Plant communities can predict the distribution of solitarious desert locust Schistocerca gregaria



    Corresponding author
    1. Wageningen University, Department of Plant Sciences, Group Crop and Weed Ecology, PO Box 430, 6700 AK Wageningen, the Netherlands;
    Search for more papers by this author

    1. Wageningen University, Department of Plant Sciences, Laboratory of Entomology, PO Box 8031, 6700 EH Wageningen, the Netherlands;
    Search for more papers by this author
    • *

      Present address: International Livestock Research Institute (ILRI), PO Box 853, Mekele, Ethiopia.


    1. Wageningen University, Department of Plant Sciences, Laboratory of Entomology, PO Box 8031, 6700 EH Wageningen, the Netherlands;
    Search for more papers by this author

    1. Food and Agricultural Organization of the United Nations, PO Box 14, Khartoum North, Sudan; and
    Search for more papers by this author
    • Present address: Commission for Controlling the Desert Locust in the Central Region (CRC), PO Box 2223, Cairo, Egypt.


    1. Wageningen University, Department of Environmental Sciences, Group Nature Conservation and Plant Ecology, Bornsesteeg 69, 6708 PD, Wageningen, the Netherlands
    Search for more papers by this author

Wopke van der Werf, Wageningen University, Department of Plant Sciences, Group Crop and Weed Ecology, PO Box 430, 6700 AK Wageningen, the Netherlands (fax +31 317 48 55 72; e-mail


  • 1The desert locust is a migratory pest whose population development in remote areas must be monitored to prevent outbreaks, upsurges and plagues. Monitoring would be very much facilitated if the area of search could be restricted to sites of likely population increase.
  • 2The spatial distribution of solitarious desert locusts on the Red Sea coastal plain of Sudan was determined over 3 years from November to March. Additional observations were made on habitat factors, such as plant community, soil texture, soil moisture and land use.
  • 3Locust densities varied according to the amount and distribution of rainfall and longevity of the annual green vegetation, with virtually no locusts being observed in the driest season.
  • 4Samples on a grid of 120 sites within a 120-km stretch of coastal plain showed that locusts were prevalent only in the millet–Heliotropium plant community, which is found at sites with a fine sandy soil texture and comparatively high and long-lasting soil moisture in wadi deltas. These sites constitute less than 5% of the area of this part of the plain.
  • 5An accessory study showed association between desert locust distribution and millet cropping in an area where no Heliotropium was found. Other samples confirmed the association between solitarious desert locust and millet agriculture.
  • 6Synthesis and applications. The results indicate that surveys for early detection and control of desert locust on the Red Sea coast of Sudan can focus on millet cropping areas. The results suggest that the efficiency of monitoring migratory pest outbreaks in remote areas could be enhanced by using associations between plant communities and herbivorous insects to predict risk areas and target survey efforts.


The desert locust Schistocerca gregaria Forsk. is a potentially devastating pest insect inhabiting the arid and semi-arid areas of northern Africa, the Arabian Peninsula and south-west Asia. Its geographical range varies from a recession area of 16 million km2 covering more than 25 countries, when locust densities are low, to an invasion area of 29 million km2 covering over 65 countries, when locust densities are high (Pedgley 1981). The physiology and behaviour of the desert locust are related to population density: low densities induce the solitarious phase whereas high densities induce the gregarious phase. Adult gregarious locusts move in swarms that can travel thousands of kilometres in search of suitable habitat, threatening agriculture and rural livelihoods. Modern locust management aims at early intervention to prevent plagues.

An ability to predict population processes is essential to the development of efficient and environmentally benign control strategies (Cressman 1996). Predictions about temporal dynamics of locusts may be derived from weather data in relation to analysis of past records (Todd et al. 2002). Locust surveys are rendered more efficient if habitats that are likely to harbour significant numbers of locusts are identified before sampling. Information on suitability of habitats is used by the Australian Plague Locust Commission to target surveys of Australian plague locust Chortoicetes terminifera Walker (Hunter 2004). This species is associated with habitats containing the host species Mitchell grass Astrebla spp. (Lindl.) (McCulloch & Hunter 1983; Hunter 1989). In the case of desert locust, the build-up of locust populations is routinely monitored using ground searches (FAO 1994). These searches cannot cover the very large breeding areas entirely and, hence, outbreaks may remain undetected (Symmons 1992).

