The conservation management of Imperata cylindrica grassland in Nepal with fire and cutting: an experimental approach

Authors


  • *Denotes IUCN Red List category: CR = critically endangered, EN = endangered, V = vulnerable (IUCN 1996).

    *Denotes IUCN Red List category; EN = endangered (IUCN 1996).

Dr Nicholas Peet, Flagstones, 4 Stortford Road, Great Dunmow, Essex CM6 1DA, UK. Tel. 01371 873054

Abstract

1.Imperata cylindrica dominated grassland in the Nepal Terai supports a number of threatened species including swamp deer Cervus duvauceli, hispid hare Caprolagus hispidus and Bengal florican Eupodotis bengalensis. It is also the main source of thatch material for local communities. Current widespread cutting and burning of the grasslands is deleterious to cover-dependent vertebrates which would benefit from patches of grassland remaining unmanaged.

2. In Royal Bardia National Park, a randomized block experiment with four treatments (cutting, burning, cutting and burning, and no management) was established to examine the effects of these treatments on plant species abundance, species richness and grassland structure. The grassland was dominated by Imperata cylindrica, with Saccharum spontaneum, Vetiveria zizanioides, Desmostachya bipinnata and Schizachyrium brevifolium the other abundant species.

3. In year 3, 2 years after the first treatment, Imperata cylindrica remained the dominant grass species under all treatments. It increased in abundance in unmanaged plots and declined in abundance in managed plots. Desmostachya bipinnata showed the opposite response. Plant species richness was significantly higher in managed plots, which were structurally more heterogeneous.

4. The effects of cutting alone, burning alone, and cutting and burning combined on the structure and composition of the grassland were similar, despite additive effects on total grass cover, total forb cover and forb species richness in cut and burned plots. This allows the effects of cutting and burning to be reduced to a single treatment (cutting and burning) in future experiments to investigate the impacts of patch harvesting on community dynamics. Simpler factorial experiments can be designed than if cutting and burning had to be considered as separate treatments.

5. There was no difference in the establishment of tall grass species or woody species between managed and unmanaged plots.

6. Results suggest that patches of grassland could be left unmanaged on a 2-year rotation without significantly altering the composition of the plant community, thereby providing refugia for cover-dependent faunal species.

Introduction

Protected areas in the lowland Terai of Nepal contain some of the few remaining examples of tall grassland/forest mosaics and their fauna, that were previously widely distributed across northern India and southern Nepal in the valleys of the Ganges and Brahmaputra Rivers and their tributaries (Bell & Oliver 1992). These important grasslands support a wide range of biodiversity (Dinerstein 1979a,b; Mishra 1982a; Bell & Oliver 1992; IUCN 1993) and are a subsistence resource for local people (Mishra 1982b; Lehmkuhl, Upreti & Sharma 1988; Vaa Saetre 1993; Brown 1997). In the two most westerly protected areas of the Terai, Royal Bardia National Park (hereafter Bardia) and Royal Suklaphanta Wildlife Reserve (hereafter Suklaphanta), grasslands dominated by the perennial grass Imperata cylindrica are particularly widespread (Schaaf 1978; Pokharel 1993; Peet et al. 1999). In some cases these open grasslands, or phantas, result from previous anthropogenic disturbance from cultivation and grazing of stock (Pokharel 1993), but in others the origin of the grasslands is unclear.

In Bardia, the I. cylindrica dominated grassland/ riverine forest mosaic supports a high ungulate biomass and in particular a dense population of chital Axis axis, with ≈ 200 animals km–2 in the core of the park, on the Karnali River floodplain; this is the richest area of the park for biodiversity (Naess & Andersen 1993). There is also a small population of swamp deer Cervus duvauceli (V)* which utilizes the I. cylindrica dominated grassland (Ghimire 1996). The high ungulate biomass helps to maintain a high density of tigers Panthera tigris (EN)*, with 4·4 adult residents 100 km–2 (Stoen 1994). Endangered small mammals such as hispid hare Caprolagus hispidus (EN)* and pygmy hog Sus salvanius (CR)*, may also remain extant in the grasslands (Oliver 1985), whilst threatened grassland birds include Bengal florican Eupodotis bengalensis (EN)*, which nests in I. cylindrica dominated grassland (Inskipp & Inskipp 1983).

In Suklaphanta, I. cylindrica dominated grassland supports over a third of the world's remaining swamp deer population, with over 1000 animals (Schaaf 1978; Henshaw 1994), together with Nepal's largest populations of Bengal florican (Inskipp & Inskipp 1983) and hispid hare (Bell 1986). Inventories for other faunal groups, particularly herpetofauna, small mammals and invertebrates, are largely absent for both protected areas. It seems likely that there remains extant a group of faunal species dependent on the threatened grasslands, but whose status is currently poorly known.

