Limitations to vegetation establishment and growth inbiofiltration swales
Greg Mazer a,*, Derek Booth b, Kern Ewing c
a URS Corporation, 500 Market Place Tower, 2025 First A�enue, Seattle, WA 98121, USAb Department of Ci�il and En�ironmental Engineering, Center for Urban Water Resources Management, Uni�ersity of Washington,
Box 352700, Seattle, WA 98195, USAc Center for Urban Horticulture, Uni�ersity of Washington, Box 354115, Seattle, WA 98195-4115, USA
Received 27 December 1999; received in revised form 29 September 2000; accepted 6 November 2000
Nonpoint source water pollution is consideredto be the major cause of US surface water qualityimpairment (USEPA, 1997). Runoff from urbanand suburban areas has become an increasinglycommon source of nonpoint source pollution and
G. Mazer et al. / Ecological Engineering 17 (2001) 429–443430
its associated habitat degradation. Acknowledgingthis problem, several governmental agencies, in-cluding the USEPA, have advocated constructionof passive storm-water filtration facilities as a bestmanagement practice (Roesner et al., 1999; Clay-tor and Schueler, 1996). One such facility fre-quently used in the Puget Sound region is thebiofiltration swale (also called biofilter orbioswale). Over 100 bioswales have been con-structed in King County over the past 10 years totreat runoff associated with residential, commer-cial, and light industrial development.
Bioswales are vegetated channels designed totreat pollutants commonly occurring in storm-wa-
ter runoff (Fig. 1). Storm water delivered by anetwork of storm sewers enters the channel at theinlet and receives treatment before it exits theoutlet. Bioswales are generally at least 30 m (100ft) long, 0.6 m (2 ft) wide, range in longitudinalslope from 0.5% to 6%, and located in series withdetention ponds, which store runoff and reducepeak discharges. Although they are designed toconvey runoff from the 100-year 24-h stormevent, they are only intended to treat runoff effec-tively from much smaller and more frequentstorms, typically up to the 2-year 24-h stormevent (Ecology, 1996; King County, 1998).Bioswales are commonly confused with grassedwaterways (also called swales), which are vege-tated channels that convey, but do not necessarilytreat, runoff (Novotny and Olem, 1994).
Pollutants in urban runoff, which include sedi-ments, nutrients, metals, synthetic organics,pathogens, and hydrocarbons, are sequestered inthe biofilter soil, sediment deposits, organic litter,and standing vegetation. Some pollutants, such assynthetic organic compounds, may then be trans-formed into less harmful substances via microbialdecomposition. All captured contaminants maybe removed when bioswale vegetation and/or soilare removed.
The treatment goal for bioswales in KingCounty is 80% removal of total suspended solids(TSS), given typical TSS concentrations at theinlet (King County, 1998). Treatment efficienciesdocumented for bioswales and sections of grassedwaterways (including roadside swales) have beendocumented to be 60–99% removal of TSS, 21–91% removal of metals, and 7.5 to more than 80%removal of total phosphorus (Kercher et al., 1983;Oakland, 1983; Harper et al., 1984; Yousef et al.,1985; Khan et al., 1992; USEPA, 1983; Goldberget al., 1993; King County, 1995; Claytor andSchueler, 1996; Walsh et al., 1997). Treatmentefficiency greatly depends upon inflow rate andpollutant concentration, both of which tend tovary considerably in storm-water runoff (Novotnyand Olem, 1994; Tarutis et al., 1999).
Herbaceous cover in bioswales is generally con-sidered to be well correlated with bioswale treat-ment efficiency. Aboveground plant parts (stems,leaves and stolons) are thought to induce sedi-
Fig. 1. Plan view and profile view schematic of typicalbioswale (not to scale).
G. Mazer et al. / Ecological Engineering 17 (2001) 429–443 431
mentation of particulates and their sorbed pollu-tants while plant roots stabilize sediment deposits,preventing sediment re-suspension (Horner et al.,1994; Claytor and Schueler, 1996; Kadlec andKnight, 1996). As most pollutants in urban runoffare in particulate form or adsorbed to sedimentparticles, sedimentation is believed to be the pri-mary means by which vegetated control facilitiesimprove runoff water quality (Horner et al., 1994;Claytor and Schueler, 1996). Vegetation and or-ganic litter may uptake or absorb dissolved pollu-tants, but bioswales do not appear to capturedissolved pollutants very effectively (Khan et al.,1992; Horner et al., 1994). Regular (annual ormore frequent) via mowing would permanentlyremove dissolved pollutants captured by vegeta-tion and incur other benefits, such as increasedflow resistance. However, mowing is costly andclippings are typically left in the biofilter, allowingpollutants to be released upon plant decomposi-tion (Schultz, 1998).
A recent study found most bioswales in KingCounty, Washington to be vegetationally depau-parate and thus functionally inadequate (KingCounty, 1995). Several environmental factorswere blamed for the frequently poor vegetativecover in bioswales including prolonged inunda-tion, high flow velocity, large fluctuations in sur-face water depth and soil moisture, excessiveshade, poor soil, and improper installation. Therelative importance of these limiting factors mayvary widely amongst bioswales, and they can be aresult of poor design, poor construction, and/orinsufficient maintenance.