Knowledge about the habitats of the desert locust is partly based on reports of reconnaissance surveys such as those by Popov & Zeller (1963), in which remote desert areas were traversed by vehicle for weeks or months at a time. By necessity, such surveys are selective, focusing on habitats where locusts are expected or reported to be present by word of mouth. As a consequence, the information obtained in such surveys gives an incomplete and potentially biased estimate of the likelihood of finding solitarious desert locusts in different habitats (Yates 1960). Another information resource for surveys is the monograph on biotopes of the desert locust in Northern Africa compiled by Popov, Duranton & Gigault (1991). This compilation is only indirectly based on field data, and the spatial scale at which land units are defined (2700–170 000 km2) is too large to be provide guidance for field surveys by vehicle, which would require information at scales from 1 to 100 km.

The Red Sea coastal plains were identified as an important breeding area for desert locust by Johnston (1926) and Maxwell-Darling (1936, 1937). There are several reports indicating that desert locusts are mostly observed in areas with millet Pennisetum typhoideum Rich and Heliotropium spp. (Maxwell-Darling 1936, 1937; Stower, Popov & Greathead 1958; Waloff 1963; Roffey & Stower 1983) within the Red Sea coastal plain. However, locusts have also been recorded in Panicum grassland and bare desert (Popov & Zeller 1963). Thus, the habitat preferences of solitarious desert locust on the Red Sea coastal plains are not clear.

The coastal plain has a hot arid to semi-arid climate, with low and erratic rainfall falling almost exclusively from October to March (Griffiths & Hemming 1963; Satakopan 1965). Wadies traverse the plain from west to east, spreading out in deltas before they reach the sea, releasing water and depositing fine soil particles. At these deltas, millet is cultivated during the wet season.

Woldewahid (2003) described four plant communities on the Sudan coastal plain: the Suaeda monoica scrubland near the coast, the Acacia tortilis scrubland near the Red Sea Hills, the Panicum turgidum grassland at intermediate location and altitude, and Heliotropium–millet, small pockets of cropland (mostly planted with millet) at the transition between the Panicum grassland and the Suaeda scrub. The croplands are characterized by relatively good moisture provision because of run-on water from the spreading wadies, and high abundance and vegetation cover of Heliotropium arbainense, a plant species strongly associated with solitary locusts (Woldewahid et al. 2004). Perennial plants are weeded out by hand (e.g. Calotropis procera) and water run-on from wadies is regulated by dykes. All plant communities produce annual undergrowth if there is enough rainfall.

The purpose of the work presented here was to refine our knowledge of those habitats that provide suitable breeding conditions for the desert locust on the Sudan Red Sea coast. The sampling design was spatially refined and unselective to identify site conditions that are indicative of high densities of locusts and might be used for devising more efficient monitoring methods.

Materials and methods

The study was conducted in four areas (Fig. 1). The main area, south of Port Sudan, covered a 120-km stretch of coastal plain approximately 20 km wide. The area contained 60 sample sites in the first season of the study, and 64 sites in the subsequent two seasons (Fig. 1b). The sites were laid out in an approximately 5 × 5-km grid. Locusts were counted during three successive rainy seasons: 1999–2000, 2000–2001 and 2001–2002. In the second rainy season, more sites were sampled in the Heliotropium-millet community, which was small in spatial extent but was found to be a significant habitat for locusts during the first season. In each season there were eight or nine sampling events. Sampling was performed using two teams, each consisting of three observers plus a driver in an all-terrain vehicle. Sampling the 5 × 5-km grid took 3 days: the date of sampling is given as the second day.

Figure 1.