The I. cylindrica grassland is not only important for biodiversity, but also for local people who rely on an annual harvest of grass to provide thatching material for their houses (Mishra 1982a; Lehmkuhl, Upreti & Sharma 1988; Brown 1997). Protected area authorities allow people access to the grasslands for a 10-day period annually, in order to harvest thatch and cane material. The grasslands are cut by hand and then burned off both illegally by local people, and for management purposes by protected area staff. Allowing local communities to enter protected areas to harvest grasses has been described as a practical trade off between conservationists and local people (Mishra 1982b, 1984).

Cutting and fire have been promoted as a method of grassland management for a number of reasons: (i) to prevent succession from grassland to forest; (ii) to prevent succession from shorter I. cylindrica dominated swards to taller grassland communities, dominated by Erianthus, Narenga and Saccharum species, which are less favoured by ungulates; (iii) to provide ungulates with high quality forage as the grassland regenerates; and (iv) to ensure a thatch crop for the following year.

With the exception of ungulate utilization of regenerating swards, these claims are largely derived from anecdotal information (Bell & Oliver 1992) with the result that management has been based on hearsay and a tradition of cutting and burning. In contrast, there is strong evidence that widespread cutting and burning has been deleterious to less mobile species and species less tolerant of disturbance, including pygmy hog (Oliver 1980) and hispid hare (Bell, Oliver & Ghose 1990). Following biomass removal from cutting and fire, both species are confined to small refugia of unburned grassland, or are forced into suboptimal habitat, where they are vulnerable to predation and disturbance (Oliver 1980, 1981; Bell, Oliver & Ghose 1990). There is ample evidence from elsewhere in the world that fire can affect the relative and absolute abundance of small mammals (Cheesman & Delany 1979; Fa & Sanchez-Cordero 1993; Friend 1993), herpetofauna (Fyfe 1980; Barbault 1983; Gillon 1983; Braithwaite 1987) and invertebrates (Gillon 1970, 1983; Ahlgren 1974; Majer 1984; Andersen 1991). With little information available on many faunal groups in the Nepalese grasslands, it seems likely that there will be other species, in addition to pygmy hog and hispid hare, that are deleteriously affected by widespread annual cutting and burning of the grasslands.

Where faunal species composition and abundance is influenced by vegetation at different stages of postfire regeneration, patch burning has been recommended as a conservation measure in fire-prone systems (Fyfe 1980; Braithwaite 1987; Fa & Sanchez-Cordero 1993; Johnson 1997). A similar approach will probably benefit conservation in the Nepalese grasslands but the success of such management requires a clearer understanding of the implications of cutting and burning, and of leaving grassland unmanaged, for the grassland plant community. This study was designed to investigate the separate effects of annual cutting and annual burning on grassland structure and community organization, and to examine those effects when cutting and burning were combined under a single management prescription, or when grassland remained unmanaged. The aim of investigating these effects was to establish whether patches of grassland could be left unmanaged on a rotational basis, in order to provide habitat for cover-dependent species such as the hispid hare.

Study site

Royal Bardia National Park (28°15′–28°40′N, 81°15′–81°40′E) occupies 968 km2 in the far west of the subtropical Terai region of Nepal (Fig. 1). Altitude in the protected area ranges from 150 to 1441 m above sea level. The climate is monsoonal, with most precipitation falling between June and September. Mean annual rainfall to the south of the park has been recorded at 1560 mm, whilst on the crest of the Churia hills mean annual rainfall is 2230 mm. A cool, dry season lasts from November to mid-February, followed by a hot, dry season until May, daytime temperatures reaching a peak in June at 45 °C and falling as low as 10 °C in January (Bolton 1976).

Figure 1.

Map of Royal Bardia National Park.

Approximately 70% of the park area is dominated by sal Shorea robusta forest (IUCN 1993), the remainder consisting of riverine forest, mixed hardwood forest and grassland. A series of phantas are located in the south-west of the protected area bordering the floodplain of the Karnali river (Fig. 1). These grasslands range in size from < 1 to 120 ha. The majority of the phanta grassland has been classified as falling into an Imperata cylindrica species assemblage, characterized by the dominance of I. cylindrica and subdominance of Saccharum spontaneum, Vetiveria zizanioides, Desmostachya bipinnata and Schizachyrium brevifolium (Peet et al. 1999). Thirty-two species of mammal occur in the park including tiger, swamp deer, Asian elephant Elaphas maximus (EN)* and greater one-horned rhinoceros Rhinoceros unicornis (EN)*; 297 species of bird have been recorded (BCN 1997).