Our study monitored bioswales that representthe range of hydrologic condition and vegetationcomposition and abundance found in bioswales inwestern Washington. Seven of the eight swalesexamined in this study were built using the 1990King County Surface Water Design Manual(King County, 1990) guidelines. For flows pro-duced by the 2-year 24-h storm event, this manualspecifies maximum permitted water depth of 10cm (4 in.), flow velocity at 0.3 m/s (1.0 ft/s), anda design Manning’s n of 0.035. No guidelines forinflow pollutant concentration limits or hydraulicloading rate (HLR) are provided by this manual.
Our study focused on the environmental factorsthat most influence vegetation establishment andabundance in bioswales. It also investigated as-pects of bioswale design and the assumed directcorrelation between herbaceous vegetation abun-dance and bioswale treatment efficiency. The prin-cipal objective of this research has been todevelop recommendations for bioswale designthat will improve their pollutant-removalperformance.
2. Methods
2.1. Retrofit
Retrofits of three sparsely to moderately vege-tated bioswales (SAY7, SAY8, and SAY9) wereconducted to determine whether vegetative covercould be improved without altering bioswale hy-drologic regime. The three swales, all located atthe Saybrook Estates development in northernKing County, were retrofitted in early fall 1996.Swale channels were widened to 1.6–1.8 m bytrackhoe excavation. The top 15 cm of soil wasremoved and replaced with sandy loam. Soil lay-ers were rototilled 8 cm below final grade andagain at the soil surface. The bioswale soil wasthen rolled with a static roller to enhance erosionresistance. Although the original longitudinalslopes were retained, previous irregularities (dipsand mounds) were smoothed.
The grasses Agrostis stolonifera (creeping bent-grass), Festuca arundinacea (tall fescue), Poapratensis (Kentucky bluegrass), Alopecurus genic-ulatus (meadow foxtail), and Festuca o�ina (sheepfescue) were hydroseeded on September 23, 1996.These species are all perennial turf and foragegrasses that are native to or naturalized in thePacific Northwest. Seeds were spread at a rate of4 kg per 100 m2 (8 lbs. per 1000 ft2). Due tosubsequent storm-induced erosion, SAY8 was hy-droseeded a second time on October 1, 1996. Tofurther prevent flow-induced erosion, hay baleswere placed every 8 m (25 ft) down the length ofthis swale. SAY9 was also re-seeded due to poorinitial establishment on August 18, 1997 by hand(using the same seed mix) after its water level had
G. Mazer et al. / Ecological Engineering 17 (2001) 429–443432
Fig. 2. Vicinity map showing bioswale locations.
dropped below the soil surface. Each swale’s plantspecies composition and cover were recorded im-mediately before and 1 year following the retrofit.The details of this sampling methodology arediscussed in the next section.
2.2. Field sur�ey
Environmental factors thought to strongly infl-uence vegetation establishment and abundance inbioswales were examined in eight bioswales, in-cluding the three that were retrofitted. Theseswales represent a wide spectrum of hydrologic,soil, and vegetative conditions and are located inthree separate areas within King County (Fig. 2).
Bioswale dimensions ranged between 29 and 84m length, 0.7–3.7 m width, and 0.23–1.95%slope. Three bioswales (PLP, PLEa, and PLEb)had check dams of crushed rock spaced 8 m (25ft) apart down the channel length. The porousnature of the dams allowed seepage, albeit atrelatively slow rates; thus, they were rarely over-topped during storm events. All swales, exceptthose at Discovery Elementary (DISC) and theCenter for Urban Horticulture (CUH), were situ-ated immediately downstream of detention ponds.
Most swale conditions were measured at dis-crete stations (Fig. 3). The first station was situ-ated 10 m (33 ft) from the swale inlet. Subsequentstations were spaced 15 m (50 ft) apart down thelength of the swale. Due to the wide range in
G. Mazer et al. / Ecological Engineering 17 (2001) 429–443 433
bioswale length, the number of measurement sta-tions per bioswale varied from 2 to 5.
Relative cover by species was measured in Juneand September, 1997. Two adjacent 0.25 m2
quadrats were used at each sampling station tovisually estimate plant cover. In September 1997,all aboveground vegetation and surface organiclitter contained within an open cylinder of diame-ter 0.12 m and height 0.1 m placed in the center ofeach quadrat was removed. This material wasthen oven-dried and weighed to determine eachquadrat’s aboveground biomass per unit area.
Aboveground biomass harvest was restricted tothe layer of vegetation and organic litter occurringbetween the soil surface and 0.1-m height. Sam-pling this layer emphasizes its disproportionatelystrong influence on sediment trapping and pollu-tant decomposition and avoids having resultsskewed by plants, such as cattail (Typha latifolia)whose greater height provides no additional treat-ment. Biomass below 0.1-m height was sampledinstead of stem density because a substantial por-tion of the species found were herbs rather thangrasses; herbs may produce abundant above-ground biomass despite exhibiting low stemdensity.
