Flowering Phenology and Diversity of Dicots in Desert-Shrub

Grassland, Southeastern Arizona, 1987-1999.

ROBERT M. CHEW

POBOX 16306, PORTAL, AZ85632

© 2004 R.M. Chew

 

 

        Abstract. _____ The timing and magnitude of flowering of a suite of 32 dicot annuals and perennials in a degraded grassland invaded by desert shrubs was recorded over 13 years (1987-1999).  Flowering was related to weekly rainfall and sum of degree days in six phenological seasons: winter (germination of annuals, dormancy of perennials), early spring (vegetative growth and limited flowering), late spring (major spring flowering), early summer (no, or limited flowering), late summer (major summer flowering), fall (little or no flowering, senescence and dormancy).  Each season varied relative to itself through the years from extremely wet to extreme drought and extremely cold to extremely warm.  Each year was unique in its pattern of six seasons. Most minimum and maximum values occurred 1994-1999.

        Fourteen species flowered only in spring, 4 only in summer, 13 spring and summer, and 1 in fall. Flowering periods overlapped, but the times of maximum bloom were often significantly separated. The duration of a species’ flowering was a function of the time of its first bloom.  Durations peaked in late spring and then declined; durations progressively declined in summer.
         Regression models were constructed for each species total flowering (Nt) in relation to the independent variables of each seasonal rainfall and temperature sum, in the current year and previous year.
Total flowerings of 18 of the 32 species were significantly predictable from rainfall and temperature, 6 for both their spring and summer flowerings. The number of significant seasonal variables in a predictive regression ranged from 1 to 5 with a median of 3. The first two variables to enter a model accounted for an average of 88% of the variability of Nt (range 56-100%).The whole models had an average adjusted R2 of 0.93.  No two species had the same first two predictive variables. For the two predominant predictive variables, 24 of 45 are for the current year and 21/45 for the previous year; Precipitation is a dominant variable in 31/45 and temperature sum 14/45.

        At least 15 different phenological patterns of flowering can be proposed for the 32 plant species, based on responses such as specific time of flowering, dependence on rain in a particular season, in the present or previous year, biseasonal flowering, seasonal flexibility, and the nature of the species: annual, type of perennial, evergreen, deciduous succulent, and time span between vegetative growth and flowering. In years when there was prolonged low moisture index, based on low precipitation and / or high temperature sum, there was not only failure of flowering, but senescence and even death of large proportions of some species populations. There were some hierarchies of recovery of species.

        Considering what is now known about the genetics of flowering in Arabidopsis, it is reasonable to speculate that the biodiversity of the dicots in this semiarid community is a result of the interaction of the great variety of rainfall and temperature conditions, season to season and year to year, with the many phenotypes of the available plants, as they have assembled into the community in the short term and further adapted in the long term.

           

INTRODUCTION

 

        The annual sequence of plant phases from germination through vegetative growth, flowering, dispersal of seeds and senescence is one of the most striking aspects of nature, and it entices one into the observation of nature.  The utility of phenological research has expanded until it is essential for understanding the functioning of ecosystems (Leith, 1974); it provides paradigms of evolution (Lyons and Fox,1996) and gives evidence for global warming (Menzel and Fabian, 1999).  Here I extend the empirical study of flowering phenology in a semiarid site in relation to precipitation and air temperature over a 13 yr period.

        Temperature, moisture and photoperiod are major clues determining blooming, but there is little documentation for individual species.  Much effort has gone into trying to relate adaptation of flowering times to different climatic patterns (Fox, 1989), and to pollinators, predators, pathogens, competitors and nutrients (Rathcke and Lacey, 1985).  As the latter authors state (p. 191), “In general there are many more hypotheses about ultimate factors that may mold flowering times than there are thorough studies that permit testing them”.  The detailed studies of single species by Fox (1989) and LeBuhn (1998) illustrate the difficulties of determining even the proximate factors.