The experimental site was located in Baghoura Phanta, the second largest I. cylindrica dominated grassland, covering ≈ 80 ha. The phanta was bordered on two sides by sal forest, on one by riverine forest and by a branch of the Geruwa River on the fourth. The phanta is regularly utilized by chital and swamp deer (Dinerstein 1979b; Ghimire 1996; Moe & Wegge 1997), especially after cutting and burning. Asian elephant, greater one-horned rhinoceros, wild boar Sus scrofa, hog deer Axis porcinus, barking deer Muntiacus muntjak and nilgai Boselaphus tragocamelus are less frequently recorded (Dinerstein 1979b; Peet 1997). Bengal florican breeds in the phanta.

The phanta is thought to have derived from an area under agricultural land use, prior to the establishment of a wildlife reserve in 1976. According to local people it was settled and cultivated in 1965 by migrants from the hills. However, with the imminent establishment of the Royal Karnali Wildlife Reserve, grazing and cultivation in Baghoura had ceased by 1975.

The history of grass cutting is obscure but it seems likely that indigenous Tharu people have harvested thatch grass for centuries. McMillan (1972) reported grass cutting as being a regular activity in the Bardiya District. Records from the Department of National Parks and Wildlife Conservation detail legal grass cutting inside the protected area from 1983 onwards. Fire is also likely to have had a long history in Bardia's grasslands as it is intimately associated with grass harvesting and the grazing of domestic stock. Hamilton (1819) noted that the grasslands of the Terai were burned once a year to ‘keep the country clear and improve pasture’, and Hooker (1893) recorded annual burning of ‘the gigantic Gramineae’ in the Terai of north-eastern India. Bolton (1976) observed uncontrolled burning in the then Royal Karnali Wildlife Reserve during the dry season.

Methods

A randomized block experiment was set up in Baghoura phanta consisting of four blocks, each containing four treatments: cutting, burning, cutting and burning, and no management. Each plot measured 35 × 35 m and was surrounded by a 3-m wide fire line. The treatment was designed to follow local practice. The plots were therefore cut immediately prior to local people entering the park to harvest grass. Cutting involved removing all above-ground biomass in a plot, with the exception of a stubble of 1–10 cm, and was carried out by local people using traditional sickles. The annual harvest began on 2 January 1995, 1996, 1997. The plots were burned immediately following the end of the 10-day harvesting period. Burning was carried out in the direction of the prevailing wind, in the same manner as park rangers would burn the grassland. The grasses were dry and senescent at the time of burning. Flame heights in the uncut plots reached 2–2·5 m but in cut plots fires were creeping with flame heights reaching no more than 15–20 cm. The grassland outside the experimental area was burned between 21 January 1995 and 5 February 1996.

Vegetation monitoring was carried out: (i) pre-treatment between 10 and 24 December 1994; (ii) 11 months after the first treatment, between 28 November and 12 December 1995; and (iii) 11 months after the second treatment, between 25 November and 10 December 1996. Monitoring involved sampling inside a 25 × 25 m core area in each plot. Twenty-five, randomly distributed, 1-m2 quadrats were surveyed in the core area of each plot. In each quadrat all plant species were identified or collected for later identification at the National Herbarium, Kathmandu. A number of species, particularly in young growth stages, could not be identified and were recorded as unknown species a, b, c, etc. The percentage cover of each species was estimated visually, as was the percentage cover of bare ground.

Any shrubs, trees or tall grass species occurring in a plot, but not falling within sampling quadrats, were recorded. Visual estimates of the percentage of live above-ground biomass, as a proportion of the total above-ground biomass, were made in six randomly located 30 × 30-cm quadrats in the core area of each plot.

Analysis

Standard techniques of Detrended Correspondence Analysis (DCA) (Hill 1994) and Analysis of Variance (anova) were employed. Data for anova were checked for heteroscedasticity and percentage data were arcsine square root transformed. In the following analyses year 1 refers to pre-treatment, with plots grouped as those ‘to be cut, burned, cut and burned, not to be managed’. Year 3 was after two treatment events.

Nomenclature

Nomenclature follows Bor (1960) for the grasses and Hara, Chater & Williams (1978–82) for the other species. Species lists follow taxonomic order.

Results

In the experimental area 62 plant species were identified, of which 14 species were Poaceae, five were Cyperaceae, 33 were species of forb, nine were species of tree and shrub, and one was a moss species.

An ordination of the plots, resulting from Detrended Correspondence Analysis, over the 3-year experimental period is presented in Fig. 2. Axis 1 accounted for 26·6% of the variance in the data and axis 2 a further 7·5% of the variance. A degree of spatial variability across the site was indicated by the spread of the plots in the first year, although differences in species composition were minor. Most plots fall between 0 and 1 average standard deviations of species turnover (Gauch 1982) along axis 1. Complete turnover of species composition between two samples occurs in about 4 standard deviations (Kent & Coker 1992). Generally, managed plots moved through ordination space in an opposite direction to unmanaged plots over time. The ordination suggests that there were differences in species composition and species abundance between managed and unmanaged plots. Plots under the three managed treatments however, did not move towards common endpoints in ordination space. Cutting, burning, and cutting and burning did not necessarily lead to plots with the same species abundances and species compositions, although the range of variation between treatments was small.

