Imperata cylindrica (also known as alang alang or cogon grass) is an aggressive grass species that colonizes land cleared for unsustainable slash-and-burn agriculture or degraded by repeated forest fires (Kinnaird & O'Brien Reference KINNAIRD and O'BRIEN1998). Once established, subsequent burning increases losses of nitrogen and carbon, thereby eroding agricultural productivity and enhancing regeneration of I. cylindrica (Chapin et al. Reference CHAPIN, ZAVALETA, EVINER, NAYLOR, VITOUSEK, REYNOLDS, HOOPER, LAVOREL, SALA, HOBBIE, MACK and DÍAZ2000, Otsamo et al. Reference OTSAMO, ÅDJERS, SASMITO HADI, KUUSIPALO, TUOMELA and VUOKKO1995). Imperata cylindrica grassland houses a depauperate and low-stature plant community compared with proximate rain forest (Slik & Eichhorn Reference SLIK and EICHHORN2003). In Indonesia, millions of hectares of land, previously covered by lowland dipterocarp forest, are dominated by I. cylindrica grasslands due to repeated logging and fire (Estrada & Flory Reference ESTRADA and FLORY2015, MacDonald Reference MACDONALD2004, Otsamo et al. Reference OTSAMO, ÅDJERS, SASMITO HADI, KUUSIPALO, TUOMELA and VUOKKO1995). During initial fires, more fuel is produced than destroyed and this acts as a catalyst for future fires, which are more likely to occur and more intense. This positive feedback destroys rain-forest trees and promotes grassland invasion (Laurance Reference LAURANCE2003).

In the present study, we sampled shrubs, trees and climbers > 1 cm dbh (hereafter called STC), ferns, grasses and herbaceous dicots (hereafter called herbs) and insects (butterflies and odonates) in 16 plots allocated randomly to a 450-ha typical I. cylindrica-dominated landscape (Cleary Reference CLEARY2003, Cleary et al. Reference CLEARY, MOOERS, EICHHORN, VAN TOL, DE JONG and MENKEN2004) in East Kalimantan Indonesia (1º03'S, 116º57'E) from 11 July–12 August 2000. STC were sampled in 10 × 20-m plots and herbs in 2 × 2-m subplots nested within the larger plots. Butterflies and odonates were sampled along 300-m transects crossing the 10 × 20-m plots following previously described methods (Cleary et al. Reference CLEARY, MOOERS, EICHHORN, VAN TOL, DE JONG and MENKEN2004). The hypotheses of the present study were that (1) I. cylindrica grassland harboured a spatially variable plant community, (2) variation in the plant community structured spatial variation in grassland insect composition and (3) I. cylindrica grassland mainly harboured widespread plant and insect species. Despite the prevalence of I. cylindrica grasslands, relatively few studies have assessed the biodiversity housed within these degraded environments. Only two studies so far have assessed butterflies (Matsumoto et al. Reference MATSUMOTO, NOERDJITO and FUKUYAMA2015, Nakamuta et al. Reference NAKAMUTA, MATSUMOTO and NOERDJITO2008) and to the best of our knowledge none assessed odonates.

All analyses were performed in R (https://www.r-project.org/). Abundance and cover matrices of STC, herbs, butterflies and odonates were imported into R and log_e (x + 1) transformed and distance matrices constructed using the Bray–Curtis index with the vegdist() function in VEGAN (https://cran.r-project.org/web/packages/vegan/vegan.pdf). Variation in composition was assessed with Principal Coordinates Analysis (PCO) using the cmdscale() function in R with the Bray–Curtis distance matrix as input. Weighted averages scores were computed for species on the first two PCO axes using the wascores() function in vegan. In each plot, we also measured/estimated a number of habitat variables, namely, altitude, slope/inclination (with a clinometer), number of STC stems, STC basal area and the percentage cover of herbaceous monocots, herbaceous dicots and ferns. Using the envfit() function in vegan, we tested for relationships between these variables, the first two axes of the PCO ordinations of STC and herbs (as proxies for gradients in vegetation composition) and the PCO ordinations of butterflies and odonates. Habitat variables with a P < 0.10 were included in the ordinations of butterflies and odonates.