Study areas in the coastal plain of Sudan. (a) The coastal plain, with the major towns and wadies. The rectangle indicates the main study area. (b) The locations of sample sites in the main study area in 1999–2000 (open circles) and in the subsequent two seasons, 2000–2001 and 2001–2002 (closed triangles). Rectangles indicate the two smaller study areas, i.e. the delta of the Wadies Gwob and Gabol (c) and the delta of the Wadi Baraka, south of Tokar (d). Symbols in (c) have the same meaning as in (b). Symbols in (d) indicate the plant community: Prosopis chilensis (open circles), Brachiaria sp. (closed circles), Sorghum–Solanum (open triangles) and Dipterygium–Medicago (closed triangles).

The second study area included the crop land (mostly millet) at the delta of the Wadies Gwob and Gabol, a few kilometres south of Suakin, as well as the nearby Panicum grassland (Fig. 1c). The cropland was sampled at a fine grained spatial resolution, using a 1 × 1-km grid of sample points, while the Panicum grassland was sampled using a rectangular 1 × 2-km grid of sample points. After the first season the sampling in the grassland was discontinued, while the sampling design for the crop area was extended to include sites in adjacent Suaeda monoica and Panicum turgidum plant communities (Fig. 1c) to determine the variation in locust density between close sites in different plant communities. The total number of sites in the Gwob Gabol area was 60 in the first season and 33 in the second and third seasons. There were nine sampling events in the first and second seasons, and only four in the third because of drought.

The third study area was in the delta of the Wadi Baraka near Tokar (Fig. 1d). Samples were collected here during the third season. The sampling grid in the Tokar delta contained 45 points in a 2 × 2-km grid. Soil in the delta was clayish and there was a dependable water supply from the Wadi Baraka. The crops (predominantly millet, sorghum and cotton) were grown under flood recession farming. Plant communities in this area were delineated using two-way indicator species analysis (twinspan) (Hill 1979), as previously described for the main study area (Woldewahid 2003).

At each sample site in the above three study areas, three observers counted all adult desert locusts along a 400 × 1-m transect. The three transects were parallel and approximately 10 m apart. The location of sites was recorded with geographic positioning (GPS Garmin 12XL, Olathe, KS) so the same sites could be revisited.

Finally, samples were taken in the second and third seasons throughout the whole Sudanese Red Sea Coast, from the Egyptian to the Eritrean border (Fig. 1a) to verify relationships observed in the three selected study areas. Sampled sites for the verification were classified as cropland (mostly millet), green grazing land (judged by the presence of substantial recent regrowth of Panicum turgidum or the presence of annuals between dry Panicum turgidum) and dry grazing land. The sites were visited from November to December 2000 and from November 2001 to February 2002. Presence or absence of locusts was checked on three parallel transects of 400 m.

Vegetation relevés were made at least once a season at all sites where locusts were sampled. Cover abundance for individual plant species was scored using a 9-point ordinal scale (van der Maarel 1979). During each sample, the status of the vegetation was recorded as greening, green, drying or dry. Particle size distribution of all sample sites used in the three seasons was determined by collecting five soil samples from each of three 10 × 10-m plots at each sample site. Soil texture was determined by dry sieving (Woldewahid 2003). Presence of moist soil was recorded every 7–14 days at five locations at each sample site. Soil was scooped away with a hand shovel and presence of moisture was assessed at three depths (0–5, 5–10 and 10–15 cm) by hand feeling. The results were coded as 0 (no moisture), 1 (moisture at 10–15 cm), 2 (moisture at 5–15 cm) or 3 (moisture throughout). The median score of the five points checked per sample site was used in analysis. Grazing pressure at each sample site was visually assessed from the presence of camels and goats and from the signs of their feeding, and classified as 0 (none), 1 (slight), 2 (moderate) or 3 (heavy).

