at SciVerse ScienceDirect

Acta Oecologica 38 (2012) 41e48

Contents lists available

Acta Oecologica

journal homepage: www.elsevier .com/locate/actoec

Original article

Seedling germination success and survival of the invasive shrub Scotchbroom (Cytisus scoparius) in response to fire and experimentalclipping in the montane grasslands of the Nilgiris, south India

Madhusudan P. Srinivasana,b,*, Ratul Kalitac, Inder K. Gurungc, Saroj K. Bhattacharjeec,Predit M. Antonyd, Suresh Krishnand, Scott K. Gleesona

aDepartment of Biology, University of Kentucky, 101 Morgan Building, Lexington, KY 40506, USAbKeystone Foundation, Groves Hill Road, Kotagiri, IndiacDepartment of Ecorestoration, Dimoria College, Khetri, Assam, IndiadWildlife Biology Department, Government Arts College, Ooty, India

a r t i c l e i n f o

Article history:Received 11 September 2010Accepted 6 September 2011Available online 23 September 2011

Keywords:Woody invasionShola-grasslandPrescribed burningSeedbank releaseFire fuelWestern Ghats

* Corresponding author. Department of Biology,Morgan Building, Lexington, KY 40506, USA. Tel.: þ1 81717.

E-mail addresses: madu@uky.edu, madhuxs@yaho

1146-609X/$ e see front matter � 2011 Elsevier Masdoi:10.1016/j.actao.2011.09.002

a b s t r a c t

The spread of the exotic shrub Scotch broom in the montane grasslands of the Nilgiris is one of the majorthreats to biodiversity and ecosystem function there. It is likely that fire suppression over the past fewdecades is the proximate cause of expansion of broom populations. This study capitalizes on a wildfireevent to examine fire effects on mature broom populations and soil seedbanks. Fire resulted in wide-spread death of mature broom stands but also stimulated broom soil seedbanks. However, this initialdifference in seedling densities in burned and unburned plots was lost over time due to continuousrecruitment in unburned plots. In a seed addition experiment, plots which were clipped prior to the fireshowed higher germination success, possibly because fire temperature was moderated by biomassremoval. In another experiment, non-dormant broom seeds were added to burned plots, which thenreceived clipping treatments; there were no differences in broom seedling survival in clipped vs.unclipped plots. Overall, these results suggest that prescribed burning might contribute to the control ofScotch broom invasion by helping eliminate mature stands without significantly increasing regenerationfrom seed.

� 2011 Elsevier Masson SAS. All rights reserved.

1. Introduction

Extensive research has focused on the causes, consequences andcontrol of exotic invasive species, and there is a substantialunderstanding of the complexities related to these topics (Drakeet al., 1989; Ehrenfeld, 2003; Hobbs and Huenneke, 1992; Hobbsand Humphries, 1995; Levine et al., 2003; Lodge, 1993; Maronand Vilà, 2001). However, this is still inadequate to effectivelymanage invasions; exotic species have the propensity to invadesites of varying conditions, and multiple species with widelydiffering traits can invade the same site, so the underlying princi-ples related to invasion mechanisms and control are not alwaysuniversally applicable. Invasion by both native and exotic woodyplants is a commonproblem in grasslands, and the pervasive effects

University of Kentucky, 10159 323 3284; fax: þ1 859 257

o.com (M.P. Srinivasan).

son SAS. All rights reserved.

of invasive woody plants on native communities and ecosystem iswell documented (Grice, 2004; Jackson et al., 2002; Wearne andMorgan, 2004).

The montane grasslands of the Nilgiris, south India, have shrunkdue to naturally expanding exotic tree plantations and shrubs,besides other unsustainable land use practices. The invasion of thegrasslands by the exotic shrub Cytisus scoparius (L.) Link (Scotchbroom, hereafter broom), though not as dramatic in extent andexpansion rate as Eucalyptus and Black wattle (Acacia mearnsii DeWild) plantations, is by no means a problem to ignore (see Fig. 1).Literature dating back to the early 1900’s (Meher-Homji, 1967;Ranganathan, 1938; Srinivasan et al., 2007; Zarri et al., 2006)cautioned about the spread of broom in the Nilgiris; recent studieshighlighted the healthy population structure of broom in the Nil-giris grassland and its adverse effects on native plant communitiesand ecosystem function (Srinivasan et al., 2007; Srinivasan,unpublished). Broom invasion in grasslands is also a seriousproblem in other parts of the world, including Australia, NewZealand, South Africa, Canada (Hosking et al., 1998; Smith et al.,2000) and the Pacific Northwest (Bossard, 2000).

Fig. 1. The study site. (a) Location of the Nilgiris in India; (b) pristine shola-grassland habitat; (c) Scotch broom invasion in Lakkadi grasslands - mature stands in the foreground andin the valley beyond; Black wattle plantations on the hilltops; (d) burned broom stands in the foreground, and unburned grasslands in the far ground.

M.P. Srinivasan et al. / Acta Oecologica 38 (2012) 41e4842

The common methods employed in the attempts to control anderadicate broom in these habitats include the use of herbicides,mechanical removal, managing seed vectors, biological control andprescribed burning (Andres et al., 1967; Bossard, 2000; Rice, 2004).Although burning broom stands appears to be an effectiveapproach, the potential unintended consequences associated withburning, particularly widespread seedbank stimulation, standregeneration, and unrecoverable fire damage to native vegetation,make its use highly debated (see DiTomaso et al., 2006; Mobley,1954). However, there have also been success stories (Agee, 1996;DiTomaso et al., 2006). There is more evidence of fire deterring,rather than promoting, woody invasions in grasslands compared toother ecosystems (Roques et al., 2001; van Auken, 2000). In theNilgiris, indigenous pastoralists, the Todas have been systematically

burning the grasslands for several centuries. However, reservemanagers have suppressed fire in the past few decades. Whetherthe reason for preventing fires stems from the fact that fire is notessential to maintain this ecosystem, or that it may attract unfa-vorable public opinion, remains a conundrum. Nevertheless,reports (Kumar, 1993; Srivastava, 2002) and personal interviews(conducted by MPS) consistently indicate that the woody exoticshave greatly expanded since fire suppression.