Peak and instantaneous surface water depthswere monitored weekly for 6- to 7-week periodsduring spring (April–May) and summer (Au-gust–September) of 1997 with crest-stage gaugesplaced on the upstream edge of each samplingstation. Water depth data for all swales except
PLEb and CUH were recorded during a 6-weekperiod in January and February, 1996, and com-pared with hydrologic data for the spring sam-pling period. The interval of time a swale isinundated above 25 mm was measured weeklywith the Inundation Sensor and Integrator (ISI),an electronic gauge developed by the senior au-thor and Dr. Robin Cleveland, professor of me-chanical engineering at Boston University. TheISI’s timer is activated by the presence of water ator above 25-mm depth and de-activated whenwater fell below this depth. During weekly moni-toring events, total time of inundation wasrecorded and the timer was reset. Soil moisturepotential was assessed at each plot during thesummer sampling period. Soil depth, percentgravel content, and bulk density were assessedafter harvesting vegetation in September 1997.
Flow resistance coefficients (Manning’s n) andstage–discharge relationships were established foreach swale during storm flows. Flow widths weremeasured with a tape measure, flow depths weremeasured by the crest-stage gauge, and flow ve-locity was estimated by repeatedly timing thepassage of a float through a measured distance.Both discharge and flow velocities were calculatedfor the mean peak, mean instantaneous, andhighest peak (‘maximum peak’) flow depths of thespring and summer sampling period. Eachbioswale’s flow character and sedimentation po-tential were estimated by calculating the hydraulicresidence time (HRT, the time required for analiquot of water to travel from inlet to outlet) andhydraulic loading rate (HLR, the ratio of inflowdischarge at the 10 m gauge to bioswale area).
County inspection records were examined todetermine construction date, design discharges,and seeded species composition for all bioswalesexcept the CUH swale, for which no records werekept. The NOAA 2-year 24-h isopluvial map wasused to estimate the precipitation amount ex-pected during the largest storm that bioswales arerequired to treat. Daily precipitation data takenfrom King County Land and Water ResourcesDivision gauges at 1900 228th Ave NE (‘MLU’)was extrapolated for the Saybrook swales, each ofwhich is within 2 km of this gauge. Data from theKing County Land and Water Resources Division
Fig. 3. Spatial distribution of bioswale measurement stations(not to scale).
G. Mazer et al. / Ecological Engineering 17 (2001) 429–443434
gauge on Old Black Nugget Road (‘464’) wasextrapolated for all other bioswales except CUH.Daily precipitation data from a University ofWashington campus rain gauge was used to esti-mate rainfall in the CUH bioswale catchment.
2.3. Greenhouse study
The greenhouse study tested the differentialresponses of four turfgrass species commonlyseeded in bioswales to varying hydrologic regimesin a nested, 4×4 factorial experiment. Seeds offour grass species were placed in small pots (toplength and width=57 mm, depth=83 mm) con-taining 30 mm of mulch/tackifier atop 40 mm ofsoil-less media (50% pumice, 35% peat, and 15%fine bark). The species used were the same asthose seeded in the retrofit minus F. ovina. Ninereplicates for each species/treatment combinationproduced a total of 144 pots. The hydrologicregime treatments were as follows:� ‘Control’……………no inundation; media kept
moist, but not saturated throughout theexperiment
� ‘Dry’………………. 2 days inundated, 12 dayswith no watering
� ‘Intermediate’………7 days inundated, 7 dayswith no watering
� ‘Wet’……………….12 days inundated, 2 dayswith no wateringInundation consisted of flooding pots 2–4 cm
above the soil surface for continuous periodswithin two 14-day cycles. To gauge seed andseedling response to the treatments, abovegroundbiomass and leaf blade density were measured.Vegetation above the soil surface in each pot washarvested, oven-dried, and weighed at the experi-ment’s end. Leaf blades per pot were countedtwice per week throughout the experiment.
3. Results
3.1. Retrofit
Of the three bioswales retrofitted, only SAY7developed a continuous and dense cover ofgrasses. The level of vegetation abundance and
distribution attained would be judged adequatefor effective biofiltration by King County andother agencies. Within two weeks of hydroseed-ing, SAY7 supported 10- to 50-mm tall F. arundi-nacea seedlings down the length of the swale. Thisspecies dominated vegetation throughout thestudy period, though A. stolonifera eventually es-tablished throughout the swale despite not germi-nating until eight months after hydroseeding. Asmall cluster of A. geniculatus established near theoutlet where soils appeared slightly more moist.Neither of the other species (P. pratensis and F.o�ina) seeded germinated in this swale. Meanherbaceous vegetation cover in SAY7 increasedfrom 41% in September 1996 (immediately beforethe retrofit) to 98% in September 1997 (1 yearafter the retrofit). The average vegetation andorganic litter biomass accrued in SAY7 was com-parable to that of the other swales observed toexhibit high herbaceous cover even though theseswales had been seeded 3–9 years beforehand.