Association between locust density and habitat variables such as plant community composition, soil particle size, soil moisture, grazing pressure and altitude were quantified and visualized with principal component analysis (PCA), using the software program canoco4 (ter Braak & Šmilauer 1998). Differences in locust density among different plant communities were tested for significance, using Kruskal–Wallis non-parametric anova in SPSS for Windows V.10. The test was conducted separately for each sampling date.


relation between solitarious locusts and plant community in the main study area

In the first season, the highest locust densities were found at sites with the Heliotropium–millet plant community (Fig. 2a). Considerably lower locust densities were found in the Panicum grassland and the Suaeda scrub. The lowest locust densities were found in the Acacia scrub. In the second sampling season, locusts were exclusively found in the Heliotropium–millet plant community (Fig. 2b). In the third season, scarcely any locusts were found (Fig. 2c).

Figure 2.

Locust density (number ha−1; average ± SEM) in four plant communities in the main study area during the seasons of (a) 1999–2000, (b) 2000–2001 and (c) 2001–2002. Bars indicate standard error of the mean.

relations between temporal trends in locust density, soil moisture and state of the vegetation in the main study area

Locust densities were consistently much higher (Kruskal–Wallis test, inline image > 7·81, P < 0·05 at each sampling date) in the Heliotropium–millet plant community than in the other plant communities throughout the first rainy season (Fig. 2a). Cumulative rainfall patterns for the 3 years are shown in Fig. 4. During January 2000 soil moisture levels decreased and the annual plants started to dry out, but this decrease was slower in the Heliotropium–millet community than in the others (Fig. 4a and Fig. 5a).

Figure 4.

Percentage of sample sites with 0–15 cm moist soil profile in four plant communities in the main study area in the seasons of (a) 1999–2000, (b) 2000–2001 and (c) 2001–2002.

Figure 5.

Percentage of sample sites with green vegetation in four plant communities in the main study area in the seasons of (a) 1999–2000, (b) 2000–2001 and (c) 2001–2002.

During the second sampling season, high locust densities were again associated with the Heliotropium–millet community (Kruskal–Wallis test, inline image > 11·34, P < 0·01 at each date) (Fig. 2b). There had been less rainfall than in the preceding year (Fig. 3), thus the percentage of sites with moist soil was lower (Fig. 4b) and the duration of green vegetation cover shorter (Fig. 5b). Locust numbers were lower than during the first season. On 31 December 2000 about 53 ± 13 locusts ha−1 were found in the Heliotropium–millet plant community and none in the other communities.

Figure 3.

Cumulative daily rainfall (mm) at Port Sudan during the seasons of 1999–2000, 2000–2001 and 2001–2002.

In the third season, the soils were quite dry (Fig. 4c) because of very limited rainfall (Fig. 3) and the vegetation stayed green for a shorter time than in the first two seasons (Fig. 5). Only a few scattered locusts were found (Fig. 2c).

multivariate analysis of relationships between locust distribution, site factors and plant communities

The PCA analysis of results from the main study area (Fig. 6 and Table 1) indicated that, during the first season, locust density was correlated positively and significantly with the first PCA species axis (r = 0·67) and was associated with the Heliotropium–millet plant community, fine textured soils (r = 0·54) and comparatively high moisture availability (r = 0·55). Locust density was negatively correlated with grazing pressure (r = −0·50) and, to a lesser extent, with altitude (r = −0·25). The PCA results for the second season were similar (correlation coefficients shown in Table 1). In the third season, too few locusts were found to establish significant correlations (Table 1).

Figure 6.

PCA ordination plot, based on plant community composition, showing the relationship of locust density (number ha−1), plant composition (axis 1 and 2) and other habitat variables in the main study area during the first (a) and second seasons (b) of the study. Symbols in the ordination plot indicate plant communities at sample sites: Suaeda monoica (▵), Heliotropium (○), Panicum turgidum (□) and Acacia tortilis (•).

Table 1.  PCA correlation matrix (r) between locust density (number ha−1), species’ first two PCA axes and habitat variables, based on results over 3 years in the main study area and results in the Tokar delta in 2001–2002
Habitat variablesLocust density (number ha−1)
Main study areaTokar delta
  • **

    Significant at P= 0·01;

  • *

    significant at P= 0·05;

  • NS

    , not significant at P > 0·05.

