The flowers of F. humilis are 15 mm–20 mm in diameter and white with a yellow centre (Figure 1a, b). The petals, especially towards the tip, are occasionally tinged pink. Flowers are either carried on very short stalks or are entirely stalkless. The pollen is yellow and tricolpate (Figure 2). The sepals resemble the leaves and form a short tube (Burgoyne et al. 2000). Frithia flowers are self-incompatible (Burgoyne et al. 2000), a common characteristic of South African Aizoaceae (Mayer & Pufal 2007). After pollination the flowers turn yellow or pink before expiring (Figure 1b), which may serve as an indication to pollinators that the flower is pollinated and without reward (Burgoyne et al. 2000).

Based on familial pollination syndromes (Hartmann 1991), F. humilis flowers could be Melittophilous because they are wide open, petals are shiny, whitish and open diurnally (from mid-morning to mid-afternoon) for several days. Pollen is less abundant than in most Melittophilous species but easily accessible. Further support for this syndrome is the self-incompatible sexual phase, with flowers reported to have a distinct male phase (protandrous) followed by a female phase approximately 4 days later (Hammer 1995; Hartmann 1991). Stigmata are initially shorter than stamens, elongating and emerging only at the beginning of the female phase when the male phase ends.

Two sites were chosen as receptor sites based on pre-selected criteria (Burgoyne & Hoffmann 2011) for plants translocated from a coal mine (Figure 3). Half of the affected population was translocated to typical Ecca Group sandstone which corresponded with the geology of the donor site. The other half of the population was translocated to less ideal outcrops of the sedimentary Wilge River Formation of the Waterberg Group to test whether other related rock types could serve as an alternative should mining activities warrant future translocations (Harris et al. 2014).

Qualitative observations for pollinators were made at the receptor sites of Goedvertrouwdt and Eagle’s Rock, and a large, natural population at Ezemvelo Nature Reserve. As the flowers open only on sunny days, observations were made on non-overcast days of March 2016 and insects that visited the flowers were collected by hand. One observation session (of 3.5 h) was performed per day to coincide with the flowering time of the species for each population, on 3 suitable days (Table 1). Three 1-m² patches with most flowers were visually estimated and demarcated at each site for observation. One person per patch observed for pollinators for the entire period of anthesis from 11:30 to 15:00. Thus 10.5 h of observation was conducted per population. Insects alighting or foraging on flowers were captured using an insect net and those species observed for the first time were euthanised in a killing jar with ethyl acetate. These specimens and a set of duplicates were then stored in individually marked containers for later identification and pollen assessments. Repeat visitors of the same species were only counted and released.

All euthanised specimens were photographed with a Nikon AZ1000 stereomicroscope to assist in identification and scanned for pollen grains. Particular attention was paid to the head and legs, where pollen was most likely to be found. Those specimens carrying pollen were sputter-coated with a thin gold-palladium layer and inspected under an FEI Quanta FEG 250 scanning electron microscope (SEM) and micrographs of relevant pollen loads were taken. Duplicate insect specimens were pinned and labelled as prescribed by Uys and Urban (2006) and submitted to the Biosystematics Division of the Plant Protection Research Institute of the Agricultural Research Council (ARC) for identification (Appendix 1). Identification to species level was hampered by the taxonomic impediment whereby large numbers of species still require classification or have gone undiscovered (Eardley, Gikungu & Schwarz 2009). Verification of F. humilis pollen was based on pollen micrographs in Burgoyne et al. (2000).

In each of the three subplots per population, all the adult plants and flowers were counted according to the technique described by Harris et al. (2014). The fruits were counted per subplot 30 days later. To estimate seed set per capsule in each subplot, 20 capsules were removed from plants bordering on the subplot (after 6 months) and the seeds were counted to determine the mean number of seeds per capsule. The sites were revisited after the first follow-up rains in October 2017 to count the seedlings per subplot.

Ethical clearance was obtained from the North-West University. This article followed all ethical standards for research without direct contact with human or animal subjects.

A total of 46 insects were captured whilst alighting on F. humilis flowers. These comprised nine different species of which six were hymenopterans and three were dipterans (Table 2). No lepidopterans were captured during this survey although species were observed making darting visits to Frithia flowers. Harris et al. (2016) did record two butterfly species alighting on Frithia flowers but these were void of pollen.