Figure 2.

Plots of axis 1 and axis 2 sample scores resulting from DCA for experimental plots under four treatment regimes: (a) no management, (b) burning, (c) cutting, and (d) cutting and burning, in an Imperata cylindrica dominated grassland. Within each graph, individual plots are identified by numbers (1–4) and the arrow indicates the movement of each plot from year 1, before treatment, to year 3, after two treatment events. Axis 1 accounted for 26·6% of the variation in the data, axis 2 a further 7·5%.

The abundance of functional groups

Grasses dominated the vegetative cover on the experimental site (Fig. 3). of the 14 grass species recorded, five contributed over 98% of the total cover of grasses in any one year: Imperata cylindrica, Saccharum spontaneum, Vetiveria zizanioides, Desmostachya bipinnata and Schizachyrium brevifolium. The total cover of grasses increased in unmanaged plots in years 2 and 3 and was higher in unmanaged plots than managed plots in year 3, with significant cut and burn effects in year 3 (Table 1, Fig. 3). Although the effects of cutting alone and burning alone were additive in cut and burned plots, the total cover of grasses remained above 70% of the total vegetative cover.

Figure 3.

Mean percentage cover (+ SE) of (a) grasses, (b) forbs, and (c) trees and shrubs, in experimental plots under four management treatments (cutting, burning, cutting and burning, and no management) in an Imperata cylindrica dominated grassland.

Table 1.  Summary of the analysis of variance for the effects of cutting, burning and block on the percentage cover of grasses, forbs and trees and shrubs in experimental plots under four treatments (cutting, burning, cutting and burning, and no management) in an Imperata cylindrica dominated grassland. The significance of the F ratios are indicated as follows: *, P < 0·05, **, P < 0·01, ***, P < 0·001
F ratios
GrassesForbsTrees and shrubs
Sourced.f.Year 1Year 3Year 1Year 3Year 1Year 3
Cut10·3338·03***0·12 8·79*0·18 3·67
Burn10·0667·23***6·21*11·03**2·7310·55*
Block31·66 3·721·36 0·520·55 4·18*
Cut×burn11·91 3·370·33 0·022·57 1·10
Error9      

The total cover of forbs was significantly higher in the three managed treatments than in unmanaged plots in year 3 (Table 1, Fig. 3), particularly in the cut and burned plots, although a significant effect was already present pre-treatment in plots grouped as those ‘to be burned’. Although the effects of cutting alone and burning alone were additive in cut and burned plots the total cover of forbs remained very low at less than 6% of the total vegetative cover.

Cover values for trees and shrubs were very low, so data should be treated with caution. The total cover of trees and shrubs increased in year 3 and was significantly higher in burned, and cut and burned plots, than in unmanaged plots (Table 1, Fig. 3).

Species abundance

Consideration was only given to the abundance of the main structural grasses, as these species contributed overwhelmingly to the total vegetation cover. Significant responses to management were exhibited by Imperata cylindrica and Desmostachya bipinnata (Table 2, Fig. 4). Although there were differences in the abundance of individual species between managed and unmanaged plots, there were no clear differences between the cutting, burning, or cutting and burning treatments.

Table 2.  Summary of the analysis of variance for the effects of cutting, burning and block on the abundance of five grass species (Imperata cylindrica, Desmostachya bipinnata, Saccharum spontaneum, Schizachyrium brevifolium and Vetiveria zizanioides) in experimental plots under four treatment regimes (cutting, burning, cutting and burning, and no management) in an Imperata cylindrica dominated grassland. The significance of the F ratios are indicated as follows: *, P < 0·05, **, P < 0·01, ***, P < 0·001
F ratios
I. cylindricaD. bipinnataS. spontaneumS. brevifoliumV. zizanioides
Sourced.f.Year 1Year 3Year 1Year 3Year 1Year 3Year 1Year 3Year 1Year 3
Cut13·34 6·33*0·1310·56*14·27**0·300·160·513·650·20
Burn10·2221·08**0·2915·52** 0·324·081·300·812·542·20
Block33·83 0·390·87 0·8149·57***0·901·480·664·32*5·42*
Cut×burn10·02 6·50*0·24 6·62* 1·862·731·480·200·052·58
Error9          
Figure 4.

Mean percentage cover (+ SE) of five grass species: (a) Imperata cylindrica, (b) Desmostachya bipinnata, (c) Saccharum spontaneum, (d) Schizachyrium brevifolium, and (e) Vetiveria zizanioides, in experimental plots under four treatment regimes (cutting, burning, cutting and burning, and no management) over 3 years, in an Imperata cylindrica grassland.

