Flowering phenology in the arid winter rainfall region of southern Africa

The impact of physical factors on the flowering phenology of a succulent karroid community in the winter rainfall region of the northwestern Cape, South Africa, based upon a three year study on permanent plots, is examined, (in the permanent plots, flowering of the shrubby species extended over a period of 4 to 4'/i> months each year, while blooming ot the therophytes peaked m the first half of the flowering season. Species composition and numbers of individuals in the therophytes and geophytes offering flowers varied greatly according to the pattern and amount of seasonal precipitation. Despite these variations a consistent flowering sequence between the years was observed. Possible relations between the flowering phenology and the climatic variables are discussed in detail. The present data suggest that the onset of flowering is determined indirectly by the first drop in temperature in autumn, indicating the beginning of the rainy season and presumably the start of the growing period, and/or by the increase of temperatures in the beginning of spring. The pattern and amount of rainfall within a given season mainly influenced the duration of anthesis and the number of flowers produced.


INTRODUCTION
A short hut copious blooming is a characteristic ieature of many arid and semi-arid regions.This is particularly true of N amaqualand in the northwestern Cape, South Af rica.which is renow ned for its flower displays during springtime.
For most plant species the growing season is restricted to the cool and moist autumn-winter-spring period and the dormant season occurs during the hot and dry summer months.The growing period usually commences in au tumn (March to April) with the sprouting of the perennials and the germ ination o f the therophytes (Van Rooyen et al. 1979;Le Roux et al. 1989).The flowering season lasts from late w inter to spring (August to early October).Dur ing this period 90% of the shrubby perennials and virtually all the annuals are blooming (Le Roux et al. I9S9).In comparison to other regions of winter rainfall climate an thesis is markedly synchronized (Orshan et al. 1989).
The overall influence o f the climatic factors on the phenological events in arid regions is obvious.With respect to the phenology o f flowering, special attention has been given to the timing and synchrony o f blooming in relation * Botany Department, University of Cape Town, Private Bag.Ronde-bosch7700.
to the influence of climatic variables (e.g.Baker et al. 1982;Beatley 1974;Noy-M eir 1973;Solbrig & Yang 1977).Generally, the importance of an opportunistic re sponse by the plants to w ater availability has been stressed.Compared to other arid areas receiving a similar amount of average rainfall, the predictability o f seasonal precipitation is relatively high in the arid winter rain fall region of the Cape Province (Hoffman & Cowling 1987).Hence, the phenology o f the vegetation should be comparatively predictable, too.However, the question of the impact o f current seasonal weather conditions on the flowering phenology has not yet been addressed ade quately.
Against this background, results of a three year study of the flowering phenology of a succulent karroid com munity of the northwestern Cape are presented.The fol lowing questions shall be addressed in particular: 1, how does the timing and vigour of flowering vary between the years?; and 2, can the phenological pattern be linked to the current rainfall and/or temperature conditions, thus indicating potential environmental cues?The present sur vey is clearly constrained by the limited number of plots and years studied, and the data do not provide a robust basis for a thorough statistical evaluation.Hence, the fol lowing observations should be regarded as indicating trends.

METHODS
The field work was carried out at the Goegap Nature Reserve during the blooming seasons of the years 1985 (late June to October), 1986 (July to late November), and 1987 (mid-August to mid-November).
The records of the actual temperatures and precipita tion at Goegap were provided by the authorities of Goegap Nature Reserve.The data were supplemented by the au thor with the aid of a thermohygrograph.Phenological data were collected for a total of 112 plant species on six permanent plots of 100 m2, each.Five plots were laid out on various sites on a mountain slope with undisturbed dwarf-scrub vegetation so as to cover some of the conspicuous variation in species composition.One plot was marked on a sandy plain with exclusively ephemeral vegetation showing a particularly high degree of variation in yearly species composition and plant cover.The permanent plots are characterized in an Appendix.
For each species onset and duration of anthesis were recorded.Furthermore, numbers of open flowers or in florescences with open flowers were repeatedly counted at time intervals of three to five (rarely up to seven) days.Inflorescences and partial inflorescences representing the functional unit of visitor attraction or pollination (i.e.blos soms sensu Faegri & Van der Pijl 1979) were taken as equivalent to single flowers.