At Goedvertrouwdt, 14 individuals belonging to six species were recorded, with Exoprosopa eluta (Loew) (Hymenoptera: Bombyliidae) being the most frequent visitor (five records) of Frithia flowers (Table 2). Species of Lipotriches (Gerstaecker) (Hymenoptera: Bombyliidae) and Quartinia (André) (Hymenoptera: Vespidae) were recorded the most at Eagle’s Rock Private Estate (Table 2). At Ezemvelo Nature Reserve, 26 individuals (four species) were captured, with Amegilla fallax (Smith) (Hymenoptera: Apidae) and E. eluta accounting for 65% of the visitations (Table 2).

Harris et al. (2016) already provided evidence that the bees Amegilla fallax and Megachile niveofasciata (Friese) (Hymenoptera: Megachilidae), together with the bee fly E. eluta, carry F. humilis pollen. Scanning electron microscope evidence reported here was, therefore, specifically aimed at the five newly captured flower visitors and also on recaptured Lipotriches sp. previously identified as a potential pollen carrier by Harris et al. (2016).

Frithia pollen was found in varying amounts on the selected insect specimens and these species are therefore considered pollen carriers (Table 2). Some pollen grains were visible on Notolomatia sp. (Figure 4), on and immediately adjacent to the mouth parts. Although only three pollen grains were found on the abdomen of Paragus sp. (Figure 5), a significant clump of Frithia pollen was found directly on the anus of the specimen. Pollen found on the Quartinia sp. (Figure 6) was sparsely distributed over its entire body. Copious amounts of pollen were found beneath the mouth parts of the Ammophila sp. (Figure 7). Halictidae specimens (Seladonea sp. and Lipotriches sp.) (Figures 8 and 9) were carrying copious amounts of Frithia pollen, particularly in their pollen baskets.

The two translocation sites and the control populations were situated in an area of approximately 200 km². Within this area, three bee species (A. fallax, Lipotriches sp. and M. niveofasciata) were recorded from at least two of the sites (Figure 10) and a bee fly species, E. eluta, was recorded from all three sites (Table 2). The 39 flower visits by these four species accounted for 85% of all recorded visitations. This indicates that the pollination system of F. humilis is dominated by generalist bee pollinators that carry medium to high loads of pollen. Although the bee fly was a more regular flower visitor (14 visits), its pollen load was low (Table 2). Reserve pollinators were observed at Goedvertrouwdt and Eagle Rock Private Estate (Figure 10). These pollinators were mostly flies and wasps and generally had medium to low pollen loads. A sweat bee (Seladonea sp.) was only recorded from Goedvertrouwdt and had a high pollen load. These five reserve pollinators accounted for 15% of flower visitations.

When mature plants, flowers, fruits and seedlings were compared across sites per 1 m², Ezemvelo Nature Reserve had significantly higher numbers, as would be expected for a natural population (Table 3). However, no significant differences were recorded between the typical and atypical habitats. Plants from Ezemvelo Nature Reserve had the highest number of flowers as a percentage of total number of mature plants, with fruit set as a percentage of flowers the highest for the sites at Goedvertrouwdt and Ezemvelo Nature Reserve (Table 3). As this study was conducted 7 years after translocation, it may be assumed that the seed bank would make a minor contribution to recruitment and that recent seed production is a consequence of recent flowering seasons. Based on this premise, seedling emergence at Goedvertrouwdt receptor site as a percentage of seed numbers was considerably higher than that of the control population and is indicative of viable seed (Table 3). Considering the 7-year time span since translocation, the presence of six pollinators (Figure 10) and seedlings, it can be deduced that effective pollination is taking place.

Exoprosopa and Notolomatia of the Bombyliidae are primarily Afrotropical in occurrence (Greathead & Evenhuis 2001). Larvae are either predators or parasitoids, whilst adult flies in most genera are adapted for nectar feeding. Female flies also feed on pollen for reproductive purposes and many of them have special adaptations for the collection of pollen (Greathead et al. 2006). Bombyliidae are most active during sunny middays that coincide with the anthesis of F. humilis. Little is known about their role as pollinators; it is generally believed that most of them are generalists (Szucsich & Krenn 2002) although some specialised relationships are known (Johnson & Steiner 1997). The small number of pollen grains on the Notolomatia specimen could be a consequence of the species or individuals (the inherent interested in pollen), the gender of the specimen (male flies may not be as interested in pollen, if at all) and the fact that the specimen was captured as soon as it landed on the flower, thereby shortening its potential visitation time.