Altitude in the study area varied from 23 to 49 m asl. We recorded a total of 43 STC species including nine morphospecies. The number of stems per plot varied from 5 to 143 and STC basal area from 0.5 to 1336 cm². The plot with a basal area of 1336 cm² contained two very large Cocos nucifera palms. The next largest basal area was only 146 cm² (plot Ka09). The most abundant STC species were Chromolaena odorata (272 stems; native to North America), the climbing bamboo Dinochloa scandens (100 stems), Melastoma malabathricum (26 stems), an unidentified Labiatae (13 stems), the climber Mikania scandens (eight stems; native to North America), Vitex pinnata (six stems), Ficus sp. (six stems), Lygodium microphyllum (five stems), Ficus grossularioides (four stems) and Dillenia excelsa, Manihot esculenta, Artocarpus heterophyllus, Cocos nucifera, Piper aduncum, Brucea javanica, Callicarpa pentandra and Lantana camara (all three stems). The main gradient in STC composition (Figure 1a) separated plots (Ka03, Ka04, Ka16) with relatively high abundances of Melastoma malabathricum, Ficus grossularioides, Mikania scandens, Lygodium microphyllum and the unidentified Labiatae from other plots (Ka06, Ka08, Ka11, Ka14, Ka15). These latter plots were characterized by very low basal areas (< 2.08 cm²). The second PCO axis separated plots (Ka09 and Ka13) with high abundances of Dillenia excelsa, Stenochlaena palustris and Ficus sp. from plots (Ka02 and Ka10) with high abundances of Brucea javanica, Callicarpa pentandra, Piper aduncum, Vitex pinnata, Cocos nucifera, Musa sapientum, Manihot esculenta, Artocarpus heterophyllus and Dinochloa scandens. Chromolaena odorata was present in all but one plot.

Figure 1. Variation in the composition of shrubs, trees and climbers (a), herbs (b), butterflies (c) and odonates (d) in an Imperata cylindrica-dominated grassland habitat in East Kalimantan, Indonesia. Ordinations show the first two axes of the PCO analyses. Red symbols and codes refer to values for plots while black codes refer to loadings for selected species (Table 1). For species codes see Table 1. Abbreviations of habitat variables (indicated by blue arrows) are: Basal: STC basal area, Fern: fern cover, Grass: grass cover, HerDic: herbaceous dicot cover, Herb.pc1: first axis of PCO analysis of herbs.

A total of 16 herb species were observed, of which eight were morphospecies, including Imperata cylindrica (64.8%), Scleria sp. (6.3%), Spermacoce sp. (5.1%), an unidentified grass (4.6%), Vernonia cinerea (0.8%), Saccharum spontaneum (0.6%), Alpinia galanga (0.3%) and Pteridium sp. (0.1%). The main gradient in herb composition (Figure 1b) separated plots with relatively low (Ka02, Ka03, Ka06 and Ka09) versus high (remaining plots) cover of I. cylindrica. The second axis separated plots (Ka02 and Ka06) with a relatively high cover of the monocot Scleria sp. from plots (Ka03 and Ka09) with a relatively high cover of the herbaceous dicots Spermacoce sp. and Vernonia cinerea. Imperata cylindrica covered > 85% of all plots with the exception of Ka02, Ka03, Ka06 and Ka09 where the cover of the grass varied from 0–1%. Spermacoce sp. and Vernonia cinerea covered > 80% of Ka06 while Scleria sp. covered 95% of Ka09. Plot Ka02 was covered by the climber Dinochloa scandens. Ka03 had relatively low herbaceous cover (< 3%) and a relatively high STC basal area (130 cm²).

A total of 67 butterfly species (3200 individuals) and 30 odonate species (1653 individuals) were recorded in the study area. Dominant butterfly species included Taractrocera ardonia, Orsotriaena medus, Spindasis lohita, Ypthima pandocus, Neptis hylas, Zizina otis, Mycalesis perseus, Potanthus omaha, Polytremis lubricans and Telicota besta (Table 1). The most dominant butterfly species (Taractrocera ardonia) accounted for 10.9% of all individuals. Dominant odonate species included Neurothemis fluctuans, Orthetrum sabina, Diplacodes trivialis, Neurothemis terminata and Rhyothemis phyllis. The most dominant odonate species (Neurothemis fluctuans) accounted for 36.1% of all individuals.

Table 1. Family, (morpho)species and abundance or percentage cover (herbs) of the most abundant STC, herbs, butterflies and odonates. The ‘Code’ field refers to loadings for species shown in the ordinations of Figure 1.