The other two retrofitted swales, SAY8 andSAY9, attained only minimal germination due topersistent high flows (SAY8) or very persistentinundation (SAY9). Continuous base flow downthe nearly 2% slope in SAY8 scoured soil andseeds alike. Drainage from the upstream retentionpond appears to have deposited seeds of emergentherbaceous plants (e.g. Alisma plantago-aquatica)that subsequently established in shallow, low-en-ergy areas. However, these areas were neitherabundant nor densely vegetated. The much shal-lower slope of SAY9 (0.23%) impeded drainage,minimizing scour but prolonging inundation. There-seeding on August 18, 1997 produced numer-ous seedling patches in the less shaded part of theswale (0–30 m). Except for a few A. stoloniferaand A. geniculatus seedlings, however, most of theseeded vegetation did not survive throughNovember 1997. Prolonged inundation combinedwith lower seasonal daylight appears to have sup-pressed seedling growth and establishment.
As the original seed mix was selected to toleratea wide range of hydrologic conditions, it wasexpected that some species would establish insome bioswales more readily than in others. InSAY7, F. arundinacea was the first species togerminate and became the dominant species in the
G. Mazer et al. / Ecological Engineering 17 (2001) 429–443 435
Fig. 4. Mean plant and organic litter biomass per bioswale.
among swales, this test used a harmonic meansample size of 3.45. SAY8, SAY9, PLEa, andPLEb are in the ‘low-biomass group’ (sparselyvegetated); SAY7, PLP, and CUH are in the‘high-biomass group’ (well vegetated); and DISCbelongs to both groups. Plant cover values werestrongly correlated with plant and organic litterbiomass (r2=0.90), though cover tended to over-estimate biomass in the mid-range values (Fig. 5).
Though nearly all of the biomass collected fromSAY7 was derived from live grasses, organic litterand dead-standing vegetation constituted a major-ity of the total biomass in the other three swalesof the high-biomass group. Most of the biomasscollected from the swales in the low-biomassgroup appeared to be composed of organic litter.
3.2.2. Hydrologic and hydraulic characteristics ofthe bioswales
The hydrologic data revealed certain seasonalpatterns common amongst all the bioswales sur-veyed in this study. One-tailed paired t-testsshowed that mean peak and maximum peakdepths were substantially, but not significantly,higher in winter than in spring. Slight significance(P�0.09) was shown for differences between win-ter and spring instantaneous water depth. Allmeasures of flow were significantly greater inspring than in summer.
swale. Upon establishing, it appeared to readilytolerate periods of surface water flow during thewinter and the dry conditions that occurred dur-ing summer. Very few turf grasses grew in theother two retrofit swales, because conditions wereso adverse to their establishment.
ences amongst the eight bioswales for mean plantand organic litter biomass. The Student New-man–Keul’s post-hoc test distinguished twogroups (Fig. 4). As the number of plots varied
Fig. 5. Mean plant cover as a function of mean plant biomass. r2=0.90, n=8.
G.
Mazer
etal./
Ecological
Engineering
17(2001)
429–
443436
Table 1Comparison of bioswale mean peak flow hydraulics, bioswale dimensions, presence of check dams, and vegetation abundance category
Spring sampling period Summer sampling periodBioswale
CheckMean peak SwaleMeanMean peak Slope (%)MeanMean Mean peakMean peak Vegetationflow andDamsinstantaneou water depthflowwater depth swale widthinstantaneou length (m)velocity(mm)s watervelocity (m)(mm) s water Organic(m/s)depth (mm) depth (mm)(m/s) Litter
BiomassCategory
44 0.15 2 2 0.02 0 1.9 1.8SAY7 65 no high70 0.20 28 39 0.14SAY8 24 1.9 1.8 50 no low59 0.10 40 39 0.08 11 0.2 1.9 50SAY9 no low88 0.13 6 104 0.14DISC 11 1.8 2.0 84 no high
However, each bioswale also has unique hydro-logic and hydraulic characteristics determined byphysical swale dimensions (e.g., longitudinalslope, presence of check dams, etc.), the presenceof groundwater, and the size and imperviousnessof the drainage area. Some aspects of bioswalehydraulics and physical dimension are comparedwith vegetation abundance category in Table 1.Hydraulic data collected during the maximumpeak flow events of the spring sampling period aregiven in Table 2. Hydraulic data collected duringthe maximum peak flow events of the springsampling period are given in Table 3.
The maximum peak flow events for both thespring and summer sampling periods occurredduring storm events of much less intensity thanwhat is regarded as the maximum treatable (2-year 24-h) for biofiltration according to KingCounty (1990, 1998). Maximum peak flow duringthe spring sampling period occurred on 31 May,1997 for most of the bioswales (SAY8, DISC,PLEa, PLEb, and CUH). Rainfall during this24-h period varied between 51 and 58 mm, only57–73% of the expected 2-year 24-h amount (90mm). During the previous week, 22–36 mm ofprecipitation fell; this amount is only slightlyhigher than average weekly rainfall for the springsampling period, which ranged from 17 to 29 mm.For the three other bioswales, precipitation dur-ing the week of 26 April–2 May produced themaximum peak water flow. Although the greatest
24-h precipitation during this week was not veryhigh (18–30 mm), the weekly rainfall total rangedfrom 63 to 68 mm, which is over twice the averageweekly total.