The primary goal of this study was to assess the effectiveness offire as a tool to deplete and possibly eliminate broom in themontane grasslands of the Nilgiris. This is a worthwhile exercisebecause other data from the same study site show that fire does notnegatively affect the native grassland community (Srinivasan,unpublished). First, we examined the efficacy of fire in killing

M.P. Srinivasan et al. / Acta Oecologica 38 (2012) 41e48 43

mature broom stands. We were specifically interested in deter-mining whether fire selectively killed younger stems as opposed tothicker stems. Since fire stimulates seedbanks of broom by breakingdormancy, we examined whether the seedling density in burnedbroom stands was greater than that in unburned stands. Further,we followed a subset of these broom stands over time to test ifseedling survival was influenced by burning. Since grazers, prom-inently sambar deer (Cervus unicolor), disperse broom seeds intoopen grasslands (Zarri et al., 2006), we assessed the effect of fire onseed germination and seedling survival in the open grasslands aswell. Seed germination in burned areas may be greater than inunburnt areas, and the seedlings may have higher survival over thelong-term due to reduced native competitors immediately after theburn. To examine the impact of the grasses on seedling recruitment,grass fuel load was reduced by clipping prior to the fire to measureeffects on germination, and grass competitors were clippedsubsequent to the fire to measure effects on seedling survival.

2. Methods

2.1. Study area

The Nilgiri Hills, an integral part of the Western Ghats moun-tains, is located between 11�100e10�300 N and 76�250e77�000 E. Thealtitude in the upper areas of the Nilgiris range from 1800 to2600 m asl. April is the warmest month with a mean maximumtemperature of 25 �C, and January is the coolest monthwith ameanmaximum temperature of 5 �C. Frost occurs at night on several daysfromNovember toMarch. The Nilgiris experiences twowet periods,the first receiving rain from the southwest monsoon between Juneand September, and a second from the northeast monsoon betweenOctober and December. The western regions of the plateau receiveabout 2500 mm of rainfall and exceeds 5000 mm in certain areas(see Caner et al., 2007 for detailed meteorological information). Thesoils are classified as non-allophanic andisols (Caner et al., 2000).The natural vegetation in the upper Nilgiris consists of patches ofstunted tropical evergreen forests (locally called shola) surroundedby grasslands. The grasslands are mainly composed of perennial C4tussock grasses. The origin of these mesic grasslands was longbelieved to be due to anthropogenic disturbances (see Thomas andPalmer, 2007), but recent evidence has shown that the grasslandswere present in the Nilgiris for at least 40,000 years BP (Caner et al.,2007), long before humanization of this landscape. The currentaccepted view is that the strong monsoon winds and frost duringwinter preclude expansion of the native tropical trees into the openslopes (see Caner et al., 2007; Thomas and Palmer, 2007).

2.2. Broom invasion in the study site

Scotch broom is a polycarpic perennial shrub, native to parts ofcentral and southwestern Europe (Rees and Paynter, 1997). Thisleguminous shrub grows to nearly 3 m in height in our study site.Broom plants exhibit plasticity in traits: in the Nilgiris, it is seen ina range of habitats from open high-altitude grasslands (>2300 m)to Eucalyptus plantations at slightly lower elevations, and the leavesare broader in plants growing under tree cover than in the opengrasslands. In the Nilgiris, broom plants usually start flowering inthe fourth year. Flowering peaks from March to May. A single plantcan produce several thousand seeds each year (Bossard, 1990).Broom seeds germinate and seedlings survive better in disturbedsites (Smith and Harlen, 1991) and seeds can remain viable in theseedbanks for decades (Bossard, 1993; Paynter et al., 2003). Broomseeds exhibit physical dormancy and germination can be achievedby scarification of the seed coat by fire (Rivas et al., 2006) or othermeans (Abdallah et al., 1989). Scotch broom is believed to have been

introduced into the Nilgiris by the British in the early 1900’s(Ranganathan, 1938). It was introduced as an ornamental for itsbright yellow blooms.

The current study was conducted in a site called Lakkadi in theUpper Bhavani reservoir area located in the southwestern corner ofthe Nilgiris plateau. A typical view of Lakkadi presents a mosaic ofvarious vegetation classes: wattle plantations on the hilltops anddistinct patches of broom amidst patches of remnant grasslands(Fig. 1a). The broom invasion tends to be thicker along the reser-voir’s shoreline and among the ruins of makeshift houses that arescattered throughout the region. This pattern suggests that local-ized sites of soil disturbance may have been the stronghold of theinvasion earlier on and it later spread to adjoining areas. Thesambar deer which heavily use broom patches are suspected to beimportant dispersers in the Nilgiris (Zarri et al., 2006). Despitea brief interruption in recruitment broom populations in the studysite are purportedly healthy (Srinivasan et al., 2007), further, therewas a decline in overall grassland species richness in broominvaded patches which were dominated by shade-tolerant weedynatives.

2.3. Field methods

A series of opportunistic experiments and surveys wereemployed to ascertain the impact of a wildfire event, whichoccurred at the end of March 2007, on mature broom stands(Fig. 1b), and on soil seedbanks in broom invaded and uninvadedgrasslands. A clipping experiment was also conducted to test howremoval of native competitors would affect broom survival.