Larvae of the hover fly genus Paragus (Diptera: Syrphidae) are recognised as aphid predators (Hayat & Claussen 1997), whilst adult flies are nectar feeders. Female flies consume pollen because it is a rich protein source necessary for reproduction (Haslett 1989). Even if pollen is consumed and digested by all syrphids, it still retains its shape after passing through the gut, making identification possible (Holloway 1976). This explains the large cluster of pollen grains from F. humilis and other plant species on the anus of the specimen. The specimen only foraged for a short period before it was captured. Although the pollen load was low on the insect itself, the pollen cluster on the anus suggested that it had recently fed on F. humilis pollen.

The Halictidae, represented by the species Lipotriches and Seladonea in this study, is the second largest family of bees with over 3500 members. Many of them are pollen specialists with distinctive adaptations for collecting nectar and/or pollen from a small number of closely related plant species (Danforth et al. 2008). In southern Africa, these non-apis bees visit numerous plant families but the most predominant is the Aizoaceae, along with Asteraceae, Fabaceae and Zygophyllaceae (Gess & Gess 2004a). Gess and Gess (2004a) reported that as many as 13 different plant families may be visited by a single species but visitation records pointed to preferences for certain families. Seladonea sp., for example, seemed to prefer Asteraceae compared with other families. However, when considering the species richness and diversity of the Halictidae, along with their polylectic manner of feeding, it can be concluded with a reasonable degree of certainty that they are important pollinators for many seed-bearing plants (Dikmen 2007). All of the collected Halictidae specimens displayed a typical pollen collection behaviour, which explains the copious amounts of pollen observed on their bodies and in their pollen baskets.

Ammophila is a genus of thread-waisted wasps in the Sphecidae well known to be parasitic and predatory in nature. Preferences for specific flowers are known amongst the genera of Sphecidae but such relationships are not as strong as in bees or masarids (Bohart & Menke 1976). Spechids, in general, feed on a variety of food including nectar, honeydew and bodily fluids of their prey. Genera with short tongues seek nectar from flowers with short corollas such as the Apiaceae, Asteraceae, Euphorbiaceae and Polygonaceae (Gess & Gess 1991). Ammophila has elongated mouthparts that enable it to feed on a greater variety of flowers. Flowers known to be visited by Ammophila include Acanthaceae, Apiaceae and Lamiaceae (Gess & Gess 1991; Herrera 1989; Weaving 1989). The captured Ammophila specimen was observed to make precise efforts to probe the nectaries of the F. humilis flower by circumnavigating the flower in such a way to ensure contact between the anthers and the underside of its mouth parts where pollen grains were found.

All species of pollen wasps (Vespidae), including those belonging to the genus Quartinia, are nectar feeders and pollen collectors (Gess & Gess 2010). Of the 40 species recorded to visit flowers, 75% preferred the Aizoaceae. Quartinia is also generally regarded as an effective pollinator for many Aizoaceae species and numerous specialised relationships have been noted (Gess & Gess 2004b, 2010). During the survey, one of the Quartinia specimens continued to forage in a F. humilis flower even after the hand net was placed over it and the flower. Even though little pollen was found on the specimen, the fact that all species of Quartinia are pollen feeders and collectors indicate that this genus may be an efficient reserve pollinator of F. humilis (Gess & Gess 2010).

Although few specimens (46 individuals) of pollen carriers were captured during this study, our observations provided enough evidence to support the hypothesis presented by Hartmann (1991) that F. humilis belongs to the Melittophilous syndrome, thus being pollinated mostly by bee and bee-like pollinators. This was especially the case for the natural F. humilis population at Ezemvelo Nature Reserve, where bee pollinators were predominant. At translocated sites, additional observations were made of flies and the Ammophila wasps which support the presence of Anemophilous syndrome and imply a bimodal pollination system. These ‘reserve’ pollinators (Johnson, Harris & Procheş 2009) associated with the Anemophilous syndrome provide plausible alternatives for the translocated populations.

However, it must be cautioned that even though specimens might carry F. humilis pollen, it does not conclusively prove that any of these species are pollinators. In addition, the amount of pollen on a pollinator’s body is regarded as an unreliable determinant of efficiency (Adler & Irwin 2006). Identifying a primary pollinator depends on the pollination efficiency of the species, which is defined as the amount of con-specific pollen transferred to a stigma in any one visit (Inouye et al. 1994). However, determining pollination efficiency is problematic because pollination is a highly variable ecological interaction and can be influenced by factors such as pollinator abundance across years, flower visit duration and frequency, contact with flower parts as well as pollen removal and deposition (Ivey, Martinez & Wyatt 2003).