        The study of perennials and annuals requires different perspectives. Perennials have a flexibility of development due to stored resources in persisting plant parts and can remain vegetative or flower.  Each year annuals have the complexities of germination, survival of seedlings, vegetative growth, flowering, setting seeds, and dispersal of seeds to seed banks in the soil.  Whereas annuals in semiarid sites usually germinate after light but extended rains, which occur in most years, perennial shrubs may germinate only after short but heavy summer rains, followed by sufficient subsequent rain, a pattern that occurs less frequently (Went, 1949). There are a number of studies of flowering of annuals:  in the ChihuahuanDesert (Kemp, 1983; Gutierrez and Whit-ford, 1987; Whitford and Gutierrez, 1989; Samson et al,1992; Philippi, 1993 a, b; Guo and Brown, 1996); in the Mojave Desert (Beatley, 1974); and in the SonoranDesert (Went, 1949; Inouye et al, 1980; Inouye, 1989; Pake and Venable, 1995, 1996). Little work has been done on flowering of perennial species, especially shrubs. The major examples are the detailed studies in Nevada by Ackerman and Bamberg (1974), Beatley (1974), Ackerman et al (1980). Turner and Randall (1987) modelled the 9 years of data of Ackerman et al for first dates of leafing, flowering and fruiting.  Kemp (1983) gives observational data for Chihuahuan desert shrubs.

        Long-term data sets have led to some general conclusions: (1) flowering can be predicted from data on rainfall and air temperature, (2) often rainfall and temperature have opposite effects, (3) the most important events occur 1-4 mo or more before blooming, and (4) sometimes predictive formulae cannot explain failures of flowering due to isolated extreme conditions, e.g.  droughts and freezes, which can affect blooming for years afterwards (Turner and Randall, 1987; Fitter et al, 1995; Sparks and Carey, 1995). I will use a 13 yr data set to test these general conclusions. Here I document the flowering times and magnitude of blooming of annual and perennial dicots of a semiarid community with the objectives of: (1) defining the seasonal patterns of blooming for the species, (2) analyzing the patterns of flowering magnitude in relation to annual variation of rainfall and air temperature, and (3) assessing the differences of species’ phenologies as the basis for plant diversity.

 

METHODS

 

        SITE____ The study site is a 9.3 ha cattle exclosure that was established July1958 in the San Simon Valley, 8 km north of Portal, Cochise County, Arizona.  The site is 1.5 km east of the foothills of the ChiricahuaMountains, at 1370 m elevation on a gentle slope (1.8%) of alluvium largely derived from limestone and volcanic tuff. Soil depth averages 29.5 + 10.8 cm (SD), over hardened calcium carbonate (caliche); the soil surface is predominantly small gravel.  The vegetation is ChihuahuanDesert scrub, which probably replaced black grama (Bouteloua  eriopoda) desert grassland that had been degraded by drought, over-grazing and erosion.  The dominant plants are creosotebush (Larrea  tridentata) and two subshrubs Parthenium  incanum  and Zinnia  pumila . The main nonforage plants, favored by grazing, are fluff grass (Tridens  pulchellus) and snakeweed (Gutierreziasarothrae). [Plant names follow Kearney and Peebles, 1960].  The vegetation in 1958 is described in Chew and Chew (1965), and subsequent changes in herbaceous and subshrub species in Chew (1982).

 

        WEATHER____  Air temperature, 5 cm above ground, was recorded continuously, 1980-1999, from a thermal element inside a 76 cm long x 48 cm wide x 71 cm high shelter with shade screen sides, with double roofs 8 cm apart. Mean daily air temperatures were calculated from field records [(max + min) / 2] and summed over weekly periods asdegree-days above 0o C. 

        I used 0o C as the threshold for biological activity of all plant species, although there is almost no information on thresholds for the species of the site. The threshold for flowering may be10o for Larreaand some other shrubs in the SonoranDesert (Bowers and Dimmitt, 1994), but there may be vegetative activity at 0o that is basic to later flowering.  French and Sauer (1974) used a threshold of 4o in their study of phenology of grasslands.

        Daily rainfall, 1980-1999, was measured in a rain gauge with an internal cylinder graduated in 0.01 inch increments to hold 1 inch total as delivered from a 10.5 cm funnel . Overflow from the graduated cylinder went into a surrounding 10.5 cm cylinder with a capacity of 28 cm of rainfall.  In 1980 and 1984-1987 rainfall was also recorded from a tipping-bucket rain gauge.

