Journal of Experimental Botany, Vol. 51, No. 343, pp. 275-286,
February 2000
© 2000 Oxford University Press
Estimation of base temperatures for nine weed species
1 Department of Botany and Plant Sciences, University of California, Riverside, CA 92521, USA
2 University of California Kearney Agricultural Center, 9240 S. Riverbend Avenue, Parlier, CA 93648, USA
Received 15 September 1999; Accepted 30 September 1999
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionEcological significance of Tbase...Supra-optimal temperaturesConclusionsReferences |
|---|
Experiments were conducted to test several methods for estimating low temperature thresholds for seed germination. Temperature responses of nine weeds common in annual agroecosystems were assessed in temperature gradient experiments. Species included summer annuals (Amaranthus albus, A. palmeri, Digitaria sanguinalis, Echinochloa crus-galli, Portulaca oleracea, and Setaria glauca), winter annuals (Hirschfeldia incana and Sonchus oleraceus), and Conyza canadensis, which is classified as a summer or winter annual. The temperature below which development ceases (Tbase) was estimated as the x-intercept of four conventional germination rate indices regressed on temperature, by repeated probit analysis, and by a mathematical approach. An overall Tbase estimate for each species was the average across indices weighted by the reciprocal of the variance associated with the estimate. Germination rates increased linearly with temperature between 15 °C and 30 °C for all species. Consistent estimates of Tbase were obtained for most species using several indices. The most statistically robust and biologically relevant method was the reciprocal time to median germination, which can also be used to estimate other biologically meaningful parameters. The mean Tbase for summer annuals (13.8 °C) was higher than that for winter annuals (8.3 °C). The two germination response characteristics, Tbase and slope (rate), influence a species’ germination behaviour in the field since the germination inhibiting effects of a high Tbase may be offset by the germination promoting effects of a rapid germination response to temperature. Estimates of Tbase may be incorporated into predictive thermal time models to assist weed control practitioners in making management decisions.
Key words: Base temperature, germination, phenology, thermal time, weeds
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionEcological significance of Tbase...Supra-optimal temperaturesConclusionsReferences |
|---|
Temperature is the single most important factor regulating germination of non-dormant seeds in irrigated, annual agroecosystems at the beginning of the growth season where light, nutrients, and moisture are typically not growth limiting (Garcia-Huidobro et al., 1982
). Consequently, the combination of temperature and time (i.e. degree-days) is a more appropriate unit of measure for predicting plant development than time alone (Ritchie and NeSmith, 1991). The degree-day approach has been implemented successfully in the prediction of agronomically important phenological stages for crop and weed species (Alm et al., 1991).
A basic requirement for this approach is an estimate of the temperature below which phenological development ceases (Tbase) for each species. Conventional approaches to estimating Tbase involve regressing germination rate on sub-optimal temperatures and computing an x-intercept from the linear equation (Holt and Orcutt, 1996; Scott et al., 1984; Wiese and Binning, 1987). Germination rate has been expressed in various ways, but no one way is universally superior in its linear relationship to temperature (Garcia-Huidobro et al., 1982; Scott et al., 1984).
Alternatively, other approaches rely on the sigmoidal shape that cumulative germination takes over time for most species (Scott et al., 1984). For example, cumulative per cent germination is probit transformed and then paired with the temperature-time (i.e. ln[T-Tbase]xtime) at which the probit was observed (Dahal et al., 1990). The probit minus five is equal to a normal equivalent deviate (NED) (Finney, 1971). The NEDs are in units of standard deviation and are defined by the standard normal cumulative density function (CDF). A straight line is fitted to the paired data, and Tbase is varied until the mean square residual for the regression is minimized. This has been deemed the best estimate of Tbase for several crop species (Covell et al., 1986; Dahal et al., 1990). The Weibull (Bridges et al., 1989) and the logistic (Hsu et al., 1984; Talbott Roché et al., 1997) functions have also been fitted to cumulative per cent germination. A method by which Tbase could be estimated according to least associated variation in degree-days or actual days would also be appropriate in the development of degree-day models. Several alternative methods have been developed based on several least variation criteria (Yang et al., 1995), which estimate Tbase from paired temperature and time to germination data.
A fundamental challenge in developing thermal time models has been deciding which method provides the most robust estimate of Tbase. This can be especially problematic when each of the methods yields a different Tbase estimate yet the mean square residuals, coefficients of determination (r2), and coefficients of variation (CV) may not be significantly different among the methods. Combining the estimates from several different methods and weighting them by the reciprocal of the variation associated with each estimate provides a mathematical approach for estimating an overall Tbase for a species. However, it is difficult to estimate the variance associated with the x-intercept, which is the negative ratio of the y-intercept and the slope of a linear equation. Additionally, a composite value for Tbase is not necessarily biologically meaningful. Alternate methods are to choose the estimate that most closely matches observed germination thresholds obtained empirically or that also predicts the time-course of germination most accurately.