With this in mind, the bees Amegilla fallax, Megachile niveofasciata and Lipotriches sp. that form part of the Melittophilous syndrome are cautiously proposed as primary pollinators of F. humilis because of their higher visitation frequencies and larger pollen loads. The sweat bee, Seladonea sp., could also be a primary pollinator because of its higher observed pollen loads, despite only being observed twice. Whilst pollen load may not be a direct indication of pollination efficiency, the larger pollen loads of bees compared with other F. humilis visitors were particularly evident on SEM micrographs. The number of observations recorded as visitation frequency is also in favour of bees compared with the various other species. One generalist pollinator, E. eluta, was observed at all sites but is likely an inefficient pollinator because of the limited amount of pollen found only on the eyes (Harris et al. 2016).

Further evidence for bee pollination is supported by the pollination system of a closely related species, Bergeranthus multiceps (Aizoaceae), from spring and autumn rainfall areas of the Eastern Cape Province, South Africa. This plant is also self-incompatible and flowers open between 15:30 and 18:30. Bees were determined to be the most important pollinators to this species, placing it in the Melittophilous syndrome (Peter et al. 2004). Other potential pollinators included butterflies, bees and hover flies belonging to the families Bombyliidae, Tachinidae and Syrphidae (Peter et al. 2004). This study and Harris et al. (2016) made similar observations for F. humilis.

Pollination syndromes may maximise successful pollination by a specific species whilst still reserving the option for chance pollination by a less suitable species. The non-Melittophilous Frithia pollen carriers are suggested to be reserve pollinators because of their frequent visits and exclusive occurrence in translocated populations. They are primarily wasp and fly species that form part of an Anemophilous syndrome. Whilst bees may be primary pollinators, the small size and isolation of F. humilis populations supports the idea that F. humilis is a generalist rather than a specialist, and it can rely on reserve pollinators as its flower structure allows pollination by a different syndrome. A bimodal pollination system has also been identified previously for a member of the Ruschioideae in the Aizoaceae (Zietsman 2013).

Pollination was equally successful at the natural and translocated Goedvertrouwdt populations although flower density differed (Harris et al. 2014). This is in accordance with the flower density hypothesis (Pufal et al. 2008). Flower density is known to have a strong influence on pollinator visits (Kunin 1997) but the size of the population had little effect on pollination efficacy. In other words, more flowers do not imply higher proportional frequencies of pollination.

Seedling recruitment per seed quantity was the highest for Goedvertrouwdt, but this could be ascribed to lower plant densities and more suitable niches (Eriksson & Ehrlen 1992). Eagle’s Rock has a different geology and as expected did not have ideal conditions for seedling germination and establishment (Harris et al. 2014). The presence of seedlings, therefore, is no indication of pollination efficacy but only that successful pollination takes place. This indicates that at least some of the pollen carriers recorded in this study are also pollinators.

This study showed that plant populations translocated to atypical habitats and geology had lower levels of fruit formation despite having higher levels of flowering compared to translocated populations of typical habitat. Site selection is important and should be performed carefully to avoid plant stressors and nutrient deficiencies which may affect plant and flower health, and the availability of pollen (Mayer 2004). Reduced flowering can affect the number of pollinators that may visit a population and might limit outcrossing in small populations resulting in pollen transfer between relatives or not being pollinated at all (Kearns & Inouye 1997).

The sparsely populated translocated sites had higher levels of fruit set and seedling occurrence per individual than the natural population. This shows that it is important to consider transplant density for xenogamous species, as too high densities influence the available pollinators and could result in many unpollinated flowers, which could restrict geneflow in the newly established population. Pufal et al. (2008) suggested that decreased density of flowers increases pollination success and fruit set per plant as pollinators are known to be scarce, with a few taxa accounting for most of the flowers at a site (Herrera 1989).

Translocation requires relocating a species to areas within its existing distribution range (IUCN 2013). This study has shown that generalist pollinators of F. humilis occur throughout its natural distribution range and are common and mobile enough to locate newly established populations. Translocation is therefore a low risk conservation option in terms of generalist plant species. Pollination is further enhanced by the bimodal pollination system that is characteristic for some of members of the Aizoaceae (Zietsman 2013), which makes them resilient to shifts within their distribution range.