Despite these relatively modest rainfall totals,five of eight swales exhibited water depths duringtheir respective maximum peak flows that weregreater than the permitted maximum depths forthe 2-year 24-h storm and thus did not meet KingCounty (1990, 1998) performance standards.None of the swales exceeded the maximum per-mitted discharge (0.14 m3/s) and only one swale(SAY8) exceeded the maximum permitted flowvelocity (0.3 m/s). However, 4 swales had HRTsless than the recommended minimum (9 min) and2 of these swales had HRTs less than the requiredminimum (5 min).
Only the CUH swale consistently exhibitedboth surface water depths and HRTs within thebounds of what is recommended for the 2-year24-h storm event. Ironically, this is the onlybioswale that was not specifically designed tomeet King County (1990) standards. Disparitiesbetween storm-water facility design and perfor-mance have also been recognized by Booth andJackson (1997), who report substantial differencesbetween model predicted and actual dischargesfrom detention ponds, and by Colwell (2000),who found that measured Manning’s n values inbioswales commonly differed greatly from designspecifications.
G. Mazer et al. / Ecological Engineering 17 (2001) 429–443438
Table 3Comparison of bioswale maximum peak flow hydraulics during the summer sampling period
3.2.3. Correlations of bioswale biomass with en�iron-mental factors
Both field observations and greenhouse experi-mentation indicated a strong influence by hydro-logic regime over herbaceous plant growth.However, hydrologic regime appears to have littleinfluence in areas with chronically low light. Theplots that were heavily shaded during the growingseason (the 40-m plot at SAY9 and all plots atPLEb) supported very little standing vegetationand received much of their organic litter fromoverhanging trees. Because heavy shade appearedto override other environmental factors and greatlyreduce bioswale vegetation and organic litterbiomass, these plots were removed from the subse-quent correlation analyses.
For the remaining plots, summer inundationpersistence (IP) was the hydrologic variable withthe most influence over bioswale plant and organiclitter biomass (r2=0.97) (Fig. 6). Persistent inun-dation during the summer sampling period, span-ning multiple days, appeared to profoundlysuppress swale biomass. Bioswales that were inun-dated for more than 35% of the time during summercontained significantly lower vegetation and or-ganic litter biomass. In contrast, spring IP is onlyweakly correlated to bioswale biomass (r2=0.18).
Mean instantaneous water depth during bothspring and summer related had moderately strongnegative correlations to bioswale biomass (r2=0.56, P�0.1 and r2=0.62, P�0.05, respectively).The correlation between swale biomass and springmean instantaneous flow velocity was significant
(r2=0.56, P�0.1). This relationship is collinearwith the one between spring instantaneous depthand swale biomass since velocity increases expo-nentially with depth.
Although swale vegetation and organic litterbiomass had no significant correlations with any ofthe physical soil variables measured (soil depthr2=0.09, bulk density r2=0.20, and relative gravelcontent r2=0.12), soil depth was important tothose plots exposed to summer drought (wheremean soil moisture potential � −15 Mpa). Vege-tation and organic litter biomass in these plotsincreased significantly with soil depth (r2=0.74,n=10).
The relationship between bioswale biomass andthe HLR calculated for mean instantaneous waterlevels (Fig. 7) is similar to the relationship betweenbioswale biomass and summer period IP. However,
Fig. 6. Mean bioswale vegetation and organic litter biomassrelated to mean weekly IP during the summer period. r2=0.97, n=7.
G. Mazer et al. / Ecological Engineering 17 (2001) 429–443 439
Fig. 7. Mean bioswale vegetation and organic litter biomasrelated to instantaneous HLR during the spring samplingperiod. r2=0.87, n=7.
Fig. 8. Mean bioswale vegetation and organic litter biomasrelated to maximum peak HLR during the spring samplingperiod. r2=0.16, n=7.
HLRs calculated for maximum peak water levelshave only a weak relationship with swalebiomass (Fig. 8). The difference between theinfluence of mean instantaneous HLR versusmaximum peak HLR upon vegetation and or-ganic litter biomass occurs in both spring andsummer data.
Within each species, final abovegroundbiomass and leaf blade accumulation washighest in ‘Control’ pots, where the media waskept continuously moist, but free from inunda-tion. Final aboveground biomass and leaf bladeaccumulation was lowest in the ‘Wet’ pots,where inundation at 2–4 cm above the soil sur-face was maintained for 12 of 14 days for two14-day treatment cycles (Fig. 9). The ‘Wet’treatment produced equally minimal germinationand growth amongst all species. For the ‘Con-trol,’ ‘Intermediate,’ and ‘Dry’ treatments, F.arundinacea produced significantly (P�0.001 foreach treatment) more biomass than the otherspecies tested. For each of these three treat-ments, A. stolonifera possessed significantly moreleaf blades than the other species during the lastthree-quarters of the experiment. For all species,the largest increases in leaf blade number oc-curred during the times when pots were freefrom flooding.
