Watershed Imperviousness and Peak Streamflow

Peabody, MA May 15, 2006

Paving Paradise: Watershed Imperviousness and Peak Streamflow Christiana Gerstner MS Thesis Environmental and Water Resources Engineering, Tufts University Advisors: Rich Vogel, Barbara Parmenter, Paul Kirshen Background Definition of imperviousness

– Surfaces that prevent natural infiltration of rain water – Examples: rooftops, roads, parking lots – Impervious Area (IA) is percentage (or fraction) of a basin’s drainage area that is covered by impervious surfaces Impact of impervious surfaces on hydrology

Runoff increases with imperviousness

Stream Corridor Restoration: Principles, Processes, and Practices (10/98). By the Federal Interagency Stream Restoration Working Group (FISRWG) Hydrologic effect of urban development

100-year flood has steadily increased from one decade to the next

Urban

Rural Two research objectives

1. Compare five common methods of estimating imperviousness at the watershed scale using data from Eastern Massachusetts

2. Explore the relationship between watershed imperviousness and peak streamflows in this region Research Part I

There are many methods for estimating impervious area (IA) at the watershed scale

How much do the results differ?

How accurate are the methods? Population density Remote sensing Land use Land cover Coefficients 26 Study basins in Eastern Massachusetts

Criteria: • Defined by USGS stream gages • Drainage area wholly within Massachusetts • Streamflow data available for 1996-2005

Watershed-scale, not site-scale Average drainage List of study basins area: 80 square miles. Ranked Contributing Watershed Station Drai nage Number ID Station Name Area (sq.mi.) 1 01095220 STILLWATER RIVER NEAR STERLING, MA 29.08 2 01095375 QUINAPOXET RIVER AT CANADA MILLS NEAR HOLDEN, MA 46.29 3 01109070 SEGREGANSET RIVER NEAR DIGHTON, MA 10.60 4 01101000 PARKER RIVER AT BYFIELD, MA 21.26 5 01094400 NORTH NASHUA RIVER AT FITCHBURG, MA 64.16 6 01105870 JONES RIVER AT KINGSTON, MA 20.03 7 01097300 NASHOBA BROOK NEAR ACTON, MA 12.68 8 01094500 NORTH NASHUA RIVER NEAR LEOMINSTER, MA 108.83 9 01109000 WADING RIVER NEAR NORTON, MA 43.40 10 01102000 IPSWICH RIVER NEAR IPSWICH, MA 124.96 11 01097000 ASSABET RIVER AT MAYNARD, MA 114.32 12 01108000 TAUNTON RIVER NEAR BRIDGEWATER, MA 261.34 13 01109060 THREEMILE RIVER AT NORTH DIGHTON, MA 84.34 14 01099500 CONCORD R BELOW R MEADOW BROOK, AT LOWELL, MA 399.75 15 01105933 PASKAMANSET RIVER NEAR SOUTH DARTMOUTH, MA 26.12 16 01104500 CHARLES RIVER AT WALTHAM, MA 250.33 17 01105000 NEPONSET RIVER AT NORWOOD, MA 34.79 18 01105730 INDIAN HEAD RIVER AT HANOVER, MA 30.10 19 01105500 EAST BRANCH NEPONSET RIVER AT CANTON, MA 27.31 20 01101500 IPSWICH RIVER AT SOUTH MIDDLETON, MA 44.50 21 01098530 SUDBURY RIVER AT SAXONVILLE, MA 105.96 22 01110500 BLACKSTONE RIVER AT NORTHBRIDGE, MA 139.98 23 01110000 QUINSIGAMOND RIVER AT NORTH GRAFTON, MA 25.52 24 01100600 SHAWSHEEN RIVER NEAR WILMINGTON, MA 36.42 25 01105600 OLD SWAMP RIVER NEAR SOUTH WEYMOUTH, MA 4.39 26 01102500 ABERJONA RIVER AT WINCHESTER, MA 24.77 Estimating imperviousness - five methods

Direct methods Inference methods

MassGIS Imperviousness NLCD land cover classification (Highest-quality method) plus coefficients developed by a Connecticut study National Land Cover Dataset (NLCD) imperviousness Population density plus equations developed by Moglen and Shivers

Population density plus equations developed by Stankowski

Study assumption: Imperviousness was stationary 2000-2005 Direct methods using remote sensing

Intersection of Route 1 and Dean Street in Norwood, MA

1m pixel based on 2005 imagery; 30m pixel based on 2001 imagery; binary (impervious yes/no) percentage (percent impervious) Inference method - NLCD Land cover with coefficients

Norwood town boundary 30m pixel based on 2001 imagery; (predominant land cover based on remote sensing) Inference methods - Population density

Norwood town boundary Moglen and Shivers (2006)

Stankowski (1972)

