The Influence of Various Physical and Biological Factors of the Environment on Honeybee Activity. An Examination of the Relationship Between Activity and Solar Radiation

In attempting to assess the value of the honeybee as a pollinator of fruit and other crops, particularly in a season when inclement weather is likely to reduce the number of wild pollinating insects, it was realized that very little is known of the influence of the physical factors of the environment on honeybee activity. It was therefore decided in the first instance to study the effects of solar radiation, humidity, actual precipitation of water vapour, and temperature on the flying activity of the honeybee. Later it is hoped to continue with a study of the effects of nectar abundance and concentration and also the biological condition of the colony itself upon activity, It should be realized, however, that although it is simplest to study each of these factors separately they are very largely interdependent upon one another and that all work together to influence the final result. In the present paper the influence of solar radiation is considered.

Various workers have studied this subject, mostly with regard to its influence on honeybees at that time of the year when they are required for orchard pollination.

Hutson (1926) states that sunlight favourably affects flight but does not urge bees to leave the hive if other conditions are unfavourable. Phillips (1930) points out that sunshine is not necessary for flight, but cloudy weather tends to keep the bees near the hive by confining them to short flights. Marshall et al. (1929) state that bees are most active on bright, warm days. These statements are without doubt in conformity with general observations made by many beekeepers; they do not, however, allow any exact determination of the importance of light intensity on honeybee activity. Brittatn et al. (1933) made very much more detailed studies of the effect of the intensity of white light, and also various wave-lengths of light, on bee activity. They point out that although sunlight is an important factor in influencing the activity of bees, it alone will not cause them to work if the temperature is too low. They also state that an interesting effect of lack of sunlight can be observed on shaded limbs of apple trees, the flowers of which are not pollinated to anything like the same extent as those on limbs exposed to sunlight. Their counts made of the number of bees present on apple flowers at different times of the day show that, even at optimum temperatures, fewer bees were present in hazy weather, even without definite cloud banks, than when the sky was clear. They claimed to show that there is a general trend upward of bee activity, corresponding with increasing light values, and a corresponding decrease when light values normally fall off in the afternoon, or from the effects of clouds, haze or fog at any time during the day. It was further claimed that, within the temperature range of bee activity, light apparently has a more important influence than slight changes in temperature. They were unable, however, by this method of sampling (i.e. counting the number of bees present on the flowers) to secure any clear indication that any particular wave-length of light had any greater stimulative effect on bee activity than any other, even though it might have been expected from the work of Bertholf (1931,1931 a), who demonstrated the stimulative effect of ultra-violet light on insects, including the honeybee. One of Brittain’s collaborators, J. M. Cameron, however, making further observations on this subject during the honeyflow from Golden-rod by means of an electrical counter which determined the number of bees leaving and entering the hive during quarter-hour periods throughout the day, claims to have shown that the curve showing the average honeybee flight throughout the day follows more closely the changes in ultra-violet light values than white-light values; there being a higher correlation between ultra-violet light values and also White-light values and honeybee activity, than between temperatures and honeybee activity. He calculated simple correlations of the average temperature, intensity of clear light and intensity of ultra-violet light, with number of bees leaving a hive in successive half-hour periods throughout the day, the correlations he gives being for temperature 0·67, for intensity of clear light 0·84, and for the intensity of ultra-violet light 0·92. He suggests that all three conditions acting together probably decide the degree of activity as shown by the multiple correlation of 0·93 which he obtained and stated that the importance of ultra-violet light is further shown by the ‘beta values’ calculated. It is unfortunately not apparent what is meant by the latter, but they are probably partial regression coefficients, the units not being specified.

Unfortunately, when more closely examined the figures given by Brittain and also by Cameron are found to be not nearly so valuable as would appear at first sight. In the general discussion of the effect of light on bee activity during the apple flowering period it is shown that there is a general trend upwards of bee activity corresponding with increasing light values, and a corresponding decrease when light readings normally fall off in the afternoon. This is undoubtedly true but not very informative, since the amplitude of the main periodic variation is in general sufficiently large to mask the effect of minor irregularities, and the same result would hold if light intensity were replaced by any measurement having a diurnal cycle. This point is discussed in greater detail in the next section.