4. Discussion
4.1. En�ironmental limitations to bioswale�egetation
Results of the field survey and the greenhouseexperiment demonstrate that persistent, multi-day inundation severely limits germination andgrowth by grasses typically seeded in bioswales.In the field survey, persistence of inundationduring the driest time of year (summer) was in-versely related to bioswale vegetation and or-ganic litter biomass. For all but the deeplyshaded bioswale plots, the less inundation above2.5-cm depth occurred during this period, thegreater the vegetation and organic litterbiomass. The relatively drier conditions in sum-mer served as a window of opportunity forbioswale vegetation, allowing seed germinationand encouraging plant growth.
For the grass species tested in the greenhousestudy, persistent inundation was shown to sig-nificantly suppress germination and seedlinggrowth. Conversely, aboveground biomass wasgreatest in pots that were kept moist but freefrom inundation (the ‘Control’ treatment).
Inundation severely inhibits germination andgrowth of plant species ill-adapted to living in
G. Mazer et al. / Ecological Engineering 17 (2001) 429–443440
Fig. 9. Final dried biomass per species and treatment (+1 standard error indicated by error bars).
frequently flooded conditions. However, even plantspecies well adapted to inundation as mature indi-viduals may be prevented from establishing in areaswhere inundation is persistent (Kozlowski, 1984;Ernst, 1990; Crawford, 1992, 1996; Ewing, 1996).This may partially explain why wetland areas withpermanent standing water typically have bothlower plant density and species richness than wet-land areas with only seasonal or intermittent stand-ing water (van der Valk et al., 1981; Emers, 1990;Cooke and Azous, 1993; Nielsen and Chick, 1997).
Bioswale vegetation abundance does not appearto be diminished by highly erosive flow, as long assuch flows and/or inundation do not persistthrough the growing season. SAY7, DISC, andSAY8 all have relatively steep longitudinal slopes(�1.5%) and high maximum peak flow velocitiesduring spring (�0.2 m/s). Yet only SAY7 andDISC support dense herbaceous vegetation be-cause unlike SAY8, these swales do not haveyear-round base flow. In the short-term (�1–14days), flowing water may be only slightly moreharmful to grasses than standing water of the sameduration (Temple, 1991).
Although often overlooked, light and soil condi-tions are also essential factors influencing vegeta-tion abundance in bioswales. Germination andvigor of herbaceous plants is typically reliant upon
at least moderate exposure to light (Gabriell, 1997).Those plots with heavy shading demonstrated con-sistently poor vegetation growth despite havingother factors beneficial to vegetation abundance.Soil provides structural support and storage fornutrients, minerals, and water (Brady and Weil,1996). Effective rooting depth appears especiallyimportant to vegetation survival in bioswales, anenvironment potentially subject to erosive flow andwide fluctuations in soil moisture.
4.2. Hydraulic indicators of bioswale performance
Given the wide variability of precipitation, pollu-tant removal efficiency in bioswales may be esti-mated by hydrologic characteristics, such as HRTand HLR better than by vegetation cover. Horneret al. (1994) state that bioswales require at least9-min HRT for adequate pollutant attenuation.However, HLR may be a more appropriate surro-gate for estimating treatment performance as itincorporates more complete information on chan-nel length, channel width, and inflow dischargerate.
The hydraulic loadings that bioswales typicallyreceive during storm events are far greater thanthose that allow effective treatment in overlandflow wastewater treatment facilities. Typical HLR
G. Mazer et al. / Ecological Engineering 17 (2001) 429–443 441
for overland flow wastewater treatment facilities,which are similar to bioswales in structure andintended function, is 0.01–0.1 m/d (Kadlec andKnight, 1996). This is approximately 3–4 orders ofmagnitude lower than the HLR that would occurin a bioswale with minimum dimensions (30-mlength and 0.6-m width) under the maximum flowvelocity (0.3 m/s) and water depth (0.1 m) permittedby King County (1990, 1998) Surface Water Man-ual for the 2-year 24-h storm event.
As mentioned above, the maximum peak flowevent for all swales occurred during storm eventsof much less magnitude than the 2-year, 24-h stormevent. Yet HLRs associated with maximum peakflow events were also very high, ranging from 6 to932 times higher than the maximum that commonlyoccurs in overland flow wastewater treatment.Summer HLRs were lower than those in spring, butonly SAY7 and PLEb supported mean peak HLRsbelow 0.1 m/d during this period.
Large HLRs in bioswales with longitudinalslopes less than 1.5% and/or flow constrictionscharacterize flow that is slow to moderately rapidand very deep. Large HLRs in bioswales withlongitudinal slopes greater than 1.5% and no flowconstrictions (e.g., check dams) indicate flow thatis rapid and is moderately deep. In both cases, theflow is generally too deep and rapid to permit muchdeposition of the clay and silt-sized suspendedsediments (�50 �m diameter). This is especiallyunfortunate given that these particles, due to theirstrong sorption capacity, are often associated withthe gravimetric majority of pollutants in urbanrunoff (Rexnord, 1984; Baker, 1992; Novotny andOlem, 1994).