2.3.1. Surveys of broom stemsA survey was carried out in October 2007 in burned broom

stands to assess the impact of the wildfire on mature broom. Thecenters of thirty-three 5 � 5 m plots were located by walkingdistances in the range of 100e200 footsteps in compass directionsselected at random. In each plot the basal area of all the stems weremeasured. Plants were examined closely and stems were broken todetermine if theywere alive. A total of 1629 stemswere censused in33 plots. The estimate may be slightly biased toward dead stems assome of the stems thought to have been killed by the fire wereprobably dead prior to the fire. However, it was usually possible toidentify and eliminate previously dead stems, as these were char-red completely, while the stems killed by the fire appeared cooked.

2.3.2. Seedlings surveys in broom patches and uninvadedgrasslands

A onetime seedling survey was conducted in October 2007 tocompare seedling densities in burned and unburned broom stands.The burned and unburned patches were in the same general areasand were separated by a road that acted as a fire barrier. Broomseedlings were counted in 82 and 88 (1 � 1 m) burned andunburned plots respectively; plots were located in broom stands asdescribed above. A seedling was defined as a broom plantmeasuring <10 cm in height and <5 mm in basal diameter, as thisseemed to be the general size of seedlings aged less than one year.

In a subset (21 in unburned broom and 20 in burned broom) ofthe plots used for the above survey, seedlings enumerated duringthe initial survey were marked with a plastic tag and these plotswere monitored for seedling germination and recruitment intosuccessive life stages over an additional 12-month period. After theinitial census, seedling densities were measured in February 2008in the dry season and finally in October 2008. New recruits weretagged during each sampling event.

A survey similar to the monitoring of seedlings in broominvaded sites was conducted in the open grasslands as well.

M.P. Srinivasan et al. / Acta Oecologica 38 (2012) 41e4844

Twenty-one plots were located in the unburned grasslands close to(5e10 m) unburned broom plots of the previous survey and,similarly, 20 plots were located in burned grasslands. All these plotswere monitored over a 12-month period similar to the survey ofseedlings in the broom stands.

2.3.3. Experiments in uninvaded grasslandsThe wildfire also burned through an existing experiment in the

open uninvaded grassland that involved broom seed addition toclipped plots. This allowed documentation of the effect of fire fuelconditions on seed viability. Fifty-four (1 � 1 m) subplots eachlocated in a 5 � 5 m plot were studied. These larger plots were laidin a randomized complete block design (18 plots in each of the 3blocks). The vegetation in half of these plots (9 per block) wasclipped to a height of 10 cm to simulate extremely intense grazingby sambar deer in this site (personal observation, MPS) and in othermountain grasslands in southern India (Sankaran, 2005), while therest were left unclipped. The clipped biomass was removed fromthe plots. Following the setup of the experimental plots thewildfireswept through all three blocks. Because each of the above subplotswere seeded with 50 broom seeds prior to the fire, it allowed forassessing the effects of fire fuel on seed viability. The seedingprocedure involvedmaking a tiny scratch on the soil surface (wherethe soil was hard and dry) so that the seed could be inserted a fewmillimeters into the soil to prevent it from being washed away.Since the amount of biomass was substantially greater in theunclipped plots than in the clipped plots just prior to the wildfire, itis reasonable to expect that on average the fire temperature mayhave been hotter in the unclipped plots. Thus, the clipped plots arereferred to as the low fire fuel treatment, while the plots that werenot clipped prior to the fire are referred to as the high fire fueltreatment. Germinated seedlings were first marked with plastictags in October 2007 and these subplots were sampled again inDecember 2007 to capture any subsequent germination.

To study the effect of clipping on seedling survival another set of36 (1 � 1 m) subplots were each seeded with 50 non-dormantbroom seeds two weeks after the fire, and a new clipping treat-ment was applied after the fire. The treatment levels includedclipping to a height of 10 cm or no clipping. Physical dormancy wasbroken by repeatedly immersing the seeds in boiling water andliquid nitrogen (Abdallah et al., 1989). Twenty-seven of the 36(1�1m) subplots used for this experimentwere located in the highfire fuel treatment plots of the earlier experiment, and theremaining nine subplots were located in nine new supplementalplots that were distributed among the three blocks. Since thesupplemental plots had not been clipped prior to the fire they weresimilar in conditions to the 27 high fire fuel plots. Half the plots ineach block were randomly assigned to be clipped, while the otherhalf were left unclipped. Plots were first clipped in October 2007;the treatment was repeated in February 2008. Germinated seed-lings were first marked with plastic tags in October 2007, theseedling subplots were monitored until late January 2009.

2.4. Data analysis

Descriptive statistics are reported for the number of broomstems that were dead and alive per plot, the basal area of dead andalive stems, and the density and total basal area of plots that hadlive stems. Box and whisker plots were used to illustrate thedistribution of basal areas of live and dead stems. A two-sample t-test for unequal sample variances was used to test for significantdifferences in mean seedling densities in burned and unburnedbroom stands. To understand broom recruitment over time inburned and unburned broom stands and grassland, repeatedmeasures ANOVAs were run separately for the number of seedlings

at each time point in broom stands and grasslands. Data weretransformed to meet assumptions of ANOVAs. Proc GLM (SAS,2008) with a repeated statement was used to analyze the data.The data from the fire fuel and the clipping experiment were bothanalyzed using repeated measures ANOVA as a randomizedcomplete block design (Proc Mixed, SAS, 2008).