The primary objective of this research was to compare several conventional and alternative methods that have been used to estimate Tbase for germination. Estimates of Tbase were made for nine common weed species differing in taxonomic classification and life history. Statistical and biological criteria were employed to choose the most robust and realistic estimates for incorporation into predictive thermal time models.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionEcological significance of Tbase...Supra-optimal temperaturesConclusionsReferences |
|---|
Seed collection
Seeds of nine weed species common to California agroecosystems were collected between August 1994 and May 1996 at the University of California Research Station at the Riverside campus (Table 1
). The species were classified into two groups according to their life-history habits. Six species were summer annuals (Amaranthus albus L., A. palmeri S. Wats., Portulaca oleracea L., Digitaria sanguinalis (L.) Scop., Echinochloa crus-galli (L.) Beauv., and Setaria glauca (L.) Beauv.), which germinate in spring and flower in summer. Two species were winter annuals (Hirschfeldia incana (L.) Lagreze-Fossatt and Sonchus oleraceus L.), which germinate in autumn and flower the following spring. Conyza canadensis L. Cronq. was classified as a winter-summer annual because it germinates in autumn or spring and grows as a rosette until it bolts and flowers in the late spring or summer (Whitson et al., 1992). Seeds were stored in paper bags at room temperature (~25.3 °C) and relative humidity (<15%).
|
Experimental design
The effect of temperature on germination response was assessed in experiments conducted between September 1995 and April 1997 on an insulated 45x114x3 cm temperature gradient bar (Holt, 1987). The bar is a solid aluminium block with hot water running through one end and cold through the other resulting in a continuous temperature gradient along its length. The hot (37 °C) and cold (-5 °C) water was provided by two thermostatically controlled water baths (Lauda model RC-20 with B-2 circulator and model C-20 with B-1 circulator). The resulting average temperature range experienced by the seeds was 36 °C to 9 °C.
All seeds were germinated on moistened filter paper, which was placed in 53 mm diameter by 8 mm deep glass Petri dishes. A wick action method provided a continuous, uniform moisture supply for the seeds (Knudson and Tibbitts, 1973). Dishes were monitored daily for water loss and refilled with distilled water as necessary. For each species, 16 dishes were placed at about 2 °C increments along the gradient bar in each experiment. Germination experiments were conducted twice, totalling 32 dishes with 20 seeds per dish for each species.
Fluctuations about the mean temperature in each dish ranged from 1–3 °C with the warmer temperatures occurring when room lights were on. Mean temperature of each dish was based on an integrated average. This was computed by measuring temperature of each dish hourly at the beginning of the experiment and every 6 h after it was apparent that temperatures did not fluctuate during the day or night hours. Temperature measurements were made with copper-constantan thermocouples (Omega Engineering, Inc., 1 Omega Drive, Box 4047, Stamford, Connecticut 06907 USA) and monitored on a CR5 Digital Recorder (Campbell Scientific, Inc., Box 551, Logan, Utah 84321 USA). The thermocouples were inserted through holes in plastic Petri dish lids and arranged so that the soldered tip was in direct contact with the moistened filter paper. Fluorescent and incandescent light banks provided at least 40 µmol m-2 s-1 for 24 h d-1, which was sufficient to meet light requirements for germination of light-sensitive species.
Dishes were subjected to treatment temperatures on the bar for up to 21 d. Seeds were checked for germination four times a day at the beginning of each experiment while germination was rapid and once a day by the end of each experiment when germination had slowed. Germinated seeds were removed from the dish once the radicle or stem had extended more than 1 mm beyond the seed coat or caryopsis, respectively. All dishes were brought to room temperature for a 14 d post-treatment period following each 21 d treatment period in order to assess final germinability. It was assumed that 5 weeks was sufficient time for non-dormant seeds to germinate. A younger seed lot was used for subsequent experiments if germinability was less than 50% in any Petri dish following the post-treatment period. On the few occasions that germinability was below 50%, the data for that dish were not utilized in subsequent analyses.
Germination indices and analyses
Analyses using four conventional germination rate indices, a repeated probit analysis and a mathematical approach were conducted with the data of each species, treating data from each dish separately. The optimum temperature was the temperature at which the highest germination rate was observed. The data utilized in calculating the indices were those from suboptimal temperatures. Suboptimal temperatures were those below the optimum temperature over which germination rate increased linearly according to visual inspection of the residuals. Total potential germinability of a species at any temperature along the gradient bar was assumed to be equivalent to that under optimal temperature, since there was no evidence for systematic differences in overall germinability that might be explained by temperature. In calculating the indices, total potential germinability was based on the total number of seeds that had germinated following the post-treatment period for each dish.