 

        PHENOLOGICAL SEASONS____  For analysis of plant and weather data I divided the year into six phenological seasons.  These were based retrospectively on generalities of the data for plant growth and flowering over 13 years:

(1) winter, weeks 45-4  (calendar days 309-28), ca. November-January, a period of dormancy of perennials and germination of spring annuals;

(2) early spring, weeks 5-16 (days 29-112), ca. February-mid April, a period of vegetative growth, but before the start of blooming of most species;

(3) late spring, weeks 17-26 (days 113-182), ca. late April-June, the period of the major spring flowering;

(4) early summer, weeks 27-30 (days 183-210), ca. July, generally before the start of summer flowering and germination of summer annuals;

(5) late summer, weeks 31-39 (days 211-273), ca. August-September, the time of major summer flowering;

(6) fall, weeks 40-44 (days 274-308), ca. October, little or no flowering, senescence or entry into dormancy.

                                                                                               

        TRANSECT COUNTS____  I walked a standard rectangular path (121.8 m x 152.4 m) and counted the flowering individuals within 2 m of the center of the path ____  an area of 0.216 ha.  The number of transect counts varied from 14 a year to 22 a year, depending on the time of  first and last blooming of any species.  Counts were made every 7 to 14 days. I counted a plant as blooming regardless of its number of flowers, which reduces the information that would be obtained if the magnitude of blooming of individual plants were noted.  To do the latter would have limited the size of transect that could have been used and the time-span of the study.
        I originally planned to limit observations to 7 yr (1987-1993), but because of the occurrence of unusual weather conditions, I added censuses in summer 1994, spring 1995 and summer 1996. During continuing unusual years of 1997-1999 full, consistent counts could not be made to establish total counts (Nt) for phenological seasons, but there were enough counts to establish numbers for peak weeks (Np).

        Counts were compiled by weeks from wk 1 (Jan 1-7) to wk 52 (Dec 25-31). Numbers were interpolated for weeks when no count was made. The first day of blooming was assumed to be the first day of the week in which flowers were first observed, and the last day of blooming as the last day of the week when flowers were last observed.  The total number (Nt) for a phenological season involves counting some individuals more than once as flowers persist from week to week and as new flowers are produced, but the count is still a measure of the magnitude of flowering of a species.  Usually there was one week when the number of individuals flowering was considerably in excess of adjacent weeks (peak week, Np).

 
       
GERMINATION QUADRATS____In winter 1997 I established eight circular 1 m2  quadrats, two randomly positioned along each side of the rectangular walking transect.  These were used to follow the number of germinating plants, and their progression to fruiting. Each quadrat was divided into quarters. In two randomly chosen quarters germinating individuals were removed as counted; in the other quarters they were marked with toothpicks and left in place for further development.

        ANALYSIS OF COUNTS____  I analyzed the data for total seasonal counts of flowering individuals of a species (Nt) for their relationship to the independent variables of rainfall and air temperature sum of appropriate phenological seasons by forward multiple stepwise regression using the JMP Program, Version 3.1 (SAS Institute 1995).  The variables chosen by the stepwise procedure were made into a model by the least squares program associated with the stepwise regression. Information from the model-constructing program provided correlation values (R2)  forNtwith each variable entered, and an adjusted correlation (AR2) for the whole model of n variables.  A model was rejected if it did not have an acceptable normality of residuals, a constant variance of residuals, and acceptably low multiple collinearity, as indicated by the program.  A model was also rejected if there was so little variation among the Ntvalues of some years that they functioned as a “single” point in determining the regression plot, or if there were years that were so exceptionally high that they biased the plot.  Models were constructed for the Nt values versus rainfall and temperature of the current year, and for the current year plus the previous year.  Models were constructed for the years 1987-1993, and for these 7 yr plus data for spring 1995, or plus summers 1994 and 1996, depending on when a species flowered.

        The multiple stepwise regression procedure has obvious defects: (1) It can be confounded by correlations among independent variables. (2) Because of the number of variables allowed into the selection process when Ntis regressed against the cur-rent plus previous year, there is the possibility, but not necessity, of spurious correlations.  However, the procedure has the advantage that its predictions are “transparent” suggestions of the predictability of Nt in terms of rainfall and temperature, rather than the “less transparent” information from procedures such as principal components analysis and lagged correspondence analysis.  The multiple stepwise procedure is retained here for its heuristic value in suggesting hypotheses to be tested by experimentation.         