I. cylindrica increased in abundance in unmanaged plots in years 2 and 3, but in all managed plots I. cylindrica declined in abundance following second treatment and was more abundant in unmanaged than managed plots in year 3, with significant cut and burn effects, although the effects of cutting alone and burning alone were not additive in cut and burned plots (Table 2, Fig. 4). The response of D. bipinnata was exactly the opposite (Table 2, Fig. 4).

Saccharum spontaneum, Vetiveria zizanioides and Schizachyrium brevifolium showed no significant response to management in the third year (Table 2). However, a pre-treatment cut effect on S. spontaneum was lost by the third year, with an associated decline in cover in unmanaged plots after second treatment, suggesting a non-significant trend for declining abundance in unmanaged plots (Table 2, Fig. 4).

Species richness

Plant species richness increased across the whole site in year 3, but was higher in plots under the three managed treatments with significant cut and burn effects in year 3 (Table 3, Fig. 5). Species richness was highest in cut and burned plots, with the effects of cutting alone and burning alone being additive in cut and burned plots.

Table 3.  Summary of the analysis of variance for the effects of cutting, burning and block on total species richness, forb species richness, grass species richness, and tree and shrub species richness in experimental plots under four treatment regimes (cutting, burning, cutting and burning, and no management) in an Imperata cylindrica dominated grassland. The significance of the F ratios are indicated as follows: *, P < 0·05, **, P < 0·01
F ratios
TotalForbsGrassesTrees and shrubs
Sourced.f.Year 1Year 3Year 1Year 3Year 1Year 3Year 1Year 3
Cut12·65 7·21*0·1510·09*7·51*0·150·690·16
Burn13·6117·33**0·0219·42**0·090·202·771·42
Block31·77 0·211·76 0·101·080·152·082·26
Cut×burn10·17 0·300·02 1·274·550·022·770·16
Error9        
Figure 5.

Changes in (a) mean species richness (+ SE), and (b) mean forb species richness (+ SE), in experimental plots under four treatment regimes (cutting, burning, cutting and burning, and no management) in an Imperata cylindrica dominated grassland.

Most of the changes in total species richness were due to changes in forb species richness which mirrored those of total species richness (Table 3, Fig. 5). Forb species richness in the different treatments was ranked as follows: cut and burned > burned > cut > unmanaged. The effects of cutting alone and burning alone were additive in cut and burned plots. There were no treatment effects on grass species richness or tree and shrub species richness (Table 3).

Rank–abundance curves

The patterns of relative abundance in year 3 are summarized using rank abundance curves (Fig. 6). The unmanaged plots were strongly dominated by a single species, I. cylindrica, whereas dominance was less extreme and a number of species were structurally important in managed plots. Species contributing to increased richness in managed over unmanaged plots were of low abundance and structurally insignificant.

Figure 6.

Rank–abundance curves for experimental plots under four treatment regimes (Cb, cutting; cB, burning; CB, cutting and burning; and cb, no management) after 3 years (two treatment events), in an Imperata cylindrica dominated grassland.

Correlations between species rank in each treatment type, for the 23 species common to each treatment, were highly significant (Table 4), suggesting that there was little change in the position of individual species in the dominance spectra between treatments. Although there were changes in the abundance of species, these were not great enough to cause large shifts in the community and new species occurring under a single treatment type were not structurally important.

Table 4.  Spearman rank correlation coefficients between the rank order of species, based on mean percentage cover, occurring in experimental plots under four treatment regimes (cB, burning; Cb, cutting; CB, cutting and burning; and cb, no management) in year 3 after two treatment events, in an Imperata cylindrica dominated grassland. The significance of the Rs coefficients are indicated as follows: ***, P < 0·001
TreatmentcbcBCbCB
cb
cB0·76***
Cb0·87***0·74***
CB0·90***0·81***0·93***

Physical structure

There were structural differences between the grassland in the three managed treatments and grassland in the unmanaged plots. Unmanaged plots were structurally more homogeneous than managed plots. The percentage cover of bare ground declined under all treatments in year 2, but particularly in unmanaged plots. The percentage cover of bare ground in managed plots then increased in year 3 after the second treatment, but remained low in unmanaged plots, with significant cut and burn effects in year 3; the effects of cutting alone and burning alone, however, were not additive in the cut and burned plots (Table 5, Fig. 7). In year 3, unmanaged plots had a significantly lower proportion of live above-ground biomass than managed plots (Table 5, Fig. 7). The effects of cutting alone and burning alone on the proportion of live biomass, however, were not additive in cut and burned plots.

Table 5.  Summary of the analysis of variance for the effects of cutting, burning and block on the percentage cover of bare ground and the proportion of live biomass in experimental plots under four treatment regimes (cutting, burning, cutting and burning, and no management) in an Imperata cylindrica dominated grassland. The significance of the F ratios is indicated as follows: *, P < 0·05, **, P < 0·01, ***, P < 0·001
F ratio
Percentage cover of
bare ground
Sourced.f.Year 1Year 3Proportion of
live biomass
Year 3
Cut10·3232·95***52·93***
Burn10·1653·31***71·65***
Block31·58 0·40 0·64
Cut×burn12·5211·87**17·92**
Error9   
Figure 7.