3. Results

3.1. Fire effects on mature boom stands

On average, fire killed 98.7% of the broom stems surveyed in the33 plots; of all the sampled plots only four had live stems. This issignificant given that these stands were thriving before the fire. Themean number of dead and living stems per plot was 48.94 (�11.995% CI) and 0.61 (�0.7), respectively. Individual plants were killedregardless of their girth (basal area) as indicated by the overlap inthe range of basal areas between the two groups (mean basal areaof stems: dead ¼ 8 cm2, range ¼ 0.08e60.4 cm2, live ¼ 1.93 cm2,range ¼ 0.08e6.6 cm2; see Fig. 2). However, live stems wereconsistently found in plots that had low broom density(mean� 95% CI: 16� 9.63 vs. 54.07� 12.3) and low total basal area(76.06 � 70.31 vs. 440.24 � 72.06 cm2). This suggests that the girthof the stems is not a factor in preventing top kill from fire, andhigher girth does not ensure better survival.

3.2. Seedlings density in burned and unburnedbroom and grassland

The mean number of seedlings in the burned broom stand wasthreefold higher than in the unburned stands when surveyed once6 months after the wildfire (burned: 30 � 3.14, N ¼ 82; unburned:9.01 � 1.13, N ¼ 88; t ¼ 6.306, df ¼ 101.75, P < 0.0001; Fig. 3).

However, there were no overall differences in mean seedlingdensities between burned and unburned broom stands when theplots were followed over a period of 12 months. There wasa significant interaction between treatment and time (F2,78 ¼ 4.23,P ¼ 0.0181). The initial significant difference seen in October 2007(Fig. 4; t ¼ 2.34, df ¼ 74.1, P ¼ 0.0221) disappeared in February andOctober of the following year. The seedling densities increasedsignificantly (via germination) over time in the unburned broomplots (Oct ’07eFeb ’08: t ¼ �4.38, df ¼ 78, P < 0.0001; Oct ’07eDec’08: t ¼ �4.31, df ¼ 78, P < 0.0001) but not in the burned broomplots.

Similarly, in the grasslands the mean seedling densities werehigher in the burned plots compared to unburned plots 6 monthspost-fire in October 2007 (Fig. 5; t ¼ 1.82, df ¼ 76.3, P ¼ 0.073). Butthis difference was not maintained over time, as there wasa significant rise in seedling densities in the unburned grasslandsdue to further recruitment from the seedbank (Oct ’07eFeb ’08:t ¼ �2.99, df ¼ 78, P ¼ 0.0038; Oct ’07eDec ’08: t ¼ �3.00, df ¼ 78,P ¼ 0.0036).

3.3. Effect of fire fuel and clipping on broom seedlings recruitment

For the seeds planted prior to the fire, the overall broom seed-ling germination was significantly lower under high fire fuelconditions (Fig. 6; F1,102 ¼ 8.96, P¼ 0.0035) at both time points (Oct’07: t ¼ 1.89, df¼ 102, P¼ 0.06; Dec ’08: t¼ 2.35, df¼ 102, P¼ 0.02)and further recruitment in both treatments was negligible.

For the non-dormant seeds planted after the fire, the clippingtreatment failed to have an overall significant effect on the broomseedling germination or survival. But seedling densities decreasedsteadily (Fig. 7; F3,134 ¼ 9.40, P < 0.0001) under both clipped andunclipped treatments, starting from December ’08.

a b

Fig. 3. Onetime seedling density under burned (N ¼ 82) and unburned (N ¼ 88) broomstands. (a) Mean � se of seedlings shownwith the data swarm for each group. (b) Box-whisker plots showing the spread of data points among the two groups. To interpretthe boxes and whiskers, see caption of Fig. 2.

0

5

10

15

20

25

30

35

mrepsgnildeeS

2

Oct '07 Feb '08

burned broom

unburned broom

Oct '08

Fig. 4. Seedling survival and recruitment in burned and unburned broom plots overa period of 18 months post-fire; the wildfire occurred in March 2007. Treatment datapoints at each time point were offset to enhance clarity and avoid overlap.

Fig. 2. Box-whisker plots showing the data spread of the basal area of individual liveand dead stems. The box represents the data ranging from the first to the third quartile,and the horizontal line within the box marks the median. The lower whisker is theminimum data range while the upper whisker marks the inner fence (i.e. 1.5 � inter-quartile range). Extreme outliers lie outside the outer fence (i.e. 3 � inter-quartilerange).

M.P. Srinivasan et al. / Acta Oecologica 38 (2012) 41e48 45

4. Discussion

Scotch boom has been shown to regenerate from stumps inEurope (Tarrega et al., 1992). The overwhelming top killing ofbroom stands without girth discrimination and the absence of anyinstances of resprouting in the subsequent 21 months, suggeststhat fire is extremely effective in killing mature broom populationsin our study site. Only four of the 33 sampled plots had live stems;these plots did not have different topographies, which couldpotentially affect fire dynamics. Since live stems were consistentlyfound in plots that had low broom density and total basal area, thissuggests that the fire temperature in the low density plots maynot have reached the critical threshold required to kill broomstems. All else being equal, fire temperatures increase with theamount of available fuel; Bailey and Anderson (1980) found that thehottest fire temperatures were recorded where shrub densitieswere the highest.

Fire breaks physical dormancy in broom by scarifying the seedcoat (Bossard, 1993). Extensive stimulation of broom seedbank inburned broom relative to unburned stands seen in the currentstudy (Fig. 3, Results) is also well documented elsewhere (Agee,1996; Bossard, 2000). Tarrega et al. (1992) found that the germi-nation success ranged from 48% at 100 �C to 58% at 70 �C, whenbroom seeds were exposed to heat for 5 min; 43% of seedsgerminated when they were exposed to 130 �C for 1 min. Similarheat treatment experiments by Rivas et al. (2006) resulted inwidespread seed germination when broom seeds were exposed to80 �C, 10 min and 110 �C, 5 min. However, heating broom seeds insoil to a modest temperature of 150 �C for 1 min rendered them

0

1

2

3

4

5

6

7

8

9

10

mrepsgnildeeS

2

Oct '07 Dec '08 Apr '08 Jan'09

clipped grasslandunclipped grassland

Fig. 7. Effect of experimental clipping of grassland vegetation on the survival of broomseedlings. Treatment data points at each time point were offset to enhance clarity andavoid overlap.