For butterfly composition, habitat variables selected using the envfit function in R (P < 0.10) were fern cover (R² = 0.346, P = 0.065), STC basal area (R² = 0.497, P = 0.061), the first axis of the herb PCO analysis (herb.pc1: R² = 0.689, P = 0.002), grass cover (R² = 0.736, P < 0.001) and herbaceous dicot cover (R² = 0.835, P < 0.001). Butterfly species associated with higher STC basal area and herb.PC1 (Figure 1c) included Elymnias hypermnestra, Ypthima pandocus, Mycalesis horsfieldi, Eurema hecabe, Neptis hylas and Jamides celeno. Elymnias hypermnestra, Eurema hecabe, Neptis hylas and Jamides celeno are all widespread species that occur well beyond Borneo and the Sundaland region while Ypthima pandocus has a more restricted range and is only found in the Sundaland region. Information on the hostplant use of these butterfly species can be found in Corbet & Pendlebury (Reference CORBET and PENDLEBURY1992), Suguru & Haruo (Reference SUGURU and HARUO1997, Reference SUGURU and HARUO2000) and http://www.nhm.ac.uk/our-science/data/hostplants/. Elymnias species feed mainly on palms and grasses while Eurema hecabe is a generalist that feeds on a wide range of herbs and trees. Neptis hylas is another generalist that feeds on herbs, lianas and trees in various plant families. Jamides celeno feeds on plants in the families Fabaceae, Meliaceae and Sterculiaceae and was the first species to recover from forest fires following the 1997/98 ENSO event where it fed on resprouts of the tree Fordia splendidissima (Cleary & Grill Reference CLEARY and GRILL2004). Ypthima pandocus only feeds on grasses.

Species mainly associated with high grass cover included Potanthus trachala, P. omaha, Orsotriaena medus and Telicota besta. All of these species are hesperids with a moderate to wide distribution. They also all feed on grass species including I. cylindrica and Oryza sativa. Butterfly species associated with herbaceous dicot plant cover included Junonia atlites and Zizina otis, both of which are widely distributed.

For odonate composition, habitat variables selected using the envfit function in R (P < 0.10) were herb.pc1 (R² = 0.356, P = 0.051), grass cover (R² = 0.536, P = 0.018) and STC basal area (R² = 0.605, P = 0.009). Odonate species associated with a greater STC basal area (Figure 1d) included Brachydiplax chalybea and Neurothemis terminata. Brachydiplax chalybea and Neurothemis terminata are both widespread and common dragonfly species that breed in ponds and marshes (http://indiabiodiversity.org/). Odonates associated with higher grass cover included the common and widespread species Ceriagrion cerinorubellum, Tholymis tillarga, Orthetrum sabina and Pantala flavescens (http://indiabiodiversity.org/).

Jones et al. (Reference JONES, SUSILO, BIGNELL, HARDIWINOTO, GILLISON and EGGLETON2003) previously showed that termite composition, abundance and richness differed strongly in I. cylindrica grasslands compared with intact forest with the former housing impoverished faunas. Matsumoto et al. (Reference MATSUMOTO, NOERDJITO and FUKUYAMA2015) and Cleary et al. (Reference CLEARY, MOOERS, EICHHORN, VAN TOL, DE JONG and MENKEN2004) also showed that butterfly richness was much lower in I. cylindrica grassland than intact forest. Matsumoto et al. (Reference MATSUMOTO, NOERDJITO and FUKUYAMA2015) recorded 14 butterfly species in I. cylindrica grassland areas close to our study area. This is much lower than the 67 species we recorded, but the present study had a much higher sampling effort; they only sampled 72 individuals compared with our 3200. All of the species reported by Matsumoto et al. (Reference MATSUMOTO, NOERDJITO and FUKUYAMA2015), with the exception of Appias olferna and Pantoporia paraka, were recorded in the present study.

This study also revealed a prevalence of common, invasive and widespread plant, odonate and butterfly species. The grasslands were not, however, entirely homogeneous. Although some areas were devoid of large woody plants, they still existed in the larger area, particularly around meandering streams. Some areas were also covered by other herbaceous species such as Scleria and Spermacoce species and the variation in vegetation structure was related to variation in butterfly and odonate composition. Although wide-ranging species dominated the I. cylindrica grassland habitat including very wide-ranging species such as Lampides boeticus that ranges from Europe to Australia, there were some exceptions. Restricted-range (mainly to Sundaland) species recorded included Amathusia ochraceofusca, Deudorix staudingeri, Elymnias harteri, Elymnias panthera, Pandita sinope and Ypthima pandocus. These species, however, can also be found along edges and in gaps of natural forest in the vicinity of the I. cylindrica grassland studied (Cleary et al. Reference CLEARY, MOOERS, EICHHORN, VAN TOL, DE JONG and MENKEN2004). The present study shows that I. cylindrica grassland harbours a variable vegetation albeit dominated by the grass I. cylindrica and that variation in vegetation composition appears to play a role in structuring resident insect communities.