Index 1. Per cent germination:
Average per cent germination per day was calculated at each temperature as:
 |
). Only data between 10–90% germination were utilized in the calculation of Index 1 since several seeds often germinated between inspections and the exact times for 10% or 90% germination could not be determined.
Index 2. Reciprocal time to median germination:
The reciprocal time to median germination was assessed at each temperature for each species (Holt and Orcutt, 1996). In most cases, germination of the median seed was not observed directly so the PROBIT procedure in SAS was utilized to estimate the time that the median cohort germinated (SAS Institute, Inc., 1989, 1995). Cumulative per cent germination at a given temperature was transformed to probits and regressed against time (Finney, 1971) and the time to median (50%) germination was estimated from this function (Fig. 1). A Pearson Chi Square test (
=0.05) was conducted to assess goodness of fit of the normal cumulative density function to per cent germination through time for each species at each suboptimal temperature (SAS Institute, Inc., 1995). Linearly interpolated values were utilized if the goodness of fit tests indicated significant divergence between predicted and observed times.
|
Index 3. Germination rate index (A):
The Germination Rate Index (A) was calculated as:
 |
).
Index 4. Germination rate index (B):
The Germination Rate Index (B) was calculated as:
 |
).
Index 5. Repeated probit analysis:
Cumulative germination as a per cent of germinable seeds at each time, i, was transformed to a normal equivalent deviate (NED) value using the NORMSINV function in Microsoft® EXCEL (Version 7), which returns the inverse of the standard normal cumulative density function with a reported accuracy within 3E-7. All cohorts of a given species were assumed to have a common base temperature (Ellis et al., 1986). These NED values from all times and suboptimal temperatures were pooled and regressed against a function of time (ti) and temperature (T) as:
 |
; Ellis et al., 1986). The intercept and slope were estimated by ordinary least squares methods and Tbase was estimated by an iterative method (Dahal et al., 1990). The iterative method involved varying Tbase in equation (4) until the mean square residual term of the regression was minimized. The Tbase producing the least residual was deemed the best estimate of base temperature according to this method (Dahal et al., 1990; Ellis et al., 1986).
Index 6 (a–d). Mathematical approach:
Estimates of Tbase were also made utilizing the mathematical formulae of Yang et al. (Yang et al., 1995). These formulae reduce to single equations the Tbase estimates for each species based on least standard deviation in degree-days (Index 6a), least standard deviation in days (Index 6b), least coefficient of variation in degree-days (Index 6c), and a regression coefficient method (Index 6d). These criteria are commonly utilized methods in the development of degree-day models. The Tbase estimate based on least standard deviation in degree-days (Index 6a) was calculated as:
 |
 |
 |
 |
Linear regressions
Germination indices 1–4 were regressed against suboptimal temperatures for each species utilizing the regression procedure in SAS (SAS Institute, Inc., 1989). The x-intercept from each of these regressions was the estimated Tbase, computed as -a/b, where a and b are the intercept and slope coefficients from each of the regressions. All comparisons among means, variances, coefficients of variation, and coefficients of determination were performed at the 5% significance level (Zar, 1984). The CVy (coefficient of variation for the dependent variable) for Index 5 was computed after transforming NED values to probits to avoid division by zero.
Estimate of variance associated with x-intercept
The standard error associated with the x-intercept (SEx-int) of indices 1–5 was estimated from the variance–covariance matrix generated with the OUTEST COVOUT options available with the regression procedure in SAS (SAS Institute, Inc., 1989). It was computed as:
 |
).
A weighted average (WA1) was computed utilizing the inverse of the standard error (SEx-int) associated with the x-intercept (xint) from each of the first four indices i as a weighting factor such that:
 |
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionEcological significance of Tbase...Supra-optimal temperaturesConclusionsReferences |
|---|
Seed germinability remained consistently high at 77.4±3.9% averaged over all species (Table 1). Germinability neither increased nor decreased significantly throughout the entire experiment.
Germination rates for all species and indices increased with temperature within the range of 10 °C to about 30 °C as indicated by significant F-tests for regression (P<0.05) (Figs 2–5). Several data points for Amaranthus palmeri extended beyond the y-axes in Figs 2–5and were not presented so that data of the other species could be interpreted more easily. In general, a straight line was adequate to describe the relationship of the different germination rate indices to suboptimal temperature for all species. Examination of the ellipses of concentration and the correlation coefficients (
r2) revealed a strong linear relationship in most cases (Table 2). Additionally, residual plots did not reveal any systematic deviation of observed values from predicted values, which would have suggested the use of an alternative model (data not presented).