The irregularities in light intensity caused by the passage of clouds are of such frequency that it is necessary to take observations at short intervals in order to show them. The data presented by Brittain are inadequate, as only half-hourly readings are shown. It is also not clear from the test in what way the average bee counts were obtained. Each point is presumably the mean of several counts, since the values are not always integral, but there is no indication whether records of a number of different observers have been combined or whether a mere average of counts for adjacent time intervals has been made.

Visual examination of the four diagrams shown gives so little impression of association between deviations from trends that it does not seem worth while to attempt any analysis of the figures. On account of the long time intervals between readings the value of these records is slight, though the observations of 25 May 1932 do show a very pronounced drop in bee activity associated with the abnormally early fall in light intensity.

The figure given by Cameron representing data collected during the honeyflow from Golden-rod is probably a summary of much more valuable information. Apart, however, from statistical errors, the data are not well presented. The legend describes one curve as ‘number of bees ×10’; this is taken in the present paper to mean that the actual number of bees is 10 times the reading of the graph and not one-tenth of this, as the latter interpretation leads to unreasonably low counts. From Cameron’s statement that figures for periods earlier than 8.30 a.m. represent the records of 2 days only, it would appear that the graph is drawn from means of a number of day’s-observations for each interval, the number of days available possibly varying considerably for different hours of the day. It is stated that 415 observations in all were made, so that they probably extended over at least 10 days. Any such inequality of representation of the different days would be in every way undesirable, as the means for some periods might be seriously biased relative to others. It is questionable, therefore, whether any reliable conclusions may be drawn from the data as presented; this is the more unfortunate as the records may originally have been excellently suited to a study of light effects on bee activity.

On 5 days during the summer of 1940 (12,19, 26 June, 3 July and 25 September), on all of which temperature and other weather conditions were favourable for bee activity, observations were made throughout the day on the numbers of bees leaving a certain hive. From early morning until evening counts were made in successive min. periods of all bees leaving the hive. The only one of the 5 days for which the bees were not already active at the beginning of the observations and had ceased flying by the middle of the afternoon was 25 September. In order that only periods of bee activity may be considered, the observations of the early and late hours of this day have been omitted and the time studied reduced from 12 or hr. to In analysing these counts it seems preferable to work with proportional rather than with absolute changes, and the records have therefore been subjected to a logarithmic transformation. The effect of this change of unit is, for example, to attach equal importance to a doubling of the number of bees leaving the hive, no matter whether the increase is from 10 to 20 or from 100 to 200, and thus to simplify the comparison of changes in activity in colonies of different strength. In order to avoid difficulty with zero counts, a. convention is made that the number of bees is increased by one before transforming. The index of bee activity in any min. period is thus taken to be the logarithm (to base 10) of ‘one plus the number of bees leaving the hive in that period’.

This measure of activity for successive periods has been considered in regard to the intensity of solar radiation as measured by a Callendar electric recorder, situated at a distance of about 400 yards from the apiary. Though the curve representing the changes in light intensity throughout any one day possesses many irregularities, on account of the frequent passage of clouds, there is generally a well-marked diurnal cycle showing a maximum in the middle of the day. This was true of four of the observation days, but 26 June was exceptional in that there occurred a prolonged dull period from about 11.45 to 13.00 G.m.t. Now the rate of egress of bees from the hive shows a diurnal cycle of similar character, rising from a very small figure in the early morning to a maximum in the middle of the day and falling again in the evening. In one sense, then, there is no doubt of the existence of a positive association between light and bee activity. But this association is of little interest, as the same property would hold for any quantity—no matter how obviously irrelevant—which shows a similar diurnal cycle.

The real importance of the data lies in their answer to the question: ‘Are the deviations of the activity rates of the bees from the trend shown by the diurnal cycle associated with similar deviations in the light intensity?’ In order to test this and to estimate the magnitude of the effect, it is essential first to eliminate from both observations the component due to the cycle. Two methods might be adopted. The first is to fit polynomials of sufficiently high order to both records and to consider only the association of deviations from these polynomials. Tables for aiding this process are given by Fisher & Yates (1938, Table XXIII). It is not immediately clear what order of polynomial should be used in order to eliminate the trend satisfactorily, and more consistent results appear to be obtained with greater simplicity by eliminating differences between successive half-hour periods. This is simply accomplished by the technique of the analysis of variance and co-variance (Fisher, 1938), treating sets of four observation pairs as ‘blocks’ and forming components of variation within these blocks.