Although most reported water-quality studies ofbioswales neglect to describe the hydraulic orhydrologic aspects of the swales they study, severaldo provide sufficient information to determine thattheir bioswales received flows of much less magni-tude than what appears to regularly occur inbioswales of King County. METRO (1992) re-ported good water-quality performance of abioswale, but its flow was always shallow (�37mm) and its HRT generally greater than 9 min.Flow in the bioswale monitored by Kercher et al.(1983) was so low that infiltration significantlyreduced the volume upstream of the outlet during
10 of 13 storm events. Had inlet flow rates beenmore like those in most King County bioswales,pollutant removal may have been much less forthese and other swales used inwater-quality studies.
4.3. Vegetation abundance and bioswaleperformance
Bioswale vegetation, regardless of abundance,did not appear to perform its intended functionadequately – namely, inducing the sedimentationof suspended sediments and their sorbed contami-nants via flow retardance. The rapid flow thatoccurred during even small rain storms did notappear to be strongly retarded in the swales withhigh aboveground biomass and longitudinal slopesgreater than 1.5% (SAY7 and DISC). Conversely,the poor vegetative cover in PLEa and PLEb waswell compensated by nonvegetal roughness (checkdams), shallow slope (�1%), and favorable swalegeometry, which contributed to long HRTs.
Overall, this study found no relation betweenbioswale vegetation abundance and maximumpeak flow HLR. Since HLR may be a plausiblesurrogate for treatment performance in bioswales,vegetative abundance is not useful or even relevant,at least under currently permitted hydrologic andhydraulic guidelines.
5. Recommendations
Establishing abundant herbaceous vegetation inbioswales is difficult, but achievable. Listed belowin descending order of importance are some guide-lines for achieving good vegetation cover:� Avoid shading by adjacent vegetation, struc-
tures, or side slopes� Minimize inundation during the dry season
drained soil (e.g. sandy loam); soil should belightly compacted with a static roller prior tofirst inundation
� Hydroseed during periods free from inundationand irrigate as necessary to facilitate plantestablishment
G. Mazer et al. / Ecological Engineering 17 (2001) 429–443442
� Minimize inter- and intra-seasonal hydrologicfluctuation by situating detention ponds up-stream from bioswales, but avoid ponds withmulti-day release periods
� Set biofilter longitudinal slope between 0.5% and2% and maintain a constant gradient throughoutthe length of the swale.Bioswales should only be located in areas that
will remain clear of invading vegetation. Intactforests should not be cleared for bioswale construc-tion due to the multitude of environmental benefitsthat forest provides. Swales should be mowed atleast once per growing season to foster high stemdensity, but soil compaction should be avoided andclippings should be removed. Other efforts must bemade as needed to prevent significant erosionand/or sediment deposition from occurring.
Although the above recommendations will im-prove the growth of herbaceous plants in bioswales,they probably will do little to improve bioswaletreatment performance given current hydrologicand hydraulic design guidelines. To achieve signifi-cantly improved performance, these guidelinesmust be altered to ensure that inflow rates aregreatly reduced. Erosion and flow convergencewould be minimized if longitudinal slopes were notpermitted to exceed 1.5%. Bioswale channel lengthand width should be sized such that resulting HLRsmore closely approach those found with overlandflow wastewater treatment facilities.
Vegetation abundance should not be consideredan accurate indicator of treatment performance.Herbaceous vegetation may serve to facilitate thecapture of sediments and their associated pollu-tants, but its influence appears minimal under thetypically high flow found in bioswales during stormevents.
Instead, we suggest that HLR be used as anestimate of sedimentation potential and treatmentperformance in bioswales and other storm-watertreatment facilities. HLR is determined by flow rateand treatment area, factors that should predicttreatment performance more accurately, and thatare widely used in other evaluations of water-treat-ment processes. Using HLR thresholds as a basisfor storm-water facility design would likely dis-courage the construction of bioswales in favor oflarge detention basins, which can more consistently
improve storm-water quality across a broad rangeof storm events but require significantly more landto construct.
Acknowledgements
We would like to thank staff from King CountySurface Water Management Division, particularlyDave Hancock and John Koon, for assistance withseveral facets of this study. Robin Cleveland’s workin developing the ISI was critical to evaluating theeffects of persistent high water on plant growth.King County provided funding through the Storm-water Technology Consortium of the Center forUrban Water Resources Management.
References
Baker, L.A., 1992. Introduction to nonpoint source pollution inthe United States and prospects for wetland use. EcologicalEngineering I, 1–26.
Booth, D.B., Jackson, C.R., 1997. Urbanization of aquaticsystems – degradation thresholds, storm-water detentionand limits of mitigation. Journal of American Water Re-sources Association 33 (5), 1077–1090.
Brady, N.C., Weil, R.R., 1996. The nature and properties ofsoils, eleventh ed. Prentic Hall, Upper Saddle River, NJ, 740pp.
Claytor, R., Schueler, T.R., 1996. Design of Storm-waterFiltering Systems. Chesapeake Bay Consortium.
Colwell, S.R., 2000. Characterization of performance predictorsand evaluation of mowing practices in biofiltration swales.M.S.C.E. Thesis. University of Washington. Seattle, WA.