|
|
|
|
|
The Tbase estimates for a given species were not more than 3 °C different among all indices for Amaranthus albus, A. palmeri, Conyza canadensis, Digitaria sanguinalis, and Echinochloa crus-galli (Table 2). Differences in estimates among the indices for the remaining four species ranged from 5.6 °C for Setaria glauca to 8.2 °C for Sonchus oleraceus. The highest variability among estimates based on a pooled SEx-int and CVx-int was observed with Portulaca oleracea and S. oleraceus (Table 3
). Estimates from Index 5 (repeated probit analysis) of Tbase for the two winter annuals, Hirschfeldia incana and S. oleraceus, were significantly lower than those of the other indices according to a standard t-test using the pooled SEx-int over Indices 1–4 for each species (Tables 2, 3). Index 5 estimates for the other species were not significantly different from Tbase estimates arithmetically averaged over the other indices.
|
The reciprocal time to median germination (Index 2) provided the most robust Tbase and best linear fit among indices across species as indicated by an average SEx-int, CVy, and r2 of 0.99 °C, 20.5%, and 0.84, respectively (Table 2). Cumulative germination was described well by the normal cumulative density function (CDF) that was used for Index 2, as Pearson Chi Square goodness of fit tests indicated no significant differences (P>0.10) between observations and values predicted by the normal CDF (Fig. 1). Similar fits were achieved for the other species (data not presented). The data in Fig. 1 represent some of the largest differences among probit predicted, linearly interpolated, and nearest observed median values found for all species and temperatures. Time to median germination was often quite different between nearest observations and those estimated by Index 2. However, the times estimated by Index 2 were very similar (within a half day) to linearly interpolated values (Fig. 1).
By these same indicators, Index 4 (GRIB) provided the poorest fit of the first four indices with an average SEx-int, CVy, and r2 of 1.70 °C, 36.0%, and 0.66, respectively (Table 2). Index 3 (GRIA) was intermediate in its estimation of Tbase and linear fit to temperature (Table 2). The germination cohorts from the 0–10% and 90–100% percentiles were omitted in the determination of Index 1 (per cent germination) since their inclusion resulted in significantly lower r2 and larger SEx-int values according to pairwise Z-tests and F-tests. Calculated in this way, the Tbase estimate and linear fit of Index 1 to temperature, averaged across all species, were also intermediate in statistical robustness (Table 2).
Index 5 (repeated probit analysis) cannot be compared with the other indices in the same manner because the independent variable was in different units (i.e. degree-days) and was also logarithmically transformed (Fig. 6). By one unitless measure, r2, Index 5 provided the poorest (0.64) fit of any index, yet by another measure, CVy, it provided the best (8.8%) fit across species (Table 2). However, these indicators are only a partial reflection of how well the normal CDF fits the linear portion of the germination progression data because including cohorts from 0–10% and 90–100% percentiles in Index 5 resulted in significantly lower r2 and higher SEx-int values, similar to Index 1.
|
The Tbase estimates based on Index 6a–d (the mathematical approach; Yang et al., 1995
) tended to be 2–3 °C higher than those based on Indices 1–5 (Table 4). Index 6b (least standard deviation in days) produced estimates that were especially high. Index 6d (regression coefficient) consistently provided Tbase estimates that were within 1 °C of the arithmetic or weighted averages for Indices 1–4.
|
The arithmetic (AA) and two weighted (WA1, WA2) averaging methods provided Tbase estimates for each species that were within 1 °C of each other across all indices except for the two winter annuals, Sonchus oleraceus and Hirschfeldia incana (Table 3). Here, the arithmetic average Tbase was not more than 1.5 °C lower than one or both of the weighted averages. The Tbase estimates for the summer annual species tended to be higher than those for the winter annual species regardless of the averaging method (Table 3; Figs 2–5). Based on WA2, the estimated Tbase for the summer species was 13.8 °C (±SD 2.7 °C), while it was 8.5 (±SD 3.9 °C) for the winter species (not including Conyza canadensis). Similar differences were observed with the other averaging methods. The estimated Tbase for C. canadensis appeared to be intermediate between summer and winter annuals, averaging close to 13 °C. The summer annual Portulaca oleracea was an exception to these trends, with an estimated Tbase of about 9 °C. The only species with a lower estimated Tbase was the winter annual, S. oleraceus, with a Tbase of 5.7 °C. Amaranthus albus, A. palmeri and Digitaria sanguinalis had the highest Tbase estimates of all species at 15.7, 17.0, and 15.1 °C, respectively, according to WA2 (Table 3). There was no apparent difference in estimated Tbase between the summer dicotyledon and summer monocotyledon groups primarily because of the variability introduced by P. oleracea and A. palmeri within the summer dicotyledons (Figs 2–5a, b).
In terms of the rate of germination response to temperature, Amaranthus palmeri and the winter annual Hirschfeldia incana ranked first and second highest according to their slope coefficients across Indices 1–4 (Table 3). A. albus and the other winter annual, Sonchus oleraceus, had the next two highest ranked slope coefficients. The average rank of the slope coefficients for Portulaca oleracea was lowest among all species at 8.5 across Indices 1–4.