0

1

2

3

4

5

6

7

8

9

mrepsgnildeeS

2

Oct '07 Feb '08

burned grasslandunburned grassland

Oct '08

Fig. 5. Seedling survival and recruitment in burned and unburned grasslands overa period of 18 months post-fire; the wildfire occurred in March 2007. Treatment datapoints at each time point were offset to enhance clarity and avoid overlap.

M.P. Srinivasan et al. / Acta Oecologica 38 (2012) 41e4846

nonviable (Bossard, 1993). Soil surface temperatures during fires inthe mature broom stands under dry conditions are known toexceed the above lethal temperature, and can potentially annihilatea large proportion of broom seeds on the soil surface and in theorganic horizon (Bossard, 1993); De Bano et al. (1977) showed thateven moderate chaparral fires produce temperatures as high as160 �C only 2.5 cm below the soil surface. However, the presence ofa large persistent subsurface seedbank (which can escape lethaltemperatures) ensures widespread regeneration. According toDowney (2002), broom seedbanks can grow to over 60,000 seedsper square meter. Hence, prescribed burning has often beenadvocated as a means to flush and deplete broom seedbanks whichotherwise remain viable for decades (Agee, 1996; Bossard, 2000).

Monitoring burned and unburned plots both in broom invadedand open grasslands revealed that there was little or no furtherrecruitment in burned plotswhile therewas a significant increase inrecruitment in both unburned broom and grassland plots. Perhapsthis recruitment occurred during the interim months from October2007 till the end of the second wet period in December 2007. Theseedbank was probably exhausted in the burned sites but seeds inthe unburned sites continued to continuously recruit from theseedbank. It is puzzling why the upward trend in seed recruitmentwas notmaintained until October 2008 (Fig. 4), because thefirstwetperiod for the calendar year is from June to August. One of thefactors could be that the southwestmonsoon in 2008was unusuallyweak; but it is likely that there could be several other reasons, such

0

0.5

1

1.5

2

2.5

3

3.5

mrepsgnildeeS

2

Oct '07 Dec '07

high fire fuellow fire fuel

Fig. 6. Effect of fire fuel conditions on broom germination success in the grassland,measured at the end of each wet season in the year the wildfire occurred. Treatmentdata points at each time point were offset to enhance clarity and avoid overlap.

as seedling predation. We observed a small green unidentifiedcaterpillar feeding on several seedlings in October 2008; this wasnot seen in the previous year. Downey and Smith (2000) have re-ported that the majority of the broom seedlings experience chronicbrowsing until they are 50 cm tall and so there is high mortality inthe first 3 years. Based on dung density data from the study site,Zarri et al. (2006) have reported that sambar deer preferentially usebroom invaded habitats more than any other natural or semi-natural habitat. According to these authors sambar deer prefer tobrowse onmature bushes rather than palatable native grasses. Also,the majority of sambar sightings were in broom invaded patches(personal observation,MPS). It is very likely that seedlings are beingculled as a result of sambar foraging in the Nilgiris but we do nothave data on how great these proportions are. In the UnitedKingdom, Paynter et al. (2000) found that rabbits foraged heavily onbroom seedlings and juveniles and significantly reduced survivaland fecundity; but such effects were not seenwith sheep grazing inAustralia (Sheppard et al., 2002). While we agree with Zarri et al.(2006) that there is widespread browsing of shrubs, the majorityof foraging in the broom habitats appears to be directed toward thesecondary herbaceous native communities encountered underbroom (personal observation, MPS).

The results of the reduction in fine-fuel (achieved by clippingthe vegetation down to a height of 10 cm) suggest that severebiomass removal can significantly increase germination bymoderating fire temperatures. The proportion of germinationsuccess in open grasslands in December 2008 was 0.05 and 0.02 forlow and high fire fuel treatments, respectively; this proportion wascalculated based on the 50 seeds that were added to each experi-mental unit. The fire in open grasslands would be similar to that inour high fuel fire treatment. According to Parker (2001), a fecundbroom shrub that produces 17,000 seeds can broadcast 50 seeds toa 1 m2 area roughly 4 m away; this estimate is based on her studypopulations in the glacial outwash prairie in Washington State.Since open grasslands are likely to have smaller seedbank sizes andlower germination success following burning, one could expectthat the stimulation of meager seed pools in the grassland byburning is unlikely to pose a significant invasion risk.

Grazing is an important grassland disturbance that has beenshown to directly and indirectly affect woody plant invasion ingrasslands. This is mediated through seed dispersal, reduction ofcompetition from the resident community, and by affecting firefrequency and intensity through alteration of fire fuel (Brown andArcher, 1989; Hobbs and Huenneke, 1992; van Auken, 2000). Theinfluence that grazers can have on broom regeneration in the Nil-giris grasslands is unknown. Grazing moderates fire temperature

M.P. Srinivasan et al. / Acta Oecologica 38 (2012) 41e48 47

(Hobbs et al., 1991; Zimmerman and Neuenschwander, 1984); thisis consistent with our result that broom can recruit better in burnedclipped sites. This suggests that burning in a heavily grazed grass-land with high background broom seed densities may conflict withthe objective of using fire to control broom.