        Stepwise multiple regression has been used successfully in the analysis of other long-term data sets: (1) for shrubs in the Mojave Desert (Turner and Randall, 1987); (2) by Fitter et al. (1995), for data taken over 36 yr in one location in England, who were able to fit regressions of first flowering date to temperature for 219 of 243 species; (3) by Sparks and Carey (1995), who found significant relationships with mean annual temperature and annual rainfall for 13 plant species in a data set extending over two centuries; (4) by Epstein et al. (1997) who found relationships explaining 67-81% of the variation in production of grasses of the US Great Plains on the basis of mean annual temperature and rainfall; (5) by French and Sauer (1974) in a study of the phenophases of three species of grasses.

 

RESULTS

 

        RAINFALL____  Rainfall was biannual (Table 1 ).  Winter and early spring rains were gentle and widespread over a broad area and  provided 35.8% of the average annual total.  Late spring rains provided only 7.6%; early and late summer rains, predominantly thundershowers, provided 50.8% of the total, and fall (October) was a hiatus of lower rainfall, 5.8%.  Variation in regional and global weather patterns resulted in exceptionally wet and dry phenological seasons in some years.  The interannual coefficient of variation ranged from 0.53 for early summer to 0.91 for late spring (Table 1). Variability resides in the shorter time periods of the phenological seasons.  Over broader time periods there is less interannual variability, to the point where annual rainfall has a coefficient of only 0.30.  Winter-spring rainfall is significantly greater in Arizona and New Mexico during El Nino events (Andrade and Sellers, 1988; Diaz and Markgraf, 1992; Dahm and Moore, 1994).  Monthly rainfalls at the site, 1980-1999, were greater March-May during El Nino events, dramatically so during the “super El Nino” of 1998, when early spring rainfall was the maximum recorded, 243% of the average for 1987-1999 (Table 1).

        Although four of the five years of records from a tipping bucket rain gauge were before the present study began  (Table 2), they are assumed to be representative of the pattern for the site;  66.5 % of the rain events were less than 5 mm, but  provided 21.9% of the average annual rainfall.  Only 33 rains were > 20 mm, and those in July-September provided 31.4% of the annual rainfall, versus 21.5 % by such events in all other months. The extremes of the distribution are of special interest.  Although rains < 5 mm may not directly affect plant growth, Sala and Lauenroth (1982) concluded they are important because they affect decomposition and nutrient processing, which occur principally near the soil surface.  They found that rains of 5 mm did directly affect the physiology of Boutelouagracilis in semiarid grassland for at least 2 days.  Summer storms > 22 mm can be sufficiently intense (0.25-1.32  mm / min) to cause runoff (Hawkinson,  1968).  Fourteen storms of this intensity were recorded by the tipping bucket rain gauge in five yr.  Winter rains of even 70 mm were of too great a duration to cause runoff.  The light rains of winter are made more effective for plants by the lower evaporation then; the heavier rains of summer are made less effective by runoff.

 

        TEMPERATURE SUMS____  Degree day summations for phenological seasons (Table 3) are inherently less variable than rainfall, with coefficients of variation ranging from only 0.03 in early summer to 0.15 in winter.  Most of the variability was in the fall through early spring.  Although temperature sum is less variable than rainfall, a small change of temperature sum may be as biologically important as a large change in rainfall (e.g. Scifres and Brock, 1969).

 

        CATEGORIZATION OF SEASONS_____ To ease conceptualization, I categorized seasons with respect to rainfall as very wet (W*), wet (W), moist (m), normal (N),  dry (d), drought (D) and extreme drought (D*), and, with respect to temperature sum, as very hot (H*), hot (H) , warm (wm), normal (No), cool (c), cold (C), very cold (C*) (Table 4). The categorization was based on the percentile distributions of values for each season across the 13 years in Tables 1 and 3: W* and H* are values >90% quartile; W and H = 75-90%; m and wm = 62.5-75%, N and No = 62.5-37.5 %; d and c =25-37.5%; D* and C* <10% quartile. The categorization within a season is relative only to itself.  For example: extreme wet (W*) is 147-154 mm rain for the winter season, but 262-297 mm for late summer.  The biological effect of extreme drought and extreme wet will depend upon the season of its occurrence.  The categorizations of Table 4 make clear the season-to-season variation within years and the interannual variation within a season.  No combination is repeated in the 13 yr tabulated.