Plots of (a) mean percentage cover (+ SE) of bare ground over 3 years, and (b) mean percentage of live biomass (+ SE) in year 3, in experimental plots under four treatment regimes (cutting, burning, cutting and burning, and no management) in an Imperata cylindrica dominated grassland.

Discussion

Imperata cylindrica dominated grasslands of the Nepalese Terai share many features with savanna systems in terms of climate, seasonality and structure. Savannas are typically dominated by grasses, Poaceae, of which a few species will be highly dominant (Sarmiento 1983; Medina & Huber 1992). Five grass species contributed overwhelmingly to the vegetation cover in the phanta grassland in Bardia: Imperata cylindrica, Saccharum spontaneum, Vetiveria zizanioides, Desmostachya bipinnata and Schizachyrium brevifolium. Of these species, all but S. brevifolium were of medium height (0·5–1·9 m) and structurally important in the grassland. In contrast, S. brevifolium was a low-growing (< 10 cm), mat-forming species. The dominance of these species was maintained under all treatment regimes and although there was an increase in the number of forb species under conditions of cutting and/or burning, they remained insignificant in terms of cover. These low-frequency, low-abundance species are another feature of savanna type grasslands (Medina & Huber 1992) and of Nepalese tall grasslands (Lehmkuhl 1989; Peet et al. 1999).

Response of imperata cylindrica and desmostachya bipinnata to management

Differences between the managed and unmanaged treatments resulted primarily from an increase in the species richness of forbs in the managed treatments, together with a decrease in the abundance of D. bipinnata and an increase in the abundance of I. cylindrica in the unmanaged treatment.

I. cylindrica rapidly colonizes disturbed sites (Brook 1989). A high root-rhizome : shoot ratio provides a source of dry matter for regeneration after cutting or burning (Ramakrishnan et al. 1983). I. cylindrica dominated grasslands are often described as a fire climax community (Seth 1970; Falvey 1981; Skerman & Riveros 1989) and in northern India they are considered to dominate following fire, cutting and grazing, being the first stage of degradation of a Phragmites–Saccharum–Imperata community (Dabadghao & Shankarnarayan 1973).

As I. cylindrica is reported to dominate regularly disturbed sites, the increase in its relative abundance in unmanaged plots was unexpected. It will, however, dominate on sites previously disturbed by slash-and-burn agriculture for up to 7 years without further disturbance events (Kushawa, Ramakrishnan & Tripathi 1983). The unmanaged plots were only undisturbed for 3 years, following previous regular disturbance, and leaving them undisturbed for a longer time period might result in a decline in the abundance of I. cylindrica, particularly as there was a build up of dense litter in these plots.

I. cylindrica is shade-intolerant (Brook 1989), shading reducing both shoot and rhizome biomass (Soerjani 1970). It is therefore possible that a build up of litter and increased shading on the unmanaged plots will result in a decline in the abundance of I. cylindrica and an increase in the abundance of shade-tolerant grasses, or perhaps oscillations in the abundance of I. cylindrica (Tilman & Wedin 1991), especially on the highly productive Terai soils (Lehmkuhl 1989). If more shade-tolerant species establish, then in Bardia it is likely that succession will result in taller grass swards dominated by Erianthus ravennae, Narenga porphyrocoma and Saccharum bengalense in unmanaged areas. If, however, an I. cylindrica dominated grassland were left unmanaged over a 2–3-year period and was then cut and/or burned, amounts of litter and shading would again be reduced and it is likely that I. cylindrica would continue to dominate.

Dabadghao & Shankarnarayan (1973) suggest that D. bipinnata may become dominant after disturbance of a Phragmites–Saccharum–Imperata grassland type in northern India, as a result of regular cutting, burning or grazing. Clearly it is a poor competitor in unmanaged swards where it rapidly declines in abundance, possibly due to shading. Although D. bipinnata increased in abundance in managed plots there was little evidence from annually cut and burned grassland outside the experimental site that it was becoming the dominant species in the grassland. Notably, the increase in D. bipinnata and S. brevifolium in year 3 was associated with a decline in the abundance of I. cylindrica in the managed plots. Whilst these changes may be correlated, they may not be a direct result of management. Fluctuations in the abundance of I. cylindrica on annually cut and burned sites, are reported by local people harvesting the grass for thatch. Whether the decline would have continued given further years of management is unknown.