Description o f the study site
The Goegap Nature Reserve is situated 12 km east of Springbok (between S 29°34 ' and 29°41' and E 17°57' and 18°02') in the Namaqualand Rocky Hills in the north western Cape, South Africa.
The area is characterized by an open dwarf-scrub vege tation which has been classified as 'Namaqualand Broken Veld' by Acocks (1988) and as part of the Succulent Karoo Biome by Rutherford & Westfall (1986).Succulent and nar row-leaved sclerophyllous species abound.The flora is clearly dominated by Asteraceae and Mesembryanthemaceae (Le Roux 1984).
The study area is situated in the realm of the winter rainfall region.About 70% of the yearly precipitation is received from April to October and about 38% during the wettest quarter of the year (June to August) (Table 1).Drought conditions prevail virtually throughout the year.All climatic factors are subject to conspicuous annual fluc tuations.

Weather conditions during the study
The amount of precipitation as well as the timing and number of rain events varied considerably between the study years (Figure 1; Tables 1 & 2).Annual rainfall was highest in 1985 (244.5 mm, i.e. 73% above the average for the years 1974 to 1987), but much lower in both of the subsequent years (110.5 mm in 1986, i.e. 22% below the average and 113.5 mm in 1987, i.e. 20 % below the average).However, rainfall figures for the main growth period (April to October) and wettest quarter of a year (June to August) provide quite a different picture (Tables 1 & 2).During the main growth period, 87 mm was re ceived in 1985 (35.6% of the year's total), 81.5 mm in 1986 (73.7% of the year's total) and 113.5 mm in 1987 (100% of the year's total).The maximum number of rainy days was recorded in 1986, especially between June and mid-September.The relatively high rainfall of 1987 was due to several exceptionally good rains during April, June, and August.In contrast, in 1985 drought conditions pre vailed in June and during August to September with only few intermediate rains.
In comparison to the rainfall pattern the annual course of air temperatures was rather uniform (Figure 1).In each year, winter minimum values were reached during a pe- yearly riod of about five m onths, after which tem peratures in-the cool season appeared to shift from year to year.In creased again during a period o f one to two months to 1985 the mean monthly temperatures dropped considcrrcach sum m er values (Table 3).Nevertheless, the onset o f ably during March, but only during April in both of the Bothalia 24,1 (1994)

Numbers o f species in flow er
In total, 112 species were observed in flower on the permanent plots during the present study.O f these, 44% flowered in 1985, 95% in 1986, and 79% in 1987.These variations were mainly attributable to the phenological performance of therophytes and geophytes (Table 4).
The number of therophytic species flowering in a given year seems to be related to the amount and distribution of rainfall occurring during the two months before the onset of the flowering season.In 1985 only 25.5 mm fell during this period, whereas in both of the subsequent years the figure was nearly twice as high (1986: 46 mm, 1987: 50.5 mm; Table 2; Figure 1).
In 1985, the flowering was also affected by unfavour able weather conditions during the flowering period itself, as shown in Figure 2. In that year, a dry period set in after the initial rains at the onset of the blooming season (Figure 1 and Table 2) causing drought stress particularly for the therophytes.A good rain in mid-September, how ever, stimulated active growth and flowering.
Very few plant species growing on the permanent plots flowered during summer.These are Othonna furcata (March and April), Tylecodon wallichii (end of November to beginning of December), the therophyte Tribulus terrestris (April), and the hysteranthous geophyte Eriospermum paradoxum (March and April; all dates after Van Rooyen et al. 1979).