We tested if clipping of native plants can promote broomseedling survival, because grazing can alter the competitive envi-ronment of the seedlings. Seedling survival was unaffected byclipping, suggesting that seedlings were not light limited. Thedecline in survival in both treatments over time could be attributedto other stress factors or predation. In a study conducted in a similargrassland community consisting of perennial tussock grasses inNew Zealand, the removal of tussocks reduced broom survival byexposing them to sheep grazing (Bellingham and Coomes, 2003).Thus broom invasion in grasslands may be influenced by the role ofgrazers as seed dispersers rather than grazers as regulators ofcompetition. Perhaps eliminating aboveground competition wouldbe a critical factor in rangelands that are dominated by turf andfodder grasses where their rhizomatous network forms a mat-likegrowth that can potentially suppress young shrub seedlingsthrough intense belowground competition. In an Australian grass-land, Sheppard et al. (2000, 2002) found that sheep grazingincreased broom seedling survival by possibly mitigating compe-tition from native grasses.

4.1. Fire management of broom: recommendations

The standard methods of weed control viz., mechanical, chem-ical and biological, have been sufficiently discussed for broom andrelated species (Swezy and Odion, 1997). The advantage anddisadvantage of each method varies with context. The efficacy ofeach control method, and its impacts on native ecosystem, need tobe evaluated in the Nilgiris using systematic long-term experi-ments. Herbicides such as 2, 4-D, and glyphosate can undoubtedlybe effective in killing broom and have successfully been used to killa variety of noxious invasives (Bossard, 2000), but these are alsonotorious for killing non-target native species. In the Nilgiris, Zarriet al. (2006) used these herbicides on broom stands and reportedthat the application killed several species of native plants. Thus, theuse of broadcast herbicides may not be an option for an eco-sensitive, biodiversity rich region such as the Nilgiris. Paintingindividual cut stems with glyphosate (LeBlanc, 2001) may perhapsbe safer; this can be an option when exercised with caution anddiscipline. Mechanical methods of control, mainly uprootingshrubs, are effective in managing adults. But both the use ofherbicides and mechanical removal fall short of depleting thepersistent seedbank.

In the Pacific Northwest and California grassland managers andresearchers have advocated using fire to control and eradicatebroom, often in conjunctionwith standard methods (DiTomaso andJohnson, 2006). For example, slashing the growth in the regener-ating stands and allowing it to dry in place before burning, orsupplementation with additional biomass where needed, is rec-ommended (Bossard, 2000). However, there is some hesitation toburn broom in France as fire causes considerable damage to theground flora (Paynter et al., 1998). This may not be a seriousconcern in the Nilgiris because, until a few decades ago, thesegrasslands were systematically burned for several centuries. Thesehistorical fires are said to have been conducted at intervals of 2e3years. Hence the native plant community may have undergonecontinuous selection for fire tolerance. A simultaneous study (Sri-nivasan, unpublished) that compared plant composition in burnedand unburned areas, both under broom and grassland, found thatthe recovered communities in the burned areas were almostindistinguishable compared to their corresponding unburned

treatments just 18 months post-fire. As seen in the Nilgiris, grass-land studies elsewhere have demonstrated the resilience of plantcommunities to burning (Jacobs and Schloeder, 2002; Sankaran,2005). Fire effects on grassland fauna are seldom generalizable.A review (Swengel, 2001) of fire effects on insect communities andpopulation in open habitats found that broadly, insect communitieswere quick to recover from fires, unless the events were frequent.Nonetheless, fire can have serious negative impacts on a smallsubset of the plant and faunal community. Using fire can bea suitable strategy when the negative effects are minimal andpersistent soil seedbanks are greatly reduced.

Repeated burning of broom stands has been advocated as aneffective control method (DiTomaso and Johnson, 2006). An initialfire will kill mature stands, stimulate the seedbanks and depletethem to a large extent; the follow-up burns will be effective ineliminating recruits. We recommend landscape scale futureexperiments to test a similar strategy in the Nilgiris to manage thebroom invasion problem here based on the documented success ofusing fire to manage broom in other sites, and evidence for largescale adult mortality and widespread seed stimulation. Accordingto Zarri et al. (2006) broom starts reproducing after four years inthe Nilgiris. Hence the follow-up burn must be done just before therecruits start reproducing and augmenting the seedbanks. Thistime interval will also allow for the growth and accumulation of thebiomass of weedy grasses that will provide the fine-fuel needed tosustain the spread of fire within the stands, as availability of amplefire fuel is often cited as a major obstacle for exercises involvingprescribed burning (Bossard, 2000). Despite the resilience of nativegrassland plants to fire, the burns must strictly target only broomstands and must be contained within them. Uncontrollable fireswill be disastrous to the rich grassland fauna of the region, besidesaccelerating the rate of soil erosion by several folds as seen else-where (Johansen et al., 2001). We also recommend that detailedstudies on the impact of fire on native biota and ecosystem processand services be undertaken within the framework of the proposedfire experiments before the implementation of repeated fire asa management practice in the Nilgiris to control and eradicatebroom.

Acknowledgments

We thank the International Foundation for Science (IFS), Swe-den for funds, and the Tamil Nadu Forest Department, Chennai,India for the research permits. Mohanraj N., WWF-India providedadditional logistical help. We thank Kausalya Shenoy, DNMcLetchie, Mary Arthur and three anonymous reviewers and thejournal’s associate editor F Badeck for inputs and comments.

References

Abdallah, M.M.F., Jones, R.A., El-Beltagy, A.S., 1989. An efficient method to overcomeseed dormancy in Scotch broom (Cytisus scoparius). Environmental andExperimental Botany 29, 499e501.

Agee, J.K., 1996. Achieving conservation biology objectives with fire in the PacificNorthwest. Weed Technology 10, 417e421.

Andres, L., Hawkes, R., Rizza, R., 1967. Apion seed weevil introduced for biologicalcontrol of Scotch broom. California Agriculture 10, 13e14.

