125 Years of Hydrologic Change in the Puget Sound Basin: The Relative Signatures of Climate and Land Cover Dennis P. Lettenmaier Department of Civil and Environmental Engineering University of Washington Purdue Climate Change Research Center October 20, 2008

Outline of this talk The role of hydrology in Earth system science What are the grand challenges in hydrology? Understanding hydrologic change: The Puget Sound basin as a case study Modeling background The physical setting Analysis and prediction Weak links and some thoughts on the path forward

The role of hydrology in Earth system science “Where is the water, where is it going and coming from and at what rate, and what controls its movement?”

The land hydrologic cycle in a modeling context

A classical hydrological problem: Predicting runoff and streamflow given precipitation

Runoff generation mechanisms 1) Infiltration excess – precipitation rate exceeds local (vertical) hydraulic conductivity -- typically occurs over low permeability surfaces, e.g., arid areas with soil crusting, frozen soils 2) Saturation excess – “fast” runoff response over saturated areas, which are dynamic during storms and seasonally (defined by interception of the water table with the surface)

Infiltration excess flow (source: Dunne and Leopold)

Runoff generation mechanisms on a hillslope (source: Dunne and Leopold)

Saturated area (source: Dunne and Leopold)

Seasonal contraction of saturated area at Sleepers River, VT following snowmelt (source: Dunne and Leopold)

Expansion of saturated area during a storm (source: Dunne and Leopold)

What are the “grand challenges” in hydrology? From Science (2006) 125th Anniversary issue (of eight in Environmental Sciences): Hydrologic forecasting – floods, droughts, and contamination From the CUAHSI Science and Implementation Plan (2007): … a more comprehensive and … systematic understanding of continental water dynamics … From the USGCRP Water Cycle Study Group, 2001 (Hornberger Report): [understanding] the causes of water cycle variations on global and regional scales, to what extent [they] are predictable, [and] how … water and nutrient cycles [are] linked?

Important problems all, but I will argue instead (in addition) that understanding hydrologic sensitivities to global change should rise to the level of a grand challenge to the community.

In an era of global change … What are the impacts of land use and land cover change on river basin hydrology? What is the climatic sensitivity of runoff? What are the impacts of water management on the water cycle?

Understanding hydrologic change: The Puget Sound basin as a case study

Topography of the Puget Sound basin

The role of changing land cover – 1880 v. 2002

The role of changing climate, 1950-2000 source: Mote et al (2005)

Insert lan’s temperature plots here

Understanding hydrologic change: The modeling context Fundamental premises Simulation modeling must play a central role, because we rarely have enough observations to diagnose change on the basis of observations alone (and in the future, the “experiment” hasn’t yet been performed) If the hydrological processes are changing, we need to represent those processes Hence, prediction approaches that are “trained” to observations won’t work well

Distributed vs Spatially Lumped Hydrologic Models Smaller Sub-watersheds More realistic Processes Lumped Conceptual Fully Distributed Physically-based Streamflow (at predetermined points) Predictive skill limited to calibration conditions Streamflow Snow Runoff Soil Moisture, etc at all points and areas in the basin Predictive Skill Outside Calibration Conditions. Visual courtesy Andrew Wood

The Distributed Hydrology-Soil-Vegetation Model (DHSVM)

Lumped Conceptual (Processes parameterized) Explicit Representation of Downslope Moisture Redistribution Lumped Conceptual (Processes parameterized)

DHSVM Snow Accumulation and Melt Model

Representing urbanization effects in DHSVM Hydrologically relevant features of urbanization not found in “natural” watersheds: surface components such as streets, rooftops, ditches subsurface components such as pipes and other manmade stormwater drainage conduits In fully urbanized catchments, these elements are linked through street curb inlets and manholes In partially urbanized catchments, these urban drainage elements are often mixed with the natural channel drainage system

Modifications of DHSVM for urban areas For pixels with land cover category “urban”, a fraction of impervious surface area is specified. For the fraction that is not impervious, DHSVM handles infiltration using the same parameterizations as for non-urban pixels. A second parameter, the fraction of water stored in flood detention, was also added. Runoff generated from impervious surfaces is assumed to be diverted to detention storage. The runoff diverted to detention storage is allowed to drain as a linear reservoir, and re-enters the channel system in the pixel from which it is diverted. Surface runoff that is not diverted is assumed to enter the channel system directly, i.e., all urban channels are connected directly to the channel system We assume that the natural channel system remains intact, and we retain the support area concept that defines the connectivity of pixels to first order channels. However, impervious surface runoff (and drainage from detention reservoirs) is assumed to be connected to the nearest stream channel directly Once impervious surface runoff has entered a stream channel, it follows the “standard” DHSVM channel flow routing processes.

Springbrook Creek catchment

Springbrook Creek simulations (left) and errors (right) with and without urban module No impervious or detention Impervious, no detention impervious and detention

Springbrook Creek mean seasonal cycle simulated current land cover and all mature forest

Forest cover change effects Measurement of Canopy Processes via two 25 m2 weighing lysimeters (shown here) and additional lysimeters in an adjacent clear-cut. Direct measurement of snow interception

Calibration of an energy balance model of canopy effects on snow accumulation and melt to the weighing lysimeter data. (Model was tested against two additional years of data)

Understanding the effects of historical land cover and climate change on the Puget Sound basin – modeling and analysis

Study Areas Puget Sound basin, Washington State, USA Temperate marine climate, Precipitation falls in October – March Snow in the highland, rare snow in the lowland Targeted sub-basins

Tmin at selected Puget Sound basin stations, 1916-2003

Tmin at selected Puget Sound basin stations, 1916-2003

Tmin at selected Puget Sound basin stations, 1916-2003

Model Calibration

Land cover change effects on seasonal streamflow for eastern (Cascade) upland gages

Land cover change effects on seasonal streamflow for western (Olympic) upland gages

Land cover change effects on seasonal streamflow at selected eastern lowland (Greater Seattle area) gages

Land cover change effects by region

Land cover change effects on annual maximum flow at eastern (Cascade) upland gages

Land cover change effects on annual maximum flow at western (Olympic) upland gages

Land cover change effects on annual maximum flows at selected eastern lowland gages (greater Seattle area)

Predicted temperature change effects on seasonal streamflow at eastern (Cascade) upland gages

Predicted temperature change effects on seasonal streamflow at western (Olympic) upland gages

Predicted temperature change effects on seasonal streamflow at selected eastern lowland gages (greater Seattle area)

Predicted temperature change effects on seasonal streamflow by region

Predicted temperature change effects on annual maximum flow at eastern (Cascade) upland gages

Predicted temperature change effects on annual maximum flows for upland gages in the western (Olympic) region

Predicted temperature change effects on annnual maximum flow for selected lowland gages in the greater Seattle area

Comparison of temperature change and land cover change effects on annual maximum flows, selected upland and lowland basins

Model/observed residuals analysis of selected basins with long gauge records, annual maximum and annual mean flows

Model/observed residuals analysis of selected basins with long gauge records, annual maximum and annual mean flows

Projected future climate change: GCMs Models Institutions BCCR Univ. of Bergen, Norway CCSM3 NCAR, USA CGCM 3.1_t47 CCCma, Canada CGCM3.1_t63 CNRM_CM3 CNRM, France CSIRO_MK3 CSIRO, Australia ECHAM5 MPI, Germany ECHO_G Max Plank Institute for Mathematics, Germany GFDL_CM2_1 Geophysical Fluid Dynamic Laboratory, USA GISS_AOM NASA/GISS, (Goddard Institute for Space Studies) USA HADCM Met Office, UK HADGEM1 Hadley Center Global Environment Model, v 1., UK INMCM3_0 Institute Numerical Mathematics, Russia IPSL_CM4 IPSL (Institute Pierre Simon Laplace, Paris, France MIROC_3.2 CCSR/NIES/FRCGC, Japan PCM1

Projected Future Climate Conditions A1B Scenario Basins Tmin/Year (˚ C) Tmax/Year (˚ C) Prcp/Year (%) Tmin hist vs. future Tmax hist vs. future (˚C) Prcp hist vs. future (%) Deschutes 0.03 2.02 2.12 2.13 4.53 Cedar 0.04 1.90 1.88 1.91 3.24 Skokomish 2.16 2.04 2.05 6.59 Dosewallips 2.00 2.09 2.10 6.63 Lowland-west 2.03 2.11 6.45 Annual change rate in 2000 – 2099; Historical vs. future change: 2000 – 2099 vs. 1960 – 1999. Average of Models: Hadgem1, Echam5, Cnrm_cm3, Hadcm, Cgcm3.1_t47, Ipsl_cm4

Future climate change: multi-model ensemble simulation: A1B scenario

Summary Upland basin mean flow sensitivity to land cover change is mostly as a result of changes in snow accumulation and ablation, and lower ET associated with reduced vegetation. However, overall upland basin seasonal flows distribution, especially in the transient snow zone, are much more sensitive to temperature change effects – both to mean and peak flows – than to land cover change Lowland basin mean flows are much more sensitive to land cover change than are upland basins, especially in the most urbanized basins. Temperature change effects on peak flows in upland basins tend to be more modest than are changes in seasonal flows (suggests rain on snow effect on peak flows may be modest).

Weak links and some thoughts on the path forward (Over?) Reliance on models – the ideal design for land cover change studies is paired catchments, but … Are the data up to the challenge? Need to understand whether the model sensitivities are correct (not necessarily evidenced by calibration/verification errors) How do we use evolving (e.g. remote sensing) data sources in methods that are highly dependent on long record lengths?