Axis 1+0·667**+0·589**+0·325*+0·608**
Axis 2−0·249NS−0·341**−0·012NS+0·467**
Soil particle size+0·535**+0·475**+0·114NS−0·663**
Moist soil+0·553**+0·589**+0·181NS+0·385**
Compaction  −0·161NS+0·114NS

locusts in the gwob gabol area

In all three sampling seasons, there was a marked difference in locust density between the Heliotropium–millet plant community and nearby sample sites in the Panicum turgidum and Suaeda monoica communities (Fig. 7; P-values in Kruskal–Wallis test were similar to those obtained in the main study area). The results confirm those of the main study area and show that important differences in the density of solitarious locusts can occur between nearby sample sites with different plant communities.

Figure 7.

Comparison of locust density (number ha−1) in the Heliotropium–millet plant community and in nearby sites in the Panicum turgidum and Suaeda monoica plant communities in the Gwob Gabol area in the seasons of (a) 1999–2000, (b) 2000–2001 and (c) 2001–2002. Bars indicate standard error of the mean.

results from the tokar delta

Based on the vegetation relevés in the Tokar delta in 2001, four plant communities were identified by two-way indicator species analysis.

The Dipterygium–Medicago plant community was characterized by Dipterygium glaucum Decne and Medicago L. This community occurred mainly on fine sand soils on which millet Pennisetum typhoideum was cultivated.

The Sorghum–Solanum plant community was characterized by Sorghum halepense L., Solanum dubium Fres and Gossypium barbadense L., and was mainly found in mixed crop fields on fine sand to silt-clay soils.

The Brachiaria plant community was characterized by Brachiaria sp. (Griseb), and occurred mainly on silt-clay soils cultivated with Sorghum bicolor (L.) Moench.

The Prosopis chilensis plant community was characterized by perennial shrubs such as Prosopis chilensis (Molina), Suaeda monoica and Cassia senna (L.).

Locusts were most numerous in the Dipterygium–Medicago plant community (millet fields) and almost absent in the Prosopis chilensis plant community (Fig. 8). The differences in locust density among plant communities were significant (Kruskal–Wallis test, inline image > 14·2, P < 0·003 at all sampling dates). According to PCA, locust density was positively correlated with soil moisture (Table 1) and negatively with the fraction silt plus clay (r = −0·66).

Figure 8.

Locust density (number ha−1) in relation to plant communities in the Tokar delta in the season of 2001–2002. Bars indicate standard error of the mean.

verification on the whole coastal plain

In 2000, locusts were observed in 82% of sites situated in wadies cultivated with millet or sorghum, whereas locusts were rarely detected in adjacent green (15%) or dry (3%) grazing sites. In 2001, locusts were detected in 92% of the cultivated sites while locusts were only rarely found in green (6%) or dry (2%) grazing sites.


The results of this study are important for designing improved desert locust survey programmes, indicating that surveys may be focused on areas with a Heliotropium–millet plant community or, in areas where Heliotropium is absent, in millet crops. These ‘target’ areas cover a small percentage of the coastal plain of Sudan and sampling resources can thus be much better deployed to areas that are likely to harbour significant locust numbers. The results further support the relationship between locust incidence and water availability to promote plant growth. When water availability was not limiting (e.g. in December 1999–January 2000; Fig. 4a), locusts were still effectively limited to the Heliotropium–millet plant community (Fig. 2a).

Several factors are likely to favour locust growth and reproduction in the Heliotropium–millet plant community compared with other communities: (i) food plants; (ii) soil type suitable for egg laying; (iii) soil moisture suitable for egg laying and conducive to plant growth; (iv) host plants inducing egg laying. Heliotropium and millet are among the best food plants for nymphal development and survival and adult reproduction and longevity (Abdel Rahman 1999). The fine sandy soil type found in the Heliotropium–millet community is the preferred soil type for locust egg laying (Abd El-Hadi & Hassanein 1968) and has the longest soil moisture duration, as demonstrated in this study. The nitrogen levels in host-plants on the Red Sea Coast in Sudan are low (2–4%), but those of food plants in the Heliotropium–millet community are up to a percentage point higher than those in the Panicum grassland (Woldewahid 2003). Such a difference is likely to have considerable effects on survival and fitness of locusts (Bernays & Simpson 1990).

Solitarious locusts can easily fly tens of kilometres, and the stark differences in locust density found in and around the Gwob Gabol cropland shows that they are quite precise about their choice of habitat. It is unknown whether the locusts aggregate in these ‘preferred’ areas through longer residence after random arrival or by directional flight. Behavioural responses to host-plants may be involved (Bashir et al. 2000).

Woldewahid et al. (2004) analysed the capacity of spatial (geo-)statistics for predicting locust densities, and found that geostatistical prediction can be made over distances not (much) greater than about 15 km. This distance shrinks as the rainy season comes to an end and the areas with green vegetation contract. Results presented here, especially those collected in the Gwob Gabol area where locust densities varied widely amongst adjacent plant communities, demonstrate that habitat (i.e. plant community) is the decisive variable for predicting locust density. Spatial correlations are especially promising when a single type of habitat occurs over a large expanse of terrain.

Recent research has highlighted the predictive value of plant species composition (plant communities) for the presence of arthropod communities (Raemakers et al. 2001). When compared with abiotic parameters and characteristics of vegetation structure, vegetation composition was found to provide the greatest predictive power with respect to the species composition of Araneae, Orthoptera, Carabidae, Curculionidae, Auchenorhyncha, Syrphidae and Apidae. Clear relations with plant communities have been found for grasshopper communities (Hemp & Hemp 2000) and molluscs (Horsák & Hájek 2003). Co-correspondence analysis, a new statistical method (ter Braak & Schaffers 2004) for comparing two corresponding data sets containing species compositions, respectively vegetation and arthropods, offers a useful tool to analyse and quantify such associations.

The important point for practical application in desert locust monitoring is that while 17–27% of our sample sites were in the Heliotropium–millet plant community, which in turn covered only about 5% of the study area, these samples contained 93–100% of the locusts detected. It is likely that similar associations will be found in other areas, potentially with other host-plants. For instance, in the Tamesna area of Niger and Mali, gregarizing locusts were reported to be confined to areas with the host-plants Tribulus ochroleurcus Maire and Schouwia purpurea L., which together occupy less than 10% of the study area (Roffey & Popov 1968). In the Thar Desert of India, Chandra (1984) reported that solitarious locusts were associated with specimens of Pennisetum typhoideum and Tribulus terrestris. Near the Wadi Wucharo in Eritrea, Stower & Greathead (1969) found gregarizing adults and hoppers on Heliotropium aegyptiacum Lehm. and rarely on other plants. Further work is necessary to confirm the relationships established for the Sudanese coastal plain in this study, especially under conditions that favour widespread gregarization, and determine whether and how they may be extrapolated to other locust breeding areas. The primary areas for extrapolation would be the other parts of the Red Sea Coastal plain, both on the African and the Arabian side. Based on the similarities in physiography of the landscape and plant communities it may be expected that similar relationships to those reported here might predict the spatial distribution of solitarious desert locust and serve as a guide for survey in this important breeding area of desert locust. The results presented here indicate that plant community relationships of desert locust could provide the key towards devising more efficient desert locust monitoring methods, which are critical to enable preventive control.


We thank Dr Tsedeke Abate of FAO and the staff of the Plant Protection Directorate of Sudan for their contributions to field work, Dr Peter Buurman of the Laboratory of Soil Science and Geology, Wageningen University, for his advice on soil particle size analysis, and Dr Magzoub Bashir of International Centre of Insect Physiology and Ecology for use of laboratory facilities at Port Sudan. Valuable comments on earlier versions of the manuscript were given by Paula Westerman, Jan Goudriaan and Keith Cressman.