The U.S. <strong>Climate</strong> <strong>Change</strong> <strong>Science</strong> <strong>Program</strong>Chapter 3The 1993 floodranks among thetop five weatherdisasters in theUnited States.large in scope and persistent, such as those thatoccurred during 1993 in the Upper MississippiValley (Kunkel et al., 1994; Anderson et al.,2003) (and again in 2008), regionally extensiveepisodes of flooding can occur. When climatesignificantly changes, as it has in the past(Knox, 2000), and will likely do in the future,changes in the overall flood regime, includingthe frequency of different size floods and theareas affected, will also occur (Kundzewicz etal., 2007).6.1 The 1993 Mississippi ValleyFloods—Large-Scale Controls andLand-Surface FeedbackThe flooding that occurred in the Upper MississippiValley of central North America in thelate spring and summer of 1993 provides a casestudy of the control of a major flood event bylarge-scale atmospheric circulation anomalies.Significant feedback from the unusually wetland surface likely reinforced the wet conditions,which contributed to the persistence of thewet conditions. The 1993 flood ranks among thetop five weather disasters in the United States,and was generated by the frequent occurrenceof large areas of moderate to heavy precipitation,within which extreme daily total rainfallevents were embedded. These meteorologicalevents were superimposed on an above-normalsoil-moisture anomaly at the beginning ofJune of that year (Kunkel et al., 1994). Theseevents were supported by the occurrence of alarge-scale atmospheric circulation anomalythat featured the persistent flow of moisturefrom the Gulf of Mexico into the interior of thecontinent (Bell and Janowiak, 1995; Trenberthand Guillemot, 1996). The frequency of seasonal(90-day long) excessive (i.e., exceeding a20-year return period) precipitation anomalieshas generally been increasing over time inthe United States (Kunkel et al., 2008 (CCSPSAP 3.3, Sec. 2.2.2.3, Fig. 2.9)).The atmospheric circulation features thatpromoted the 1993 floods in the MississippiValley, when contrasted with the widespreaddry conditions during the summer of 1988,provide a “natural experiment” that can be usedto evaluate the relative importance of remote(e.g., the tropical Pacific) and local (over NorthAmerica) forcing, and of the importance offeedback from the land surface to reinforcethe unusually wet or dry conditions. For example,Trenberth and Guillemot (1996) used acombination of observational and “reanalysis”data (Kalnay et al., 1996), along with somediagnostic analyses to reveal the role of largescalemoisture transport into the midcontinent,with dryness occurring in response to less flowand flooding in response to greater-than-normalflow. Liu et al. (1998) used a combination ofreanalysis data and simple models to examinethe interactions among the different controlsof the atmospheric circulation anomalies inthese 2 years.Although initial studies using a regional climatemodel pointed to a small role for feedback fromthe wet land surface in the summer of 1993 toincrease precipitation over the midcontinent(Giorgi et al., 1996), subsequent studiesexploiting the 1988/1993 natural experimentusing both regional climate models and generalcirculation models point to an important rolefor the land surface in amplifying the severityand persistence of floods and droughts (Bonanand Stillwell-Soller, 1998; Bosilovich and Sun,1999; Hong and Pan, 2000; Pal and Eltahir,2002). These analyses add to the general patternthat emerges for large moisture anomalies(both wet and dry) in the midcontinent ofNorth America to have (a) local controls (i.e.,atmospheric circulation and moisture fluxover North America), (b) remote controls (e.g.,Pacific SST anomalies), and (c) a significantrole for feedback that can reinforce the moistureanomalies. The 1993 floods continue to bea focus for climate model intercomparisons(Anderson et al., 2003).6.2 Paleoflood HydrologyThe largest floods observed either in theinstrumental or in the paleorecord have avariety of causes (O’Connor and Costa, 2004),for the most part related to geological processes.However, some the largest floods aremeteorological floods, which are relevant forunderstanding the nature of abrupt climatechanges (Hirschboeck 1989; House et al. 2002)and potential changes in the environmentalhazards associated with flooding (Benito etal., 2004; Wohl, 2000). Although sometimesused in an attempt to extend the instrumentalrecord for operational hydrology purposes (i.e.,fitting flood-distribution probability density110