Germination of Conyza canadensis appeared to be the most variable compared to the other species as indicated by an r2 value of 0.66 averaged across Indices 1–5 (Table 3). The r2 values and the standard errors associated with the Tbase estimates (SEx-int) were negatively correlated (r=-0.85) for the first four indices. Consequently, variability was also high for this species as indicated by an SEx-int of 1.53 °C averaged across Indices 1–4. Portulaca oleracea and Sonchus oleraceus had SEx-int estimates that were higher at 1.94 °C and 1.80 °C, respectively (Table 3). The two Amaranthus spp. were least variable by these two indicators with an average r2 of 0.86 and a pooled SEx-int of 0.65 °C compared with the 0.72 and 1.46 °C of all other species for the same indicators. A. palmeri, P. oleracea and S. oleraceus had less than 20% relative variability across all indices according to the coefficients of variation in germination rate (CVy) while the CVy for the other species ranged from 20.5 to 26.0% (Table 3).
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionEcological significance of Tbase...Supra-optimal temperaturesConclusionsReferences |
|---|
The indices utilized here represent transformations that linearize germination data with respect to suboptimal temperature or the logarithm of thermal time. The transformations simplify parameter estimation and comparison. Several problems may arise with methods that rely on linear models, including lack of fit and an inflated error term due to the fact that Tbase is extrapolated (Phelps and Finch-Savage, 1997
). However, the indices used in this research provided transformations that were represented adequately by linear models. Consistent x-intercept estimates were obtained for most species using several different germination indices. Based upon residuals, parameter estimates did not suffer from non-linearity (as described in Marshall and Squire, 1996). This is partially due to the fact that few data points from very low or high temperatures (e.g. close to the estimated Tbase or above optimal temperatures) were included in the analyses. Data within 2 °C of an estimated Tbase were not included in the analyses because daily temperatures in the laboratory fluctuated several degrees. Including these data would underestimate Tbase because, during periods when temperatures decreased below the true Tbase for a species, the average daily temperature would decrease, but the average germination rate would remain constant (i.e. zero). Linear methods of estimating germination parameters are common, as in Angus et al., who compared linear and non-linear equations and found that linear models provided satisfactory estimates of thermal responses and Tbase of emergence for 44 crop species (Angus et al., 1981). In that research, biologically meaningful Tbase estimates were obtained by excluding both lower and upper tails of cumulative germination curves, as was done here (Angus et al., 1981).
Another concern regarding linear models is that they may not characterize seed populations accurately since seed lots may be comprised of subpopulations, each having a different Tbase (Phelps and Finch-Savage, 1997). In this research it was assumed that seeds of each species were drawn from that population at random so that any effect of subpopulations would be minimized. The relatively constant germinability across suboptimal temperatures observed here supports the assumption that subpopulations with different physiological responses to temperature either were not present or, if present, did not influence results. However, it is possible that such effects could occur undetected, for example, if slow germinating seeds also did not germinate at colder temperatures, such that Tbase were overestimated.
Evaluating differences among indices was among the objectives of this investigation. However, statistical comparisons by conventional methods are not valid because the indices use much of the same data and therefore are not entirely independent. Data from different temperatures and species are independent as they come from different Petri dishes and completely randomized experiments. Approximations to SEx-int provided by Equation 9 allow direct comparisons of x-intercepts among species (Cox, 1990). The weighted averaging methods that utilize SEx-int provide a robust approach to combining estimates from different indices. The weighted average WA2, which utilizes the inverse of CVx-int as a weight, tends to be biased against Index 5 (repeated probit analysis) since in that index the variability associated with the x-intercept (thermal time) includes the variability associated with time to germination in addition to the variability associated with temperature. These composite estimates are useful for determining the variability among estimates derived from different calculation methods.
Investigators have utilized analyses of germination responses to temperature that do not rely heavily on the day that the first and last cohorts germinate because of the high variability associated with these cohorts compared with that of the median cohorts (~40–60%) (Dahal et al., 1990, Hsu et al., 1984). As cumulative germination values approach 0% and 100%, NED values approach negative and positive infinity, respectively. Therefore, a cumulative germination of 0% or 100% could not be utilized in Index 5 (repeated probit analysis). Significant reductions in r2 values and increases in variability associated with the x-intercept were found when the 0–10% and 90–100% cohorts were included in Index 1 (per cent germination) and Index 5. For this reason, when estimating Tbase it seems prudent to utilize indices that rely primarily on the observed or estimated median cohorts such as Indices 2 (reciprocal time to median germination) and 5. However, variability in the early and late cohorts is what makes weed germination so unpredictable, and thus, weed management so difficult. Thus, over- or underestimates in Tbase that might occur from focusing on the majority germination response and omitting very early and late cohorts may need to be examined.