Plant species richness

Managed plots, in particular cut and burned plots, had a greater species richness than unmanaged plots, a function of an increase in the number of forb species and a result of disturbance from the combined effects of cutting, burning and grazing. These disturbances resulted in an increased number of gaps in the sward. Gaps provide sites that can be colonized by a range of species that would otherwise be excluded by competition from the dominant species. Higher species diversity on the managed plots is consistent with a trade-off between colonizing ability and competitive ability (Horn & MacArthur 1972; Crawley & May 1987; Tilman 1994). Implicit here is the assumption that the forbs are inferior competitors but have a greater capacity to colonize gaps than the dominant grasses. The dynamics of gap colonization in the phantas are, however, unknown. Recruitment is likely to occur following canopy removal after cutting and burning in the dry season, but only a few individuals will persist in gaps, most being competitively excluded as the dominant grasses regenerate.

Management and succession

A primary reason given by protected area managers for cutting and burning I. cylindrica dominated grasslands has been to prevent succession to taller grass swards dominated by Narenga, Saccharum and Themeda species, thereby maintaining a grazing resource for ungulates. The taller species form a dense sward unsuitable for some ungulates and are of low palatability (Lehmkuhl 1989; Peet 1997). This has made managers unwilling to consider leaving patches of grassland unmanaged. Leaving I. cylindrica dominated grassland unmanaged over the short-term (2–3 years), however, does not appear to have successional consequences, in terms of the establishment and persistence of taller grass species or trees and shrubs. There was no evidence that tall grasses were able to establish more successfully in unmanaged plots than managed, although mature individuals of the two tall grass species Erianthus ravennae and Saccharum bengalense were only present at low frequencies.

The small increase in total tree and shrub cover in year 3 in managed plots was a result of recruitment of Adina cardifolia seedlings across the phanta. Plants were small (< 5 cm high), however, and such events are reported to be a regular occurrence with seedlings usually failing to survive their first dry season and the accompanying fires (S.R. Jnawali, personal communication). Such small, temporary changes would not affect ungulate utilization of the grasslands.

In the longer term there is evidence that I. cylindrica dominated swards, in the protected areas, have declined in area as a result of succession to tall grassland, despite regular cutting and burning (Pokharel 1993; Lehmkuhl 1994). Whilst the available literature refers to I. cylindrica as dominating under fire regimes (Seth 1970; Dabadghao & Shankarnarayan 1973; Kushawa, Ramakrishnan & Tripathi 1983; Soerjani, Eussen & Tjitrosudirdjo 1983; Brook 1989; Riswan & Hartanti 1995), succession to secondary forest has been recorded under conditions of regular burning (Hubbard et al. 1944; Soerjani, Eussen & Tjitrosudirdjo 1983). Invading species, which include a number of tall grasses, are typically deep-rooted, fire-resistant and can outcompete I. cylindrica for light (Soerjani, Eussen & Tjitrosudirdjo 1983). Given time, therefore, they may be expected to dominate the I. cylindrica grasslands under the current management regimes, once they become established.

Implications for future management

Leaving plots unmanaged over 3 years did not cause a dramatic turnover of species, a change in the composition of the main structural grass species or lead to succession to tall grassland or scrub. These results allow rotational patch management, whereby patches of grassland are left unmanaged for 2–3 years and are then cut and burned again, to be considered as one of three possible management scenarios for I. cylindrica dominated grassland: (a) maintenance of the status quo, with annual cutting and burning of nearly all I. cylindrica dominated grassland; (b) stopping annual cutting and burning; and (c) rotational patch harvesting of grassland, on a biennial or triennial basis.

Option (b), ending the practice of cutting and burning would remove the benefit of regenerating grassland to ungulates. Whilst I. cylindrica might temporarily increase in abundance, mature grasses are of poor palatability to ungulates, and new shoots would not be available or would be difficult to locate in the dense litter. A build up of fuels would mean that natural fires would be hotter, more destructive and less easy to control than at present. Local people would also no longer have access to grassland resources. The lack of obvious conservation value and the drastic impact on local people make this option undesirable.

Option (a), continued widespread cutting and burning, would provide ungulates with an important forage resource from the regenerating grassland (Mishra 1982a; Moe & Wegge 1997; Peet 1997) and local communities with an important subsistence resource. This form of management, however, does not ultimately prevent succession to forest and has been demonstrated to be deleterious to disturbance intolerant or cover-dependent species such as the hispid hare (Bell 1986) and pygmy hog (Oliver 1980) and is probably deleterious to a range of other small mammals, herpetofauna and invertebrates. For these species, rotational patch cutting and burning of the grassland, option (c), would provide suitable habitat for population persistence, with unmanaged patches of grassland having a dense cover of grasses and litter.