Abrupt <strong>Climate</strong> <strong>Change</strong>functions; Kochel and Baker, 1982; Baker etal., 1988), paleoflood hydrology also providesinformation on the response of watersheds tolong-term climatic variability or change (Ely,1997; Ely et al., 1993; Knox, 2000), or to jointhydrological-climatological constraints onflood magnitude (Enzel et al., 1993).Knox (2000, see also Knox, 1985, 1993) reconstructedthe relative (to present) magnitude ofsmall floods (i.e., those with frequent returnintervals) in southwestern Wisconsin during theHolocene using radiocarbon-dated evidence ofthe size of former channels in the floodplains ofsmall watersheds, and the magnitude (depth) oflarger overbank floods using sedimentologicalproperties of flood deposits. The variationsin flood magnitude can be related to the jointeffects of runoff (from precipitation and snowmelt)and vegetation cover (Fig. 3.13). The largestmagnitudes of both sizes of floods occurredduring the mid-Holocene drought interval,when tree cover was low, permitting morerapid runoff of flood-generating snowmelt andprecipitation (see Knox, 1972). As tree coverincreased with increasing moisture during theinterval from 6 ka to 4 ka, flood magnitudesdecreased, then increased again after 3.5 ka aseffective moisture increased further in the lateHolocene.The paleoflood record in general suggests aclose relationship between climatic variationsand the flood response. This relationship maybe quite complex, however, inasmuch as thehydrologic response to climate changes ismediated by vegetation cover, which itselfis dependent on climate. In general, runofffrom forested hillslopes is lower for the sameinput of snowmelt or precipitation than fromless well-vegetated hillslopes (Pilgrim andCordery, 1993). Consequently, a shift from dryto wet conditions in a grassland may see a largeresponse (i.e., an increase) in flood magnitudeat first (until the vegetation cover increases),while a shift from wet to dry conditions maysee an initial decrease in flood magnitude,followed by an increase as vegetation coveris reduced (Knox, 1972, 1993). This kind ofrelationship makes it difficult to determine thespecific link between climate variations andpotentially abrupt responses in flood regimewithout the development of appropriate processmodels. Such models will require testing underconditions different from the present, as is thecase for models of other environmental systems.Paleoflood data are relatively limited relative toother paleoenvironmental indicators, but workis underway to assemble a working database(Hirschboeck, 2003).6.3 Floods and Global <strong>Climate</strong> <strong>Change</strong>One of the main features of climate variationsin recent decades is the emergence of a packageof changes in meteorological and hydrologicalvariables that are consistent with global warmingand its impact on the hydrological cycle andthe frequency of extreme events (Trenberthet al., 2007, IPCC AR4, WG4, Ch. 3). Themechanisms underlying these changes includethe increase in atmospheric moisture, the intensityof the hydrologic cycle, and the changesin atmospheric circulation as the atmospherewarms (Knight et al., 2008). As described inone of the key findings of Gutowski et al. (2008;CCSP SAP 3.3, Ch. 3), “Heavy precipitationevents averaged over North America haveincreased over the past 50 years, consistentwith the increased water holding capacity of theatmosphere in a warmer climate and observedincreases in water vapor over the ocean.” (Seealso Easterling et al., 2000, Kunkel, 2003;Kunkel et al., 2003.) In addition, the frequencyof season-long episodes of greater-than-averageprecipitation is increasing (Kunkel et al., 2008;CCSP SAP 3.3, Sec. 2.2.2.3), and the timing ofsnowmelt is changing in many parts of the country(see Sec. 7). All of the meterological controlsof flooding (short- and long-duration heavyprecipitation, snowmelt) are thus undergoing111