Cooke, S.S., Azous, A., 1993. Effects of urban storm-waterrunoff and urbanization on palustrine wetland vegetation.A report to the US Environmental Protection Agency –Region 10. Seattle, WA.
Crawford, R.M.M., 1992. Oxygen availability as an ecologicallimit to plant distribution. Advances in Ecological Research23, 93–185.
Crawford, R.M.M., 1996. Whole plant adaptations to fluctuat-ing water tables. Folia Geobotany and Phytotaxonomy. 31,7–24.
Department of Ecology (Ecology), 1996. Selecting Best Manage-ment Practices for Storm-water Management. Water QualityProgram – Urban Nonpoint Unit. Olympia, WA.
Ewing, K., 1996. Tolerance of four wetland plant species toflooding and sediment deposition. Environmental and Ex-perimental Botany 36, 131–146.
Emers, M., 1990. Community Patterns along a water levelgradient in freshwater wetlands. M.S. Thesis, University ofWashington, Seattle, WA.
Ernst, W.H.O., 1990. Ecophysiology of plants in waterloggedand flooded environments. Aquatic Botany 38, 73–90.
G. Mazer et al. / Ecological Engineering 17 (2001) 429–443 443
Gabriell, T., 1997. Starting right with seeds: adequate light isthe key to successful germination. Organic gardening andfarming 26 (1), 32–36.
Harper, H.H., Yousef, A.Y., Wanielista, M.P., 1984. Effi-ciency of roadside swales in removing heavy metals fromhighway associated nonpoint source runoff. Options forReaching Water Quality Goals. American Water ResourceAssociation.
Horner, R.R., Skupien, J.J., Livingston, E.H., Shaver, H.E.,1994. Fundamentals of Urban Runoff Management: Tech-nical and Institutional Issues. Terrene Institute, Washing-ton, DC.
Kercher, Jr., W.C., Landon, J.C., Massarelli, R., 1983. Grassyswales prove cost-effective for water pollution control.Public Works, pp. 53–54.
King County, 1995. Evaluation of Water Quality Ponds andSwales in the Issaquah/East Lake Sammamish Basins.Final Report for Task 5 of Grant AgreementcTAX90096-Issaquah/East Lake Sammamish NonpointPlans. King County Surface Water Management Division,Seattle, WA.
King County, 1990. King County, Washington Surface WaterDesign Manual. King County Surface Water ManagementDivision. Seattle, WA.
King County, 1998. King County, Washington Surface WaterDesign Manual. King County Water and Land ResourcesDivision. Seattle, WA.
Kozlowski, T.T., 1984. Extent, causes and impacts of flooding.Flooding and plant growth. In: Kozlowski, T.T. (Ed.),Academic Press, New York, pp. 1–9.
METRO (Municipality of Metropolitan Seattle), 1992. Bi-ofiltration swale performance, recommendations, and de-sign considerations. Municipality of Metropolitan Seattle.Seattle, Washington.
Novotny, V., Olem, H., 1994. Water Quality: Prvention, iden-tification, and management of diffuse pollution. Van Nos-trand Reinhold, New York, NY, 1054 pp.
Oakland, P.H., 1983. An evaluation of urban storm-waterpollutant removal through grassed swale treatment. Inter-national Symposium on Urban Hydrology, Hydraulics andSediment Control. University of Kentucky, Lexington,KY, pp. 183–185.
Rexnord, 1984. Source and migration of highway runoff pollu-tants. vol. 3: Research Report prepared for the FHWA.FHWA/RD-84/059, NTIS PB86-227915.
Roesner, L.A., Urbonas, B., Sonnen, M., 1999. Design ofurban runoff quality controls. Proceedings of an Engineer-ing Foundation Conference on Current Practice and De-sign Criteria for Urban Quality Control. American Societyof Civil Engineers, New York, NY.
Schultz, D., 1998. Current status of vegetation management inroadside ditches and storm-water management facilities:implications for storm-water quality. University of Wash-ington, Seattle, WA.
Temple, D.M., 1991. Changes in vegetal flow resistance duringlong-duration flows. Transactions of the American Societyof Agricultural Engineers 4, 1769–1774.
USEPA, 1983. Results of the Nationwide Urban Runoff Pro-gram (NURP) – Final Report. USEPA NTIS PB84-185552. Water Planning Division, Washington, DC.
USEPA, 1997. Nonpoint Source Pollution: The Nation’sLargest Water Quality Problem. �http://www.epa.gov/OWOW/NPS/facts/point1.htm�
van der Valk, A.G., Swanson, S.D., Nuss, R.F., 1981. Theresponse of plant species to burial in three types ofAlaskan wetlands. Canadian Journal of Botany 61, 1150–1164.
Walsh, P.M., Barrett, M.E., Malina, Jr., J.F., 1997. Use ofvegetated controls for treatment of highway runoff. Re-search report 2954-2. Center for Transportation Research,Austin, TX.
Yousef, Y.A., Wanielista, M.P., Harper, H.H., 1985. Removalof highway contaminants by roadside swales. Transporta-tion Research Record 1017, 62–68.
