1. Flood Hydroclimatology: Insights into Mixed Flood Populations Katie Hirschboeck Laboratory of Tree-Ring Research University of Arizona April 24, 2009

2. How do we transfer the growing body of knowledge about global and regional climate change and variability to individual watersheds to develop useful scenarios about hydrologic extremes? Key Question: Key Need: to understand the processes that deliver precipitation (or the lack thereof) to individual watersheds, at relevant time and space scales

3. 1. UNCERTAINTY: The Challenge of the “Upper Tails” 2. ASSUMPTIONS: The Standard iid Assumption for FFA 3. RE-THINKING: New Insights from “Flood Hydroclimatology” 4. ANTICIPATING THE FUTURE: Scenario building for a post-stationary world A “Story” in Four Chapters:

4. 1. UNCERTAINTY

5. SKEWED DISTRIBUTION Extreme events  tails of distribution Gaged Flood Record -- Histogram ( Standardized Discharge Classes) The Challenge of the “Upper Tails” Standardized Mean o = partial series  = annual series

6. Flow Time Series The gage was shut down in 1980 A fairly long record with lots of variability.... Flow Time Series The flood of October 1983! (WY 1984)

7. Santa Cruz River, Tucson Arizona Example The record flood of October 1983! Typical dry river bed or minor low flow vs. The Challenge of the “Upper Tails”

8. Flood Frequency Analysis: Theoretical Dilemmas (SOURCE: modified from Jarrett, 1991 after Patton & Baker, 1977)

9. ... can fail when “outlier” floods occur ! SOURCE: modified from Jarrett, 1991, after Patton & Baker, 1977 Curves A & B indicate the range (uncertainty) of results obtained by using conventional analysis of outliers for 1954 & 1974 floods. Pecos River nr Comstock, TX The Challenge of the “Upper Tails”

10. 2. ASSUMPTIONS

11. http://acwi.gov/hydrology/Frequency/B17bFAQ.html#mixed http://acwi.gov/hydrology/Frequency/B17bFAQ.html#mixed “Flood magnitudes are determined by many factors, in unpredictable combinations. It is conceptually useful to think of the various factors as "populations" and to think of each year's flood as being the result of random selection of a "population”, followed by random drawing of a particular flood magnitude from the selected population.”

12. “ iid ” assumption: independently, identically distributed The standard approach to Flood Frequency Analysis (FFA) assumes stationarity in the time series & “iid” The Standard iid Assumption for FFA

13. 3. RE-THINKING

14. Meteorological & climatological flood-producing mechanisms operate at varying temporal and spatial scales FLOOD-CAUSING MECHANISMS

15. Summer monsoon convective event Synoptic- scale winter event Tropical storm or other extreme event The type of storm influences the shape of the hydrograph and the magnitude & persistence of the flood peak This can vary with basin size (e.g. convective events are more important flood producers in small drainage basins in AZ) Storm type  hydrograph

16. HYDROCLIMATOLOGY  Weather, short time scales  Local / regional spatial scales  Forecasts, real-time warnings vs.  Seasonal / long-term perspective  Site-specific and regional synthesis of flood-causing weather scenarios  Regional linkages/differences identified  Entire flood history context  benchmarks for future events HYDROMETEOROLOGY

17. It all started with a newspaper ad.... Re-Thinking the “iid” Assumption

18. THE FFA “FLOOD PROCESSOR” With expanded feed tube – for entering all kinds of flood data including steel chopping, slicing & grating blades – for removing unique physical characteristics, climatic information, and outliers plus plastic mixing blade – to mix the populations together

19. Alternative Conceptual Framework: Time- varying means Time- varying variances Both SOURCE: Hirschboeck, 1988 Mixed frequency distributions may arise from: storm types synoptic patterns ENSO, etc. teleconnections multi-decadal circulation regimes

20. FLOOD HYDROCLIMATOLOGY is the analysis of flood events within the context of their history of variation - in magnitude, frequency, seasonality - over a relatively long period of time - analyzed within the spatial framework of changing combinations of meteorological causative mechanisms SOURCE: Hirschboeck, 1988

21. This framework of analysis allows a flood time series to be combined with climatological information To arrive at a mechanistic understanding of long-term flooding variability and its probabilistic representation.

22. APPROACH  Meteorological / Mechanistic / Circulation-Linked  Flood Hydroclimatology Framework / Link to Probability Distribution  “ Bottom–Up ” Approach (surface-to-atmosphere)  Observed Gage Record

23. WINTER & Seasonality of Peak Flooding

24. Flood Hydroclimatology Example Peaks-above-base: 30+ gaging stations in Arizona Synoptic charts + precipitation data  causal mechanisms

25. ANALYSIS Peaks-above-base -- 30+ gaging stations in Arizona Synoptic charts + precip data + decision tree  assigned causal mechanism / flood type Analyzed floods grouped by type -- spatially -- temporally / interannually

26. Sample Distributions of Gila Basin Gaged Peak Flows: Flood Hydroclimatology Example Are there climatically controlled mixed populations within?

27. Santa Cruz River at Tucson Peak flows separated into 3 hydroclimatic subgroups Hirschboeck et.al. 2000 Tropical storm Sumer Convective Winter Synoptic All Peaks

28. Remember the Santa Cruz record? What does it look like when classified hydroclimatically? What kinds of storms produced the biggest floods?

29. Hydroclimatically classified time series...

30. Hirschboeck et.al. 2000 Verde River below Tangle Creek Peak flows separated into 3 hydroclimatic subgroups Tropical storm Sumer Convective Winter Synoptic All Peaks

31. Historical Flood

32. Thinking Beyond the Standard iid Assumption for FFA.... Based on these results we can re- envision the underlying probability distribution function for Gila Basin floods to be not this....