By census tract from 2000 census Range: from 3.9% Results to 35.1% impervious Estimated Percent Impervious Ranked NLCD Moglen/ Watershed Land Shivers Stankowski Number Watershed MassGI S NLCD Cover Pop Dens Pop Dens 1 Stillwater 3.9 1.4 5.9 4.9 4.5 2 Quinapoxet 6.0 4.1 7.5 6.6 6.3 3 Segreganset 7.1 3.9 7.0 7.5 7.3 4Parker 7.2 4.3 7.6 8.3 8.1 5 N. Nashua - Fitchburg 7.9 5.9 8.9 8.6 8.4 6 Jones 9.8 3.7 8.2 8.9 8.8 7 Nashoba Brook 10.7 7.2 10.8 8.8 8.6 8 N. Nashua - Leominster 11.3 10.5 12.2 10.8 10.8 9 Wading 11.6 9.2 11.3 9.9 9.8 10 Ipswich 11.6 11.5 12.7 11.4 11.5 11 Assabet 11.9 11.2 12.9 10.6 10.7 12 Taunton 12.1 9.7 11.6 11.7 11.9 13 Threemile 12.6 10.7 12.3 10.5 10.6 14 Concord 13.5 13.1 14.2 12.4 12.6 15 Paskamanset 14.1 13.2 14.2 12.7 12.9 16 Charles 14.8 13.6 14.5 13.9 14.2 17 Neponset 16.0 15.1 15.6 13.8 14.1 18 Indian Head 16.2 14.9 15.4 13.4 13.7 19 E. Br. Neponset 16.3 18.2 17.6 14.7 15.1 20 Ipswich 16.4 17.5 17.3 14.0 14.3 21 Sudbury 17.1 17.2 16.9 14.7 15.1 22 Blackstone 18.9 18.9 18.0 16.6 17.1 23 Quinsigamond 21.0 25.5 22.6 17.5 18.1 24 Shawsheen 21.9 25.9 23.4 14.6 15.1 25 Old Swamp 24.3 26.0 23.3 16.8 17.4 26 Aberjona 35.1 41.8 33.8 22.2 23.0 Comparison of estimates 45 Watershed Imperviousness Estimates 40 The estimates 35 diverge the most at higher levels of 30 imperviousness

o 25 The study assumed 20 the MassGIS estimate to be accurate 15 Percent Impervi 10

5

0 0 5 10 15 20 25 30 Watershed NLCD Imperviousness NLCD Land Cover MassGIS Population Density M/S Population Density St % Error in estimates 60 At low levels of Error in Imperviousness Estimates development, error As compared with MassGIS Estimate as high40 as 60% NLCD Land Cover Population density method performed methods underpredict at 20 the best high levels, overpredict at low levels

0 0 5 10 15 20 25 30 35 40 % NLCD Imperviousness-20 underpredicts at low

levels, ( Error overpredicts at high levels -40

-60 NLCD Imperviousness NLCD Land Cover Population Density M/S Population Density St -80 Impervious Area (%) Research Part II

What is the relationship between watershed imperviousness and peak streamflows in Eastern Massachusetts? Previous studies

USGS Urban Flood Frequency Studies – equations for estimating peak discharge based on drainage area and imperviousness

Equations for estimating the T-year peak discharge in ungauged urban basins in some states* take the form:

b cOR b c QT = aA IA QT = aA (1+IA) Where A is drainage area IA is percent impervious cover

The exponent c can be called the “urban elasticity”

*AL, GA, MO, NC, WI plus national equations developed by Moglen/Shivers 1.2 Meta-analysis of Urban Elasticity 1 Imperviousness has less impact on larger less- 0.8 frequent floods, like the 50- year flood

0.6 NC MO 0.4

Elasticity GA Elasticity is the exponent c AL in this equation for WI 0.2 predicting floods of a given frequency T Milwaukee

0 0 25 50 75 100 Frequency (years) Imperviousness has the greatest impact on smaller b c Q = aA IA more-frequent floods, like T Imperviousness increases the 2-year flood Where peak streamflows A is drainage area (elasticities are positive) IA is percent impervious cover Streamflow data

Annual instantaneous peak discharge 1996-2005 80 Comparing Peak Unit Discharge - Urban vs. Rural

60 Parker River (7% Impervious) Aberjona River (35% Impervious)

40

20 Unit Peak Discharge (cfs/sqmi) Discharge Peak Unit

0 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 Year Streamflow analyses

Regression analysis to see if Correcting for other factors: peak streamflow could be •Climate/precipitation norm predicted based on area and •Surface storage percent imperviousness for these basins. Using other streamflow statistics: Results were inconclusive. •Daily mean, exceedance probability .01, .02, .05, .10, .20 Peak streamflow can be predicted from area alone but accounting for percent Focusing study: imperviousness does not •Basins < 100 square miles improve the fit. •Basins < 20% surface storage •Sub-regions Issues complicating the streamflow analysis

Other factors affecting peak Methodology: streamflow: •Short time series – only 10 •Baseflow/groundwater years of streamflow data •Connected vs disconnected •Actual vs. synthetic discharge imperviousness data •Low-impact development •Annual vs. event-specific discharge •Insufficient data at high end of imperviousness Conclusions

• Quantified variation and error in imperviousness estimates at the watershed scale • Could not confirm a simple correlation between imperviousness and the 2-year flood in Eastern Massachusetts 1996-2005 Questions