Bailey, A.W., Anderson, M.L., 1980. Fire temperatures in grass, shrub and aspenforest communities of central Alberta. Journal of Range Management 33, 37e40.

Bellingham, P.J., Coomes, D.A., 2003. Grazing and community structure as deter-minants of invasion success by Scotch broom in a New Zealand montaneshrubland. Diversity and Distributions 9, 19e28.

Bossard, C.C., 1990. Tracing of ant-dispersed seeds: a new technique. Ecology 71,2370e2371.

Bossard, C.C., 1993. Seed germination in the exotic shrub Cytisus scoparius (Scotchbroom) in California. Madrono 40, 47e61.

Bossard, C.C., 2000. Cytisus scoparius. In: Bossard, C.C., Randall, J.M., Hoshovsky, M.C.(Eds.), Invasive Plants of California’s Wildlands. University of California Press,Berkeley, CA, USA, pp. 145e150.

M.P. Srinivasan et al. / Acta Oecologica 38 (2012) 41e4848

Brown, J.R., Archer, S., 1989. Woody plant invasion of grasslands: establishment ofhoney mesquite (Prosopis glandulosa var. glandulosa) on sites differing inherbaceous biomass and grazing history. Oecologia 80, 19e26.

Caner, L., Bourgeon, G., Toutain, F., Herbillon, A.J., 2000. Characteristics of non-allophanic andisols derived from low-activity clay regoliths in the Nilgiri Hills(Southern India). European Journal of Soil Science 51, 553e563.

Caner, L., Seen, D.L., Gunnell, Y., Ramesh, B.R., Bourgeon, G., 2007. Spatial hetero-geneity of land cover response to climatic change in the Nilgiri highlands(southern India) since the last glacial maximum. The Holocene 17, 195e205.

De Bano, L.F., Dunn, P.H., Conrad, C.E., 1977. Fire’s effect on physical and chemicalproperties of chaparral soils. In: Mooney, H.H., Conrad, C.E. (Eds.), Symposiumon Environmental Consequences of Fire and Fuel Management in Mediterra-nean Ecosystems. U.S.D.A. Forest Service, Washington, pp. 65e74.

DiTomaso, J.M., Brooks, M.L., Allen, E.B., Minnich, R., Rice, P.M., Kyser, G.B., 2006.Control of invasive weeds with prescribed burning. Weed Technology 20,535e548.

DiTomaso, J.M., Johnson, D.W., 2006. The Use of Fire as a Tool for ControllingInvasive Plants. Cal-IPC Publication 2006-01. California Invasive Plant Council,Berkeley, CA, USA, p. 56.

Downey, P.O., 2002. Disturbance and Invasion Ecology of Scotch Broom (Cytisusscoparius (L.) Link) in Australia. University of New England, Armidale, Australia.

Downey, P.O., Smith, J.M.B., 2000. Demography of the invasive shrub scotch broom(Cytisus scoparius) at Barrington Tops, New South Wales: insights for manage-ment. Austral Ecology 25, 477e485.

Drake, J.A., Mooney, H.A., Di Castri, F., Groves, R.H., Kruger, F.J., Rejmanek, M.,Williamson, M., 1989. Biological Invasions: A Global Perspective. John Wiley &Sons, Chichester, New York, Brisbane, Toronto, Singapore.

Ehrenfeld, J.G., 2003. Effects of exotic plant invasions on soil nutrient cyclingprocesses. Ecosystems 6, 503e523.

Grice, A., 2004. Weeds and the monitoring of biodiversity in Australian rangelands.Austral Ecology 29, 51e58.

Hobbs, N.T., Schimel, D.S., Owensby, C.E., Ojima, D.S., 1991. Fire and grazing in thetallgrass prairie: contingent effects on nitrogen budgets. Ecology 72, 1374e1382.

Hobbs, R.J., Huenneke, L.F., 1992. Disturbance, diversity, and invasion: implicationsfor conservation. Conservation Biology 6, 324e337.

Hobbs, R.J., Humphries, S.E., 1995. An integrated approach to the ecology andmanagement of plant invasions. Conservation Biology 9, 761e770.

Hosking, J.R., Smith, J.M.B., Sheppard, A.W., 1998. Cytisus scoparius (L.) link subsp.scoparius. In: Panetta, F.D., Groves, R.H., Shepherd, R.C.H. (Eds.), The Biology ofAustralian Weeds. R. G. & F. J. Richardson Publishers, Melbourne, Australia,pp. 77e88.

Jackson, R.B., Banner, J.L., Jobbágy, E.G., Pockman, W.T., Wall, D.H., 2002. Ecosystemcarbon loss with woody plant invasion of grasslands. Nature 418, 623e626.

Jacobs, M.J., Schloeder, C.A., 2002. Fire frequency and species associations inperennial grasslands of south-west Ethiopia. African Journal of Ecology 40, 1e9.

Johansen, M.P., Hakonson, T.E., Breshears, D.D., 2001. Post-fire runoff and erosionfrom rainfall simulation: contrasting forests with shrublands and grasslands.Hydrological Processes 15, 2953e2965.

Kumar, S., 1993. Survey and mapping of Shola forests and grasslands in the UpperNilgiri Plateau and assessment of human utilization of the vegetation. WorldWild Fund for Nature-India.

LeBlanc, J., 2001. Getting a Handle on Broom. University of California Agricultureand Natural Resources, Oakland, USA.

Levine, J.M., Vila, M., Antonio, C., Dukes, J.S., Grigulis, K., Lavorel, S., 2003. Mecha-nisms underlying the impacts of exotic plant invasions. Proceedings of the RoyalSociety of London, Series B: Biological Sciences 270, 775e781.

Lodge, D.M., 1993. Biological invasions: lessons for ecology. Trends in Ecology &Evolution 8, 133e137.

Maron, J.L., Vilà, M., 2001. When do herbivores affect plant invasion? Evidence forthe natural enemies and biotic resistance hypotheses. Oikos 95, 361e373.

Meher-Homji, V.M., 1967. Phytogeography of the South Indian hill stations. Bulletinof the Torrey Botanical Club 94, 230e242.

Mobley, L., 1954. Scotch broom, a menace to forest, range, and agricultural land. In:Proceedings of the 6th Annual California Weed Conference, pp. 39e42.

Parker, I.M., 2001. Safe site and seed limitation in Cytisus scoparius (Scotch broom):invasibility, disturbance, and the role of cryptogams in a glacial outwash prairie.Biological Invasions 3, 323e332.

Paynter, Q., Downey, P.O., Sheppard, A.W., 2003. Age structure and growth of thewoody legume weed Cytisus scoparius in native and exotic habitats: implica-tions for control. Journal of Applied Ecology 40, 470e480.

Paynter, Q., Fowler, S.V., Memmott, J., Shaw, R.H., Sheppard, A.W., 2000. Determi-nants of broom (Cytisus scoparius (L.) Link) abundance in Europe. PlantProtection Quarterly 15, 149e155.

Paynter, Q., Fowler, S.V., Memmott, J., Sheppard, A.W., 1998. Factors affecting theestablishment of Cytisus scoparius in southern France: implications formanaging both native and exotic populations. Journal of Applied Ecology 35,582e595.

Ranganathan, C.R., 1938. Studies on the ecology of the grassland vegetation of theNilgiri Plateau. Indian Forester 64, 523e541.

Rees, M., Paynter, Q., 1997. Biological control of Scotch broom: modelling thedeterminants of abundance and the potential impact of introduced insectherbivores. Journal of Applied Ecology 34, 1203e1221.

Rice, P.M., 2004. Fire as a Tool for Controlling Nonnative Invasive Plants. Center forInvasive Plant Management, Bozeman, MT, USA.

Rivas, M., Reyes, O., Casal, M., 2006. Influence of heat and smoke treatments on thegermination of six leguminous shrubby species. International Journal of Wild-land Fire 15, 73e80.

Roques, K.G., O’Connor, T.G., Watkinson, A.R., 2001. Dynamics of shrub encroach-ment in an African savanna: relative influences of fire, herbivory, rainfall anddensity dependence. Journal of Applied Ecology 38, 268e280.

Sankaran, M., 2005. Fire, grazing and the dynamics of tall-grass savannas in theKalakad-Mundanthurai Tiger Reserve, south India. Conservation and Society 3,4e25.

SAS, 2008. SAS for Windows Version 9.2. SAS Institute Inc, Cary, NC, USA.Sheppard, A.W., Hodge, P., Paynter, Q., 2000. Factors affecting broom regeneration in

Australia and their management implications. Plant Protection Quarterly 15,156e161.

Sheppard, A.W., Hodge, P., Paynter, Q., Rees, M., 2002. Factors affecting invasion andpersistence of broom Cytisus scoparius in Australia. Journal of Applied Ecology39, 721e734.

Smith, J.M.B., Harlen, R.L., 1991. Preliminary observations on the seed dynamics ofbroom (Cytisus scoparius) at Barrington Tops, New South Wales. Plant ProtectionQuarterly 6, 73e78.

Smith, J.M.B., Sheppard, A.W., Hosking, J.R., 2000. An introduction to the biogeog-raphy and ecology of broom (Cytisus scoparius) in Australia. Plant ProtectionQuarterly 15, 140e144.

Srinivasan, M.P., Shenoy, K., Gleeson, S.K., 2007. Population structure of Scotchbroom (Cytisus scoparius) and its invasion impacts on the resident plantcommunity in the grasslands of Nilgiris, India. Current Science 93, 1108e1113.

Srivastava, R.K., 2002. Forest fire and biotic interferences - a great threat to Nilgiribiosphere. In: Goldammer, J.G. (Ed.), International Forest Fire News. UnitedNations Economic Commission for Europe. Food and Agriculture Organizationof the United Nations, Germany, pp. 32e36.

Swengel, A.B., 2001. A literature review of insect responses to fire, compared toother conservation managements of open habitat. Biodiversity and Conserva-tion 10, 1141e1169.

Swezy, M., Odion, D.C., 1997. Fire on the mountain: a land manager’s manifesto forbroom control. In: Symposium Proceedings: California Exotic Pest Plant Council,pp. 1e7.

Tarrega, R., Calvo, L., Trabaud, L., 1992. Effect of high temperatures on seed germi-nation of two woody Leguminosae. Vegetatio 102, 139e147.

Thomas, S.M., Palmer, M.W., 2007. The montane grasslands of the Western Ghats,India: community ecology and conservation. Community Ecology 8, 67e73.

van Auken, O.W., 2000. Shrub invasions of North American semiarid grasslands.Annual Review of Ecology and Systematics 31, 197e215.

Wearne, L.J., Morgan, J.W., 2004. Community-level changes in Australian subalpinevegetation following invasion by the non-native shrub Cytisus scoparius. Journalof Vegetation Science 15, 595e604.

Zarri, A.A., Rahmani, A.R., Behan, M.J., 2006. Habitat modifications by scotch broomCytisus scoparius invasion of grasslands of the upper Nilgiris in India. Journal ofthe Bombay Natural History Society 103, 356e365.

Zimmerman, G.T., Neuenschwander, L.F., 1984. Livestock grazing influences oncommunity structure, fire intensity, and fire frequency within the Douglas-fir/ninebark habitat type. Journal of Range Management 37, 104e110.