33. Alternative Model to Explain How Flood Magnitudes Vary over Time Schematic for Gila River Basin based on different storm types Varying mean and standard deviations due to different causal mechanisms... but this:

34. IMPORTANT FLOOD- GENERATING TROPICAL STORMS Tropical storm Octave Oct 1983 Hurricane Lili Oct 2002 Tropical Storm Flood Events

35. When the dominance of different types of flood- producing circulation patterns changes over time, the probability distributions of potential flooding at any given time (t) may be altered. Conceptual Framework for Circulation Pattern Changes El Nino year La Nina year Blocking Regime Zonal Regime... or this:

36. Conceptual Framework for Low-Frequency Variations and/or Regime Shifts:... or this: A shift in circulation or SST regime (or anomalous persistence of a given regime) will lead to different theoretical frequency / probability distributions over time. Hirschboeck 1988

37. By definition extreme events are rare... hence gaged streamflow records capture only a recent sample of the full range of extremes that have been experienced by a given watershed. To fully understand flood variability, the longest record possible is the ideal... especially to understand and evaluate the extremes of floods and droughts! ADVANTAGES OF INTEGRATING THE PALEORECORD

38. Using Paleo-stage Indicators & Paleoflood Deposits... -- direct physical evidence of extreme hydrologic events -- selectively preserve evidence of only the largest floods...... this is precisely the information that is lacking in the short gaged discharge records of the observational period

39. Flood Frequency Analysis (SOURCE: Jarrett, 1991 after Patton & Baker, 1977) Curves A & B indicate range (uncertainty) of results obtained by using conventional analysis of outliers for 1954 & 1974 floods. Curve C is from analyses of paleoflood data. Q (discharge) uncertainty R.I. uncertainty Pecos River nr Comstock, TX

40. Compilations of paleoflood records combined with gaged records suggest there is a natural, upper physical limit to the magnitude of floods in a given region --- will this change? Envelope curve for Arizona peak flows

41. Historical Flood Largest paleoflood ( A.D. 1010 +- 95 radiocarbon date) 1993 FLOOD HYDROCLIMATOLOGY  evaluate likely hydroclimatic causes of pre-historic floods

42. 4. ANTICIPATING THE FUTURE

43. How do we transfer the growing body of knowledge about global and regional climate change and variability to individual watersheds to develop useful scenarios about hydrologic extremes? Key Question: Key Need: to understand the processes that deliver precipitation (or the lack thereof) to individual watersheds, at relevant time and space scales

44. Web-based “course” by UA’s Roger Caldwell: “Anticipating the Future” http://cals.arizona.edu/futures/ http://cals.arizona.edu/futures/ Represent Events by Simple Curves Question Assumptions Watch for Groupthink and Fixed Mindsets Expect Both Surprises & ‘Expected Results’ Several Solutions are Likely

45. MIXED POPULATION FAQ Question: “Floods in my study area are caused by hurricanes, by ice-affected flows, and by snowmelt, as well as by rainfall from thunderstorms and frontal storms. How do I determine whether mixed- population analysis is necessary or desirable?” Flood Hydroclimatology “in practice?”

46. “In practice, one determines whether the distribution is well-approximated by the LPIII by: -- comparing the fitted LPIII --- with the sample frequency curve defined by plotting observed flood magnitudes versus their empirical probability plotting positions... If the fit is good, and if the flood record includes an adequate sampling of all relevant sources of flooding (all "populations"), then there is nothing to be gained by mixed-population analysis.” Answer:

48. from Hirschboeck 2003 “Respecting the Drainage Divide” Water Resources Update UCOWR (Def): Interpolation of GCM results computed at large spatial scale fields to higher resolution, smaller spatial scale fields, and eventually to watershed processes at the surface. ONE APPROACH: DOWNSCALING

49. PROPOSED COMPLEMENTARY APPROACH:

50. RATIONALE FOR PROCESS-SENSITIVE UPSCALING: Attention to climatic driving forces & causes: -- storm type seasonality -- atmospheric circulation patterns with respect to: -- basin size -- watershed boundary / drainage divide -- geographic setting (moisture sources, etc.)... can provide a basis for a cross-scale linkage of GLOBAL climate variability with LOCAL hydrologic variations at the individual basin scale...

51. CONCLUSIONS Insights on Flood Hydroclimatology & Mixed Populations for Anticipating Future Floods

52. Mixed Distributions 1. Implications for predicting the tails of a distribution The distributions of key subgroups may be better for estimating the probability and type of extremely rare floods than the overall frequency distribution of the entire flood series. Suggestion: Separate out causes & linkages by stratifying by subgroup.

53. Hydroclimatic Regions 2. Implications for spatial homogeneity -- Basins can be grouped according to how their floods respond to different types of mechanisms and circulation patterns -- This grouping can change from season to season -- This grouping is also basin-size dependent

54. Non-Stationarity & iid Implications for time series homogeneity, stationarity & the iid assumption The conceptual framework of climate- driven time-shifting means, variances and/or mixed distributions provides a useful explanation for non-stationarity in flood times series and challenges the iid assumption.

55. For floods, climatic changes can be conceptualized as time-varying atmospheric circulation regimes that generate a mix of shifting streamflow probability distributions over time. This conceptual framework provides an opportunity to evaluate streamflow-based hydrologic extremes under climatic scenarios defined in terms of shifting modes or frequencies of known flood- producing synoptic patterns, ENSO, etc. Climatic Variability Implications for evaluating how flood time series may vary under a changing climate

56. PROCESS SENSITIVE UPSCALING: