly responsible for the observed trend in flood damages. The analysis shows that a relatively small fraction of the increase in damages can be associated with the small increasing trends in precipitation. Indeed, after adjusting damages for the change in national wealth, there is no significant trend in damages. This would tend to support the assertion by Lins and Slack (1999) that increasing precipitation is not inconsistent with an absence of upward trends in extreme streamflow. In other words, there is no paradox. As they write,
We suspect that our streamflow findings are consistent with the precipitation findings of Karl and his collaborators (1995, 1998). The reported increases in precipitation are modest, although concentrated in the higher quantiles. Moreover, the trends described for the extreme precipitation category (>50.4 mm per day) are not necessarily sufficient to generate an increase in flooding. It would be useful to know if there are trends in 24-hour precipitation in the >100 mm and larger categories. The term "extreme", in the context of these thresholds, may have more meaning with respect to changes in flood hydrology.
 Karl et al. document that the increase in precipitation occurs mostly in spring, summer, and fall, but not in winter. H. Lins (1999, personal communication) notes that peak streamflow is closely connected to winter precipitation and that "precipitation increases in summer and autumn provide runoff to rivers and streams at the very time of year when they are most able to carry the water within their banks. Thus, we see increases in the lower half of the streamflow distribution."

Furthermore, McCabe and Wolock (1997) suggest that detection of trends in runoff, a determining factor in streamflow, are more difficult to observe than trends in precipitation: "the probability of detecting trends in measured runoff [i.e., streamflow] may be very low, even if there are real underlying trends in the data such as trends caused by climate change." McCabe and Wolock focus on detection of trends in mean runoff/streamflow, so there is some question as to its applicability to peak flows. If the findings do hold at the higher levels of runoff/streamflow, then this would provide another reason why the work of Lins and Slack is not inconsistent with that of Karl et al., as it would be physically possible that the two sets of analyses are complementary.

In any case, an analysis of the damage record shows that at a national level any trends in extreme hydrological floods are not large in comparison to the growth in societal vulnerability. Even so, there is a documented relationship between precipitation and flood damages, independent of growth in national population: as precipitation increases, so does flood damage. From these results it is possible to argue that interpretations in policy debate of the various recent studies of precipitation and streamflow have been misleading. On the one hand, increasing "extreme" precipitation has not been the most important factor in documented increase in flood damage. On the other hand, evidence of a lack of trends in peak flows does not mean that policy makers need not worry about increasing precipitation or future floods. Advocates pushing either line of argument in the policy arena risk misusing what the scientific record actually shows. What has thus far been largely missed in the debate is that the solutions to the nation's flood problems lie not only in a better understanding of the hydrological and climatological aspects of flooding, but also in a better understanding of the societal aspects of flood damage.
So if you want to connect increases in precipitation to flood damages, you need to ask at least two more questions:

What has been the influence of increased precipitation on peak streamflow?
What has been the influence of increased streamflow on damage?

Too often such questions are left unasked and unanswered.

5 comments:

  1. Extreme weather linked to climate change: study

    http://www.abc.net.au/am/content/2011/s3141099.htm

    the research considered the time period 1950-1999. Hardly more than 1 "climate cycle" and this is considered "robust". What a weird world we live in where most of recorded history is ignored!

    ReplyDelete
  2. Roger:

    If the outlet channels of rainfall -- that is to say, the streams and rivers that carry rainfall away -- have not changed -- they how can more rainfall *not* lead to more flooding? It's simple physics, isn't it?

    David
    --
    David Appell
    davidappell.blogspot.com

    ReplyDelete
  3. -2-David Appell

    A good question ... think of this example -- suppose that the increase in precipitation occurs during the time of year when streamflow is at its lowest. Runoff could increase but the river never even approaches flood stage.

    In our 2000 paper we found that different measures of precipitation (e.g., 1-day, 2-day, 5-day etc.) have different relationships with flood damage. In some places 2 inches of rain is a ho-hum event, where I live it'd be a good share of our annual average.

    This is why it is necessary to address the two questions at the bottom of this post, if you want to make a better case for attribution. You can't just say more rain equals more flood damage.

    Ask again if unclear, Thanks!

    ReplyDelete
  4. Hi Roger, I think it may be interesting to discuss under this heading the two papers now out in Nature, on global attribution of changes in rainfall extremes and flood return periods in the UK:

    http://www.nature.com/news/2011/110216/full/470316a.html

    ReplyDelete
  5. David Appell,

    I see you live in Oregon (UO alum here). Think of the difference in rainfall and ground saturation between, say, April and August. In April the ground is saturated from about half a year of frequent rain so that it cannot absorb more, meaning there is greater runoff, which adds to the volume of water in the rivers. At the same time, snowpack in the Cascades will be melting, which also adds to the volume of the rivers. A two-inch in one day rain event then could easily push rivers into flood stage.

    In August the ground is usually dry so it has more absorptive ability, and most of the mountain snow that is going to melt already has, so the rivers aren't running as full. A two-inch rain event then will temporarily flood parking lots, for sure, but the soil will absorb quite a bit and the rivers can take the rest without rising above their banks.

    If climate change caused a seasonally specific increase in severe precipitation events during the spring, then I'm sure we'd see a correlation with increased flooding. Still, the finding is a little surprising, because if the severe precipitation events are evenly distributed, approximately 25% should be in spring, so I'd expect to see some degree of increased flooding. I wonder if R. Pielke can provide some insight on that?

    ReplyDelete