As in other reports, the reciprocal time to median germination (Index 2) provided the best estimate of Tbase in this research. The robustness of the probit analysis utilized for Index 2 likely occurred because the first and last cohorts carry little weight in probit analysis (Finney, 1971) and the entire germination progression was utilized in the estimate of median time rather than one or a few data points. Similar observations have been made by other authors (Hsu et al., 1984; Talbott Roché et al., 1997), who utilized a logistic function to estimate median times, which is similar to the normal CDF. The PROBIT procedure in SAS allows for logistic and Gompertz functions to be utilized as CDFs in the analysis, but these provided worse fits than the normal CDF according to the lack of fit tests. The degree to which germination progression follows a normal CDF is indicative of the robustness of the approach, but this CDF may not always provide the best fit to germination (Oryokot et al., 1997).
Index 4 (GRIB) tended to provide the poorest fit among indices for all species, possibly because this index included both early (0–10%) and late (90–100%) cohorts. Index 3 (GRIA) also included these same cohorts yet was less affected by their variability since dividing by the total per cent germination of seeds in the Petri dish had a stabilizing effect. Germination rate and per cent germination are often combined into one term to estimate Tbase; however, there are objections to this approach since it confounds the effect of response time and per cent germination (Heydecker, 1973; Scott et al., 1984).
Estimates made by the regression coefficient method (Index 6d; Yang et al., 1995) appeared to correspond fairly well with Indices 1–4 for all species. Minimizing the standard deviation in days (Index 6b) has been used as a criterion to validate degree-day models (Wilen et al., 1996). However, this formula (Yang et al., 1995) provided Tbase estimates that were consistently higher than those provided by any other method utilized in this research. From a statistical standpoint, Indices 5 (repeated probit analysis) and 6 (mathematical formulae) are attractive methods because they do not rely on extrapolation to obtain the Tbase estimate.
|
Ecological significance of Tbase estimates |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionEcological significance of Tbase...Supra-optimal temperaturesConclusionsReferences |
|---|
For most species, similar Tbase estimates were obtained regardless of the calculation method, although methods differed in statistical robustness and goodness of fit. However, Index 2 (reciprocal time to median germination), the most statistically robust index tested, is also expected to be the most biologically relevant index since parameter estimates are obtained from an analysis that utilizes the progression of cumulative germination over time (Angus et al., 1981
; Hsu et al., 1984). Such an analysis may be used to estimate other biologically meaningful parameters in addition to Tbase, such as time to onset of germination, maximum germination rate, and germination response to various factors (Hsu et al., 1984). Thus, the validity of Index 2 for estimating Tbase is strengthened by its broader applicability for predicting the time-course of germination in the field. In addition, Tbase estimates from Index 2 were consistent with observations of seed germination during experiments; that is, no seeds germinated below the estimated values.
The two winter annuals (Hirschfeldia incana and Sonchus oleraceus) appear well adapted for germination in cooler temperatures as they had relatively low Tbase estimates compared to the summer annual species. Conyza canadensis had Tbase estimates intermediate to those of the winter and summer annuals. Therefore, its classification as either winter or summer annual seems appropriate (Whitson et al., 1992). The greatest variability in Tbase estimates among indices was found for the summer annual, Portulaca oleracea, which had a Tbase close to the values for winter annual species. In other research, P. oleracea seeds were dormant upon dispersal and had greater germination under alternating than constant temperatures, although sensitivity to temperature fluctuations disappeared as dormancy was broken by seed storage for several months at room temperature (Kruk and Benech-Arnold, 1998). These authors also estimated a low Tbase for P. oleracea (7 °C), consistent with our findings. Since seed age and physiological status are known to influence germination responses to temperature (Probert, 1992), seed dormancy status and sensitivity to alternating temperatures should be determined prior to conducting experiments such as those reported here.
In the field, a species with a relatively high Tbase would experience little germination in early spring when temperatures would frequently drop below its Tbase. However, a rapid germination response to an increase in temperature (i.e. large slope) could offset the low early germination caused by the high Tbase. For example, Amaranthus palmeri had the highest estimated Tbase (17 °C), but also the highest ranked slope of all species studied. A large percentage of the germinable seeds of this species would likely germinate rapidly after the first hours that temperatures exceeded 17 °C in the field. Therefore, even though its Tbase was high relative to other species, A. palmeri may not be underrepresented in the early season weed population because of its high response rate. A. palmeri is native to the hot, dry, high light conditions of the Sonoran desert in the south-western United States where moisture is limiting; therefore, germinating quickly following imbibition is an adaptive characteristic of this species (Pearcy and Ehleringer, 1984).
Portulaca oleracea had the lowest Tbase of the summer annuals, but also the lowest slope of the species studied. With a small germination response to an increase in temperature, only a small percentage of the germinable seeds of P. oleracea would likely germinate over a given period of time. However, due to a low Tbase, the duration of germination by this species would likely be much longer than for A. palmeri. Thus, it appears that each of the summer annuals could be equally represented in the early spring weed population due to these offsetting characteristics. A pattern of decreasing thermal time to germination with increasing Tbase was also observed for 44 crop species and attributed to selection for early competitive advantage (Angus et al., 1981).
The differences in Tbase and slope estimates observed for the two Amaranthus species is not unusual for this genus. Different estimates were also found for two Amaranthus species (A. powellii S. Wats. and A. retroflexus L.) (Oryokot et al., 1997). As with these results, they found that the species with the higher Tbase also had the higher estimated slope. Base temperature estimates have been reported by others for the species studied here, including Digitaria sanguinalis (Alm et al., 1988), Echinochloa crus-galli (Wiese and Binning, 1987), and Portulaca oleracea (Kruk and Benech-Arnold, 1998). Different estimates among published reports are expected, however, due to selection of local ecotypes.
|
Supra-optimal temperatures |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionEcological significance of Tbase...Supra-optimal temperaturesConclusionsReferences |
|---|
Temperatures above which germination does not occur were not estimated here. Direct temperature measurements at the University of California, Riverside Research Station at 1–5 cm depths in soil with adequate moisture levels for germination indicated that maximum temperatures were between 5–10 °C cooler than maximum air temperatures (unpublished data). The year 1996 was unusually warm in southern California, having eight days that were 40 °C or higher (University of California Statewide Integrated Pest Management Program, www. ipm.ucdavis.edu). The germination rates of Hirschfeldia incana, Sonchus oleraceus and Conyza canadensis had just begun to plateau or decline slightly at 35 °C on the gradient bar. Germination rates of all the other species were still increasing albeit not linearly at 35 °C. Therefore, it appears that maximum air temperatures are not likely to limit germination in soil with non-limiting moisture levels. Estimating an upper temperature threshold for germination for these species is probably unnecessary for thermal time models. While maximum summer temperatures may inhibit growth of the winter annual species, this would likely occur only following emergence.
|
Conclusions |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionEcological significance of Tbase...Supra-optimal temperaturesConclusionsReferences |
|---|
Several of the methods tested for quantifying temperature responses of germination provided consistent estimates of Tbase and were statistically robust. However, the reciprocal time to median germination method was the best choice based on both statistical criteria and biological relevance. This method distinguished summer annual species with relatively high Tbase values from winter annual species with lower values. However, differences were found within each life history type, among related taxa, and between these results and reports in the literature for the same species. Thus, estimates of germination response are not valid for all members of a species, but should be derived for locally adapted ecotypes. Data from this research provided reasonable estimates of a parameter needed for predictive thermal time models of plant development, which have utility in making weed management decisions. These initial estimates should be further validated with observations made under field conditions.
|
Acknowledgments |
|---|
This research was funded in part by a competitive grant from the University of California Statewide Integrated Pest Management (IPM) Project. The technical assistance of Ms Nanette Pratini is gratefully acknowledged.
|
Notes |
|---|
3 To whom correspondence should be addressed. Fax: +1 909 787 4437. E-mail:jodie.holt{at}ucr.edu
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionEcological significance of Tbase...Supra-optimal temperaturesConclusionsReferences |
|---|
Alm DM, McGiffen Jr ME, Hesketh JD. 1991. Weed phenology. In: Hodges T, ed. Predicting crop phenology. Boca Raton, FL: CRC Press, 191–218.
Alm DM, Pike DR, Hesketh JD, Stoller EW. 1998. Leaf area development in some crop and weed species. Biotronics 17, 29–39.
Angus JF, Cunningham RB, Moncur MW, MacKenzie DH. 1981. Phasic development in field crops. I. Thermal response in the seedling phase. Field Crops Research 3, 365–378.
Bridges DC, Wu HI, Sharpe JH, Chandler JM. 1989. Modeling distributions of crop and weed seed germination time. Weed Science 37, 724–729.
Covell S, Ellis RH, Roberts EH, Summerfield RJ. 1986. The influence of temperature on seed germination rate in grain legumes. I. A comparison of chickpea, lentil, soyabean, and cowpea at constant temperatures. Journal of Experimental Botany 37, 705–715.
Cox C. 1990. Fieller's theorem, the likelihood and the delta method. Biometrics 46, 709–718.
Dahal P, Bradford KJ, Jones RA. 1990. Effects of priming and endosperm integrity on seed germination rates of tomato genotypes. Journal of Experimental Botany 41, 1431– 1439.
Ellis RH, Covell S, Roberts EH, Summerfield RJ. 1986. The influence of temperature on seed germination rate in grain legumes. II. Intraspecific variation in chickpea (Cicer arietinum L.) at constant temperatures. Journal of Experimental Botany 37, 1503–1515.
Finney DJ. 1971. Probit analysis, 3rd edn. Cambridge, UK: Cambridge University Press.
Garcia-Huidobro J, Monteith JL, Squire GR. 1982. Time, temperature, and germination of pearl millet (Pennisetum typhoides S. & H.). Journal of Experimental Botany 33, 288–296.
Heydecker W. 1973. Seed ecology. London, UK: Butterworth and Co Ltd.
Holt JS. 1987. Factors affecting germination in greenhouse-produced seeds of Oxalis corniculata, a perennial weed. American Journal of Botany 74, 429–436.
Holt JS, Orcutt DR. 1996. Temperature thresholds for bud sprouting in perennial weeds and seed germination in cotton. Weed Science 44, 523–533.
Hsu FH, Nelson CJ, Chow WS. 1984. A mathematical model to utilize the logistic function in germination and seedling growth. Journal of Experimental Botany 35, 1629–1640.
Knudson LL, Tibbitts TW. 1973. Wick-action method for germination of seeds. HortScience 8, 472.
Kruk BC, Benech-Arnold RL. 1998. Functional and quantitative analysis of seed thermal responses in prostrate knotweed (Polygonum aviculare) and common purslane (Portulaca oleracea). Weed Science 46, 83–90.[Web of Science]
Marshall B, Squire GR. 1996. Non-linearity in rate-temperature relations of germination in oilseed rape. Journal of Experimental Botany 47, 1369–1375.
Oryokot JOE, Murphy SD, Thomas AG, Swanton CJ. 1997. Temperature- and moisture-dependent models of seed germination and shoot elongation in green and redroot pigweed (Amaranthus powellii, A. retroflexus). Weed Science 45, 488–496.
Pearcy RW, Ehleringer L. 1984. Comparative ecophysiology of C3 and C4 plants. Plant, Cell, and Environment 7, 1–13.
Phelps K, Finch-Savage WE. 1997. A statistical perspective on threshold type germination models. In: Ellis RH, Black M, Murdoch AJ, Hong TD, eds. Basic and applied aspects of seed biology. London, UK: Kluwer Academic Publishers, 361–368.
Probert RJ. 1992. The role of temperature in germination ecophysiology. In: Fenner M, ed. Seeds: the ecology of regeneration in plant communities. Wallingford, UK: CAB International.
Ritchie JT, NeSmith DS. 1991. Temperature and crop development. In: Hanks J, Ritchie JT, eds. Modeling plant and soil systems. Agronomy No. 31. Madison, WI: American Society of Agronomy, 5–29.
SAS Institute Inc. 1989. SAS/STAT user's guide, Version 6, 4th edn. Cary, NC: SAS Institute Inc.
SAS Institute Inc. 1995. Probit analysis for estimating an LD50. In: Logistic regression examples using the SAS system, Version 6, 1st edn. Cary, NC: SAS Institute Inc, 145–150.
Scott SJ, Jones RA, Williams WA. 1984. Review of data analysis methods for seed germination. Crop Science 24, 1192–1199.
Talbott Roché C, Thill DC, Shafii B. 1997. Estimation of base and optimum temperatures for seed germination in common crupina (Crupina vulgaris). Weed Science 45, 529–533.
Whitson TD (ed), Burrill LC, Dewey SA, Cudney DW, Nelson BE, Lee RD, Parker R. 1992. Weeds of the west. The Western Society of Weed Science in Cooperation with Western United States Land Grant Universities Cooperative Extension Services and the University of Wyoming. 630.
Wiese AM, Binning LK. 1987. Calculating the threshold temperature of development for weeds. Weed Science 35, 177–179.
Wilen CA, Holt JS, McCloskey WB. 1996. Predicting yellow nutsedge (Cyperus esculentus) emergence using degree-day models. Weed Science 44, 821–829.
Yang S, Logan J, Coffey DL. 1995. Mathematical formulae for calculating the base temperature for growing degree-days. Agricultural and Forest Meteorology 74, 61–74.
Zar JH. 1984. Biostatistical analysis, 2nd edn. Edgewood Cliffs, NJ: Prentice-Hall Inc.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  G. R. Chantre, D. Batlla, M. R. Sabbatini, and G. Orioli Germination parameterization and development of an after-ripening thermal-time model for primary dormancy release of Lithospermum arvense seeds Ann. Bot., June 1, 2009; 103(8): 1291 - 1301. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. P. HARDEGREE Predicting Germination Response to Temperature. III. Model Validation Under Field-variable Temperature Conditions Ann. Bot., October 1, 2006; 98(4): 827 - 834. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. P. HARDEGREE and A. H. WINSTRAL Predicting Germination Response to Temperature. II. Three-dimensional Regression, Statistical Gridding and Iterative-probit Optimization Using Measured and Interpolated-subpopulation Data Ann. Bot., August 1, 2006; 98(2): 403 - 410. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. P. HARDEGREE Predicting Germination Response to Temperature. I. Cardinal-temperature Models and Subpopulation-specific Regression Ann. Bot., June 1, 2006; 97(6): 1115 - 1125. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||