Timing and sequence o f flowering
In each year, the blooming season lasted 4 to 4'/i> months.However, the exact timing varied markedly be tween the years.In 1985.most plants flowered from early July until the end of October, whereas in both of the sub sequent years blooming commenced about four weeks later, in early August, and continued until the end of No vember (Figure 1).(Heinrich 1976;Heithaus 1974;Hocking 1968;Kratochwil 1984;Mooney et al. 1974;Pierce 1984;Reader 1975;Stiles 1977).Considerable yearly variations in flowering dates, but unaltered flowering sequences have also been reported from southwestern Spain (Arroyo 1990).In Goegab the shrubby species make up the "backbone' o f the sequence, while therophytes and geophytes close the ranks in sea sons when their specific requirements in terms o f moisture and temperature conditions are met.This basic pattern re mained unchanged, though the vigour o f blooming varied greatly between years (see below).Taxonomically related species usually flowered sequentially, with their flowering periods slightly overlapping (e.g.Euphorbia, Hemumnia, Ruschia spp.; for further details see Struck 1992).
In the present observations no recurring pattern be tween the timing and/or amount of precipitation and the timing o f flowering in single species or the timing of the bkxmiing season in general were detected, though the vig our of flowering appeared to be stimulated by single rain fall events in single cases (see below).
Moreover, the present data do not indicate that flow ering of the therophytes was stimulated by drought.This is in contrast to a view widely accepted for desert annuals (Fox 1990a, b;Rathcke & Lacey 1985).Yet, recent ex perimental and field demographic studies (Aronson et al. 1992;Fox 1989Fox . 1990b;;Van R<x)yen et al. 1991) indicate that water stress had little or no effect on the induction of flowering.
Conversely, a correlation was found between the au tumnal drop in temperatures and the onset of the blooming season: as mentioned above, mean monthly temperatures dropped considerably during March in 1985, but only dur ing April in both o f the subsequent years (Table 3).The bkxmiing season started about three months later in each year, namely in early July in 1985 and about early August in both of the subsequent years.It should be noted that the amount and distribution of rainfall during autumn dif fered notably between the years (Figure I).The autumnal drop in temperature corresponds to the beginning o f the winter rainy season and presumably marks the onset of the grow ing period for (he plants (Van Rooyen et al. 1979).
With regard to the weekly temperatures (or mean tem peratures of 7-day-intervals not corresponding to calendar weeks) during autumn (Figure 1), the first cool interval occurred in mid-March in 1985, during the second week of April in 1986, and about the end o f April in 1987 (with 5.0, 5.9, and 3.1°C below the means of the respective months).These temperature events shifted for about 25 days between 1985 and 1986 and for about 10 days be tween 1986 and 1987.This situation is reflected in the yearly shift of flow ering dates of selected perennial species (shift 1985/86: about 3 weeks, 1986/87: 4 to 7 days, see Table 5).Though the data show a high degree of variation, the differences appeared to be significant.In contrast, the average shift in flowering dates for selected annuals was 14 and 7 days, respectively.Yet, due to the generally low proportion of annuals in the plots, particularly during drier blooming seasons, the results remain ambiguous (Table 5).However, the lesser yearly shift of flowering dates of annuals may also indicate other influences, e.g. a delay of flowering by low tem peratures during the w inter period (Van Rooyen et al. 1979).Low temperatures have been shown to lengthen the time between flower initiation and anthesis in therophytes (Van Rooyen et al. 1991).One could expect such an effect for 1985 and 1987 when the cool winter period overlapped with the onset of the bkxmiing period, i.e. when most annuals started to flower.Given this, the differences in flowering dates should be less between 1985/86, but more between 1986/87 compared to the per ennials.This is what Table 5 indicates.However, the pos sibility that this is merely coincidence, cannot be ruled out.
In some species flowering dates differed conspicuously in various places of the study area (e.g. up to a month in Ruschia robusta).Similar observations have also been re ported from other arid regions (e.g.Turner & Randall 1987 and references cited therein).W hether these differ ences reflect small-scale variations in soil conditions or intra-specific genetic variability, is not known.Some o f the species under present study were reported to flower one to two months earlier (in 1974, Van Rooyen et al. 1979) or later (in 1981, Le Roux et al. 1989) than observed during the present survey.Nevertheless, the se quence of flowering of the species involved generally matches the present results.Unfortunately, a further com parison is not possible, because the authors collected phenological data in a notably larger area, notwithstanding the small-scale variations o f flowering dates.Besides, rel evant weather data are lacking for the above-m entioned years.