 

        THE PLANT COMMUNITY____  The 32 species that were censused (Table 5) are listed according to three categories of maximum densities of flowering individuals.  A few species that did not bloom in at least 4 yr in the period 1987-1993 are omitted. Except for Astragalus spp. and Lepidium  lasiocarpum the omitted species are annuals that were present erratically or in very small numbers, although some of these were exceptionally abundant in 1995 (see later). The community is predominantly Compositae (12 of the 24 species with densities >400 stems / ha) and perennials (21 of the 32 species).

        The censused dicots are predominantly perennials: 2 shrubs (Sh, Table 5), which were up to1.5 m high; 2 subshrubs (SSh) of < 0.5 m height; 4 suffrutescents (SP), still smaller and only slightly woody; and 11 herbaceous perennials (HP). Ten censused species were annuals (A).

        The times of blooming of species (Table 5) as established by censuses are: spring only (Sp), spring and summer (Sp/Su), and summer only (Su), with some minor cases (Sp* and Su*) when less than 5% of blooming individuals were counted in the other season. The dominance of spring flowering by annuals (Table 6) is almost significantly different from random (X2 = 5.33, 0.10> P > 0.05, df = 2). There was no difference for perennials. All the annuals listed in Table 5 germinate in winter, and then bloom in the spring, except Eriogonum Abertianum, which does not bloom until summer. There are few summer-germinating annuals on the site, none of which was abundant enough to be listed in Table 5.

                                                                                                                       

        SPECIES BLOOMING PERIODS ____   Spring flowering begins with  an herbaceous perennial with a long tap root, Cymopterus  multinervatus, and an annual, Draba  cuneifolia (Table 7). Thirteen other spring-only species (one Sp*) are scattered throughout the spring period, among 13 species that bloom both spring and summer.  The year ends with 3 perennials (one Su*) and one annual that bloom only in late summer and 1 shrub that blooms only in fall.

        The average beginnings of spring blooming progress from day 61 to 147. The average duration of blooming of species in spring (Fig. 1a) increases from day 60 to day105 and then declines to day 150. The second degree polynomial regression for this relationship is:

 

Y (days duration) = -92.934 + (2.72219 x ave 1st day) - (0.01348 x ave 1st day2), which has an F value with probability of 0.072.

 

The average duration of summer flowering progresses from day 196 to day 254 (Fig. 2a) with a significant decline from beginning to end. The 2nd degree polynomial regression is:                                                                                                                                   

Y  = 1592.06 - (12.9558 x ave 1st day) + (0.02703 x ave 1st day2). P > F = 0-.046.

 

The mean duration of blooming was not significantly different for annuals and perennials, but was significantly less for cacti, which have only a few flowers per plant and  flowers that persist only one or a few days.

 

        COMMUNITY PATTERNS OF BLOOMING____ The rainfall and temperature sum categorizations of Table 4 were used to generalize soil moisture indices (MI) for spring and summer of each year (Table 8). Rainfall categories, as they related to the input of water to the soil,  were assigned  arbitrarily values of: W* = 3, W = 2, m = 1. N = 0, d = -1,  D = -2, D* = -3  Temperature sum categories, as they related to evaporation of soil moisture, were given half the weight of the rainfall categories: C* = 1.5, C = 1, c = 0.5, No = 0, wm = -0.5, H = -1, H* = -1.5. The earliest blooming of plant species in spring was compared to moisture indices summed for winter and early spring; earliest summer blooming was compared to MI values summed for late spring and early summer.

        The eight spring annuals bloomed earliest in years with soil moisture indices in the middle range of values (2.5 to -1.5) (Table 8).  Earliest bloomings did not occur in years with high MI (5.0 to 3.0), nor lowest MI (-5.5). The distribution of species bloomings was highly significantly different from random (X2 = 30, df = 7,  P < 0.005).  In two years with earliest bloomings (1995, 1989) the contribution of winter to the summed MI was positive (1.5-2.0 respectively) but early spring was negative (-1.0,  -3.0). In 1988 both winter and early spring were positive MI (0.5 and 2.0 respectively).  The earliest bloomings of 17 perennials were spread over the range of MI values (except the lowest, -5.5) with a tendency towards medium to low values (2.5 to -1.5), which was marginally significant (X2 = 12.9, df 7, 0.10> P < 0.05).  In spring no annuals bloomed in 1990 when MI was -5.5; 71% of perennials also failed to bloom that year.