Given the paucity of data available for most cover-dependent species, initial patch design could be based on the limited data available for the hispid hare. Hispid hare densities have been estimated at 0·68 animals ha–1, although this density is calculated from narrow strips of tall unburned grassland postfire, and densities in areas of more suitable habitat are unknown (Bell, Oliver & Ghose 1990). This suggests that unmanaged patches would need to be several hectares each in area. Hares are also associated with mixed shorter and tall grassland (Peet 1997) and patches would best be sited on the boundary between shorter I. cylindrica grassland and tall grassland. In Bardia, the total area of I. cylindrica dominated grassland is small (< 340 ha), so the total area removed from cutting and burning should not in the first instance be more than 25–50 ha. In Suklaphanta, the area of grassland is larger and a greater area of grassland could be left unmanaged.

Whilst cover-dependent species would benefit from the adoption of a patch management system, decisions regarding the removal of grassland from the current cutting and burning regime should take account of species that benefit from it, principally ungulates (Moe & Wegge 1997; Peet 1997). In Bardia, I. cylindrica dominated grassland is utilized by chital, swamp deer and nilgai, following fire (Dinerstein 1979b; Moe & Wegge 1997; Peet 1997), and in Suklaphanta by swamp deer and chital (Schaaf 1978). The grasslands are probably most important for the swamp deer and nilgai populations, as chital also feed extensively in riverine forest (Mishra 1982a; Moe & Wegge 1994). Being highly mobile, these species would be able to move between patches of burned grassland to exploit the regenerating grass shoots. Removing small amounts of the grazing resource is unlikely to impact on ungulate populations as the resource is super-abundant following cutting and burning.

Local communities that harvest thatch must also be considered if grassland management is altered (Brown 1997). Local communities could continue to harvest thatch under a patch management system, but the available resource would be smaller. To avoid an increase in park–people conflict, provision of alternatives to thatch, especially tiles, would need to be explored to compensate for any increased shortfall in the thatch supply (Peet 1997).

Rotational patch harvesting: implications for the grassland plant community

Experiments to investigate the impacts of cutting and burning on plant communities can rapidly become very complicated, when the frequency of treatments is also altered. Although the effects of cutting alone and burning alone were sometimes additive in cut and burned plots, the removal of biomass, whether that was through cutting, burning, or cutting and burning, had similar impacts on plant community structure and composition in this experiment. Additive effects had little impact on grassland structure and composition. For instance, despite the additive effects of cutting and burning on the total cover of grasses and total cover of forbs, total grass cover remained at over 70% and total forb cover below 6% in cut and burned plots. The additive effects would not have implications for animal utilization of the grassland or for biodiversity conservation. This means that future experiments to investigate the impacts of patch harvesting on community dynamics can use a single treatment (e.g. cutting and burning). Simpler factorial experiments can now be designed to investigate the impacts of the frequency of biomass removal on community dynamics, than if cutting and burning had to be considered as separate treatments.

A priority is to investigate the impact of returning unmanaged patches to a regime of fire and cutting after 2–3 years and to determine whether the characteristics of the fires are different from those of fires in annually burned grassland. Increased litter, in patches left unmanaged for 2–3 years, might mean hotter, more intense fires, although variations in intensity are less in grasslands than woody communities (Bond & van Wilgen 1996). Variation in fire intensity can affect plant survival and regeneration, especially for woody species (Trollope 1974; Walker 1981; Wright & Bailey 1982; Mallik & Gimingham 1985; Hobbs & Atkins 1988). Relatively little is known about the interaction between species survival and fire intensities in grassland (Bond & van Wilgen 1996). As very little living biomass of grasses is destroyed by fires, because the grasses are largely senescent at the time of burning, the influence of different intensity fires may be more important for woody herbs such as Desmodium spp. and for tree saplings, which are less likely to survive hotter fires. Less frequent but hotter fires may therefore retard succession to forest.

While the results from the experiment reported here indicate that patch management is a viable option within the I. cylindrica grassland assemblage it must be remembered that it is only one of a number of assemblages (Peet et al. 1999) in the protected areas. Patch management therefore requires investigation in a range of these other assemblages, with the aim of understanding the implications of patch management for the grassland system rather than a single species assemblage. Adoption of a system of patch management will allow conservation of those species that require cover and those that benefit from cutting and burning of grassland.

Acknowledgements

This work was primarily funded by the Darwin Initiative for the Survival of Species. Support was also received from the NERC (grant no. GR9/1276′A′), Fauna and Flora International, the International Trust for Nature Conservation and the Sarnia Charitable Trust. Permission to work in the protected areas was generously granted by The Department of National Parks and Wildlife Conservation, Nepal. We are grateful to Director Generals of the DNPWC, Dr U.R. Sharma and Dr T.M. Maskey, and the warden and staff of Royal Bardia National Park for their assistance, in particular Mr P.B. Shrestha, Mr R. Thapa, Mr Pathak, Mr Sharma and Mr Rupaketi. Dr K. Rajbhandari from the National Herbarium Kathmandu provided invaluable assistance in plant identification. Logistical support in the field was provided by the King Mahendra Trust for Nature Conservation.

Received 20 March 1998; revision received 12 February 1999

Ancillary