From this analysis may be computed, for each of the five days, a regression coefficient to represent the average increase in activity resulting from a unit increase in the rate of radiation at any time of the day. This regression coefficient leads to an estimation of the proportionate increase in the rate of egress of bees from the hive corresponding to a unit deviation of the rate of radiation from its trend line. The results are summarized in Table 1.

There is no doubt of the significance of the effect of the light intensity. There are considerable differences in the magnitude of the effect on the 5 days, but it is not apparent what the cause may be. The largest value of the regression coefficient occurs on 25 September when the temperature was low and there were very few bees flying. Conditions of activity at that time of the year are, however, scarcely comparable with those of June and July, and for the four earlier days there is some indication that the higher values of the regression coefficient are associated with the brighter days. On the other hand, the mean activity for the day appears to bear little relationship to the mean rate of radiation. Within any one day an average increase of about 15% in the number of bees leaving the hive may be expected when the radiation rate increases by 0·1 cal./sq. cm./min. Approximately 30 % of the variability of bee activity about its trend may be ascribed to irregularities in light intensity.

It would in theory be possible to make an exactly similar analysis of the counts, using temperature records instead of light intensities, or alternatively obtaining regressions on both observations. The temperature is, however, to the order of accuracy of its measurement, much too regular in its changes throughout the day, and the elimination of components ascribable to its diurnal cycle would leave little information for a study of its effects on bee activity.

In spite of the criticisms which have been made earlier of the presentation of Cameron’s data, it seemed of interest to attempt a new analysis of them by the technique used for the records now under discussion. For this purpose the possible inequalities in the presentation of days in the means were neglected, and the mean counts and light intensities read with two-figure accuracy from the graph. Between 8.30 a.m. and 6.15 p.m. thirty-nine pairs of records for successive quarter-hour periods were obtained, and the trends removed by eliminating differences between sets of three. Again a logarithmic transformation was employed.

The results of this analysis cannot, unfortunately, be put in a form directly comparable with those of Table 1, as Cameron does not state his units of light intensity. In terms of the scale units actually employed by Cameron for his light measurements, without correcting for the factor given in the legend, the partial regression coefficients of bee activity on clear and on ultra-violet light are 0 · 0072±0 · 0030 and –0 · 0020±100024. There is no indication that ultra-violet light has any significant effect other than the component due to its correlation with clear light, and the regression coefficient on clear light alone, 0 · 0049 ± 0 · 0014, sufficiently represents the situation. The mean intensities of clear and ultra-violet light over the periods studied were 48 · 2 and 54 · 3 and the mean index of activity of the bees 2 · 35 for a quarter-hour period. The mean number of bees leaving the hive was 239 per quarter hour. The regression coefficient shows an average increase of 1·1 % in the number of bees leaving the hive for a rise of one unit in the rate of radiation of clear light. The correlation coefficient between clear-light intensity and bee activity after the elimination of trends, 0 · 57, shows about 32% of the variations in activity to be accounted for by variations in radiation. The mean temperature rose from 67° F. at 8.30 a.m. to 73° F. at 1.00 p.m., and afterwards fell steadily to 65 ° F. at 6.15 p.m. As with the other data, its trend was too regular for information to be obtained on its effects on activity.

It seems, then, that Cameron’s conclusions as to the relative importance of clear and ultra-violet light in determining bee activity must be rejected. The reversal of the situation appearing in his figures is, of course, little dependent on the use of a logarithmic transformation of the data and almost entirely due to the elimination of the effects of the diurnal cycle. Unfortunately it is impossible to reconstruct and re-analyse the full 415 observations mentioned by Cameron, but, assuming that his means for different periods of the day have not been unduly distorted by non-orthogonal representation of the different days, it appears that deviations of clear-light intensity from its trend are directly reflected in the activity of the bees, but that ultra-violet light is of little importance in this respect apart from the correlation of its measurements with those of clear light.

We wish to express our thanks to Dr C. B. Williams and the members of the Bee Department for their interest and assistance in this investigation.