        In summer, with data for only three species of annuals, there was no significant difference of distribution of earliest blooming from random.  Likewise for 16 perennials (in both cases X2  0.50 > P > 0.25).

        When the magnitude of flowering (Nt) of species is compared to the moisture indices summed over all phenological periods relevant to spring flowering and summer flowering (Table 9 ) the flowerings of annuals were greatest in spring at MI values of -1.0 to 3.5, very significantly different from random Flowerings of spring perennials were significantly distributed above MI 2.0, with major weighting at 2.0.  In summer, flowerings of annuals were greatest only at MI values 2.0 to 5.0, whereas perennials were spread over the range of values.

        This attempt to find community patterns in flowerings is, of course, confounded by the adaptations of the phenotypes of individual species, which are dealt with next.

 

FLOWERING CHARACTERISTICS OF INDIVIDUAL SPECIES

 

        WEEK OF FIRST BLOOM____ The week of first bloom each year was regressed against rainfall and temperature sum of appropriate phenological seasons for the 14 species that had a range of variation of first flowering of at least 5 weeks during the years of observation (Table 10).  Variables were significantly predictive of first week of spring blooming for two of three annuals; blooming was advanced for Baileyamultiradiata by an increase in winter precipitation, and advanced for Gilia  longifora by increase in temperature sum in early spring. Only one of nine spring-flowering perennials, Bahia, had a significant relationship; flowering was retarded by higher winter temperature sum.

            First summer flowering was significantly predictive for one of three summer annuals; Baileya;flowering was advanced by increase in early summer precipitation. Five of eight perennials were significantly predictive by summer rainfall or temperature sum (Table 10). Gutierrezia, which does most of its vegetative growth in winter and does not bloom until early or late summer, showed complicated predictive relation- ships; first bloom was delayed by increased late spring rainfall and advanced by higher late spring temperature sum, but was advanced by increased early summer rainfall and retarded by higher early summer temperature sum.

 

        MAGNITUDE OF TOTAL BLOOM (Nt)____   For almost all speciesthere was considerable variation in the number of flowering individuals censused from year to year (Tables 11a and 11b). These detailed data are provided for the use of anyone wishing to do a different analysis than provided herein.  For 18 species (Table 11a) the data for 1987-1994 provided significant whole models for the regression of Nt based on rainfall and temperature sum. For nine species (Table 11b) there was no significant relation-ship, but the data for five species show remarkable low and high values of Nt in the “exceptional years” 1994-1996.

        For the 18 species of Table 11a the models of Nt are highly significant with AR2 > 0.86 in all but one case. Models have a median of three independent variables (range 1-5).  For spring bloomings (Table 12), Nt is predicted by a mixture of variables of the current year (10 of 24) and previous year (14/24); the latter variables range from the previous fall through the previous early spring. Only one annual and three perennials are principally predicted only by current year variables. Five perennials are predicted principally by previous year variables. Precipitation is a dominant variable in 17 of 24 cases, whereas temperature sum is so in 7 of 24.

        For summer bloomings (Table 13) variables of the current year are more often predominantly predictive (14/21) than those of the previous year (7/21), as compared with spring data.  Precipitation is predictive in 14 of 21 cases (7+, 7-) and temperature sum 7/21 (2+, 5-). In one case, the number of Gilia flowering in spring is the second most predictive variable of Gilia summer Nt, the only species for which this is true.  Spring Nt was made an independent variable only  for models of summer flowering.

        There is no case in which two species have the same predictive relationship for the two most significant variables. Aster  hirtifolius  and Dyssodia  acerosa do share the same two variables (Table 12) but for one of them (Pfall*)  the relationship is  different.  

 

        “EXCEPTIONAL” YEARS”_____ In early 1994 it was sensed that the year was unusually dry__flowering by spring species was zero or nearly so (Table 11a, b).  Censusing of Nt was then extended to summer 1994, spring 1995 and summer 1996. For spring 1994 and 1996 no censuses were necessary, except for a few species, since there was no or very little flowering; The symbol (*) represents a “census” of zero in these cases  Limited observations of phenology continued through 1999.  In comparison to 1987-1993, 1994-1999 were exceptional in one way or another. Five of the six minimum values of rainfall for the six phenological seasons occurred in 1994-1999 (Table 1). The lowest total annual rainfall occurred in 1994. Every seasonal maximum for temperature sum occurred 1994-1999, with the highest annual total being in 1999 (Table 3).  Minimum rainfalls together with maximum temperature sums suggest low soil moisture 1994-1999 except as ameliorated by the second highest late summer rainfall in 1995. The coincidence of certain plant phenologies with the different precipitation and temperature sum patterns of 1994-1999 are very suggestive of cause and effect.

        Year 1994   was distinctive in having warm to very hot weather (wm, H, H*, Table 4) from winter through fall. Annual rainfall was a minimum in 1994; only early spring and late summer were normal rainfall (N) while other seasons were dry (d or D).  There were several negative consequences for flowering: (1) There was a failure of spring flowering more complete than after the 1990 spring drought (62 mm total for winter through late spring for 1990, Table 1, compared to 101 mm for 1994.  Larrea tridentata  was the only species that bloomed sufficiently to be censused, and unexpectedly had its maximum spring Nt in1994 (Table 11a). (2) Progressive leaf stress and senescence and some mortality of subshrubs was observed May through July. (3) Although rains began July 7, recovery of shrubs delayed their blooming by an average of 29 days (range 10-50 d) beyond their average beginning times.  A first summer census was not needed until August 24.  (4) Success of summer flowering was varied (Tables 11a, 11b).  Only a few seedlings of the biseasonal annuals, Baileya and Gilia  longiflora survived from winter to bloom.  Conversely, five biseasonal perennials had an average or above average summer Nt so that the community aspect was dominated by the flowers of Zinnia, Parthenium and Menodora.  Summer-only perennials varied: two bloomed near average; GutierreziaNtwas only 4% of its summer maxi-mum (to be expected because of its winter-spring vegetative growth), whereas fall-blooming Flourensia had its maximum bloom. (5) Although Gutierrezia  appeared to have greened-up normally through early spring, by Nov. 425 of 472 individualswere dead.  Although even moderate mortality of stems had not been observed previously for subshrubs, by Nov. 15.6% of Zinnia appeared dead; 25.5% of Dalea appeared completely dead and 50% had some dead stems.

        In 1995   weather was unusual in the way in which seasonal temperature sums enhanced or counteracted seasonal rainfalls, as they could affect soil moisture from winter through fall: (Table 4): WNo, NH, D*C, NC, WH, DH*. Annuals and most perennials responded in opposite ways. Winter rainfall was the second highest value (136 mm) and was equitably distributed over 13 of 14 weeks.  Species that were keyed to winter rainfall were very successful: (1) Seven annuals had spring flowerings that ranked first for all years. (2) Four annuals were present that had not been censused before, two of which had not been recorded for the site (Table 14). (3) Eleven annuals bloomed 2.2 wk earlier than on average (range 1-4 wk). (4) Two early flowering species, the geophytes Cymopterus and Dichelostema had their maximum spring Nt along with Dyssodia  acerosa, which showed an ability to respond to rainfall regard-less of season. Spring/summer perennials bloomed very poorly, with five species failing to bloom, or almost so (Tables 11a, 11b). Consequently no census was made in summer.

        In 1996: The drought, very warm (D,H*) of fall 1995 continued as a stress into the (D,H) of winter 1996, and dry, warm (d,wm) of early spring, even more severe than 1994. The above average rainfall (W) of late spring was partly offset by the maximum temperature sum (W,H*), which gave way to a normal early summer. Shrubs (Larrea, Flourensia) and subshrubs (Dalea,  Parthenium,  Zinnia) showed leaf stress until they began to green-up at the end of early spring. Almost all species had negligible spring blooms in 1996, as in 1994.  The years differed in the varied responses of summer flowerings (Table 11a).  <