Numbers o f flo w ers and duration o f flow erin g
The great majority o f the observed species produce large num bers o f flowers over a tim e span o f several weeks ( 'cornucopia', Gentry 1974).In these species the maximum num bers of flowers (or inflorescences, if re garded as the functional unit o f visitor attraction or pol lination) offered sim u ltan eo u sly , v aried from a few thousand to tens o f thousands (up to 40 (XX) per 100 m2 in Galenia sancophylla |A izoaceae) on plot 6 in 1986).
Beside the seasonal moisture availability, single rainfall events appeared to stimulate the vigour of flowering in certain cases.An example is the good rain in early Sep tember 1985, which ended a short drought period (Figure 1).After that shower, several therophytes and chamaephytes produced a further minor bloom (Figure 3).Given a causal relationship between these incidents, the species did, nevertheless, not respond to similar rain events (in terms of amount and timing of precipitation) in the sub sequent years.
Furtherm ore, 'small rainfall events' (sensu S ala & Lauenroth 1982) seemed to stimulate flowering in thero phytes: several species in plot 6 (e.g.Senecio arenarius, Leysera tenella) showed a slight, though remarkable in crease in the numbers of flowers after sparse rains (below 5 mm) occurring during September and the beginning of October 1986 and 1987.This is in contrast to the findings of various authors (for references see Sala & Lauenroth 1982) who only regard rain events of at least 8-10 mm ( 'effective rainfall event' sensu Noy-Meir 1973) as eco logically significant.
Conversely, all observed geophytes as well as some chamaephytes and therophytes produced but a few flowers per day over a period o f several weeks or more ( 'steady state'.G entry 1974).O f these the perennial clim ber Microlonui sagittaturn (Asclcpiadaceae) showed the long est flowering period (120 days).A transitional phenological pattern was shown by some species which produced moderate numbers o f flowers over an extensive period (e.g. the succulent shrublet Crassula muscosa subsp.muscosa flowering more than 70 days, the therophytic Osteospermum hyoseroides and O. amplectens flowering for 95 days).
As already m entioned, bloom ing o f the therophytes was generally rather poor in 1985.but was markedly better in 1986 and 1987.The shrubby species responded rather differently to seasonal m oisture availability: generally, non-succulent shrubs showed the highest production of flowers in the moist years 1986 (most Asteraccac) and 1987 (Lebeckia sericea (Fabaceaej), whereas succulents (most M csem bryanthcm aceae, Euphorbia spp.) flowered most prolifically in the com paratively dry season of 1985.
The duration of flowering of most taxa ranged from 10 to 80 days.Only the geophytes showed short flowering periods of 20 to 30 days in all three years.Due to the adverse climatic conditions during the growth period of 1985, most therophytes had a comparatively short flow ering period (and low flower numbers) in that year.On the other hand, duration of flowering of most therophytes was clearly longer in 1987 than in 1986 (the year with the maximum number of flowering therophytic species, see Table 4).
No specific pattern o f skew ness in the flow ering phenologies could be detected: flow ering may begin abruptly and then tail off in one year but may start slowly and end more quickly in another year.In species showing the 'cornucopia' type of flowering pattern the duration of full bloom (defined as the time interval with at least 50% of a species' maximum number of flowers in anthesis) was longest in the year with the highest numbers of flow ers and shortest in the year with the lowest numbers of flowers, i.e. a lower flower production was not com pen sated for by an extended duration o f flowering.
A high degree of phenological variation on the species level, and to a lesser extent between populations, is an outstanding feature of the plant community studied.The present observations revealed that the timing of flowering and the vigour of blooming varied independently within the species under study.Conversely, the flowering se quence of given species remained largely unaltered during the years.
Regarding the possible impact of climatic variables, it is inferred that the timing of flowering is indirectly deter mined by the first considerable drop in air temperatures in autumn and/or by the increase of temperatures in the beginning of spring.In contrast, species composition and number of individuals of therophytes and geophytes in flower are greatly affected by the pattern and amount of seasonal precipitation.The same applies to the number of blossoms produced.Due to fluctuations in seasonal rain fall the vigour of a species' blooming may be promoted in one year and retarded in another.Yet, potential pheno logical responses seem also to be related to morphological and physiological constrains: e.g.non-succulent shrubs produced the greatest number of flowers in moister years, whereas succulents flowered most prolifically in a dry year.

APPKNDIX
Habitat, spccies composition, percentage of plant cover and flowering phenology of the permanent plots.Where species flowered twice during a (lowering season, number of days of flowering (fourth column) are given separated by '+ '. '+ ' put before the number ol days indicates that flowering started before an observation period, *+' put behind means that flowering continued after an observation period.Floral units counted (sixth column) were single flowers unless stated otherwise.Names of plant communities follow Le Roux (1984).

+68
monthly temperature; x (max.), mean maxima; x (min.), mean minima; abs.max., absolute maxima; abs.min.. absolute minima; d. differ ences between means of consecutive months.TABU-! 4.-Number of flowering species observed on the Despite the temporal fluctuation in the onset of the flowering season the sequence o f flowering o f the species involved rem ained nearly constant (see flowering se quence for plot I in F igure 3), i.e. independent o f the current pattern o f precipitation in a given year.Such con sistent sequences o f flowering phenologies have been de scribed from areas o f virtually all climatic zones I K U Kl, 2. Percentage of species flowering simultaneously (cumula tive for all plots, I (K)c/r = 112 species).Solid line = 1985; stippled line = 1986; dotted line = 1987.

TABLE 1 .
-Rainfall (mm) in Goegab NatureReserve 1974-1987, calculated from data from Le Roux (1984)and unpublished weather data of the Reserve

TABLE 2 .
-Monthly rainfall and number of rainy days (in brackets) dur ing the main growth period (April to October) 1985 to 1987

TABLE 3 -
Temperature regime in Goegab during the study period 1985 to 1987.Data from unpublished weather report Goegab Nature Reserve and own measurements by thermohygrograph