Monitoring of hive weight changes in various landscapes

Bee development is determined by the quantity and quality of food resources available in the environment that can be used by bees only to a certain extent. Therefore, this study aimed to evaluate, by means of changes in hive weight, the amount of food collected by honey bees in various landscapes, in relation to meteorological conditions, landscape, and forest structure. The monitored colonies (n = 104) were placed on electronic beehive scales, which provided data about hive weight, temperature, and humidity over two beekeeping seasons. The landscape valuation within a 2 km radius from each beehive was prepared based on digital maps. The seasonal distribution of food resources collected by honey bees was determined based on hive weight changes analysed in 5-day-periods (i.e. pentads). The hive weight changes mainly depended on the observation period and the air temperature. The hive weight increased from April to the last 10 days of June and when the air temperature was over 24 °C. The proportions of agricultural lands and forests, as well as the air humidity, had a small effect on hive weight changes. Continuous monitoring of the weight of honey bee colonies on a large scale showed that honey bees collected food stores in their nests mainly in the spring. Then, they collected a small amount of food, probably due to limited access to food resources from the last 10 days of June until the end of the growing season.


INTRODUCTION
The honey bee (Apis mellifera L.), one of the most important pollinators, is well-known mainly as a species managed by beekeepers. Increasingly, it is suggested that managed honey bees threaten wild bees, especially wild honey bees and bumble bees (Alaux et al. 2019;Requier et al. 2019). For wild honey bees, threats are mainly connected with the lack of nest sites, diseases, hybridisation, and restrictions on access to food resources, which is emphasised more and more (Fürst et al. 2014;Requier et al. 2019). The quantity and quality of food resources in the environment determine the diversity and abundance of wild bee species. Diminishing food resources in the environment is considered to be one of the major causes of the widespread decline in wild and managed honey bee populations (Goulson et al. 2015 The main reasons for the substantial reduction in plant diversity and abundance are landscape changes due to anthropogenic effects (e.g. agricultural intensification, urbanisation, habitat loss or fragmentation, deforestation; Goulson et al. 2015;Belsky and Joshi 2019;Grab et al. 2019;Neumüller et al. 2020).
The availability of food resources for bees mainly depends on land use. Arable lands, then forests, affect bees most strongly due to the area occupied. Access of bees to food sources is more unpredictable in arable areas compared to forest areas, where food resources are similar from year to year but can meet the nutritional needs of a limited population of bees (Donkersley et al. 2014(Donkersley et al. , 2017Odoux et al. 2014). Forests also provide better habitat conditions (e.g. nest and overwinter sites) and are less exposed to pesticides compared to arable areas (Donkersley 2019;Requier and Leonhardt 2020).
Bees feed on nectar and pollen collected from various plant species. In the environment, nectar and pollen resources are limited and allow the nutritional needs of a limited number of bees to be met. Introducing to the environment, beyond any control, commonly managed honey bees (A. mellifera) increase the density of bees per unit of surface area, which may disturb the nutritional balance among bees and lead to food competition (Hudewenz and Klein 2015;Lindström et al. 2016;Mallinger et al. 2017;Geldmann and González-Varo 2018;Wojcik et al. 2018;Alaux et al. 2019;Renner et al. 2021). Food competition is increasing as the diversity of plant species and the surface area occupied by plants decreases (Kennedy et al. 2013). Bees mainly compete over finite resources of pollen, which is a rich source of proteins critical for brood rearing (Lindström et al. 2016;Cane and Tepedino 2017;Geldmann and González-Varo 2018;Wojcik et al. 2018). A high density of managed honey bees in the environment can be especially dangerous for pollinators using the same food resources (Goulson and Sparrow 2009;Lindström et al. 2016).
The abundance and diversity of food sources available to bees change over the growing season. Seasonal changes mean that the composition of plant species does not always ensure the continuous availability of enough nectar and pollen to correspond to the actual nutritional needs of bees (Requier et al. 2015;Kaluza et al. 2016;Donkersley et al. 2017;DeGrandi-Hoffman et al. 2018). Long gaps in access to food may cause malnutrition, which can have a negative impact on bee development and health (Alaux et al. 2010;Di Pasquale et al. 2013).
The food resources in the environment can be evaluated in many ways, for example by monitoring land use, floral resource phenology, and the nectar or pollen productivity of flowers (Odoux et al. 2012(Odoux et al. , 2014Sponsler and Johnson 2015). However, these methods only provide information about the floral resources available to bees or what kind of potential food resources are present within the flight range of honey bees. They do not allow quantitative evaluation of how much of these resources are collected and used by bees. Such quantitative data can be collected on a large scale by means of continuous monitoring of the weight of honey bee colonies housed in hives placed on beehive scales (Steffan-Dewenter and Tscharntke 2000; Meikle et al. 2008Meikle et al. , 2016Meikle et al. , 2018Lecocq et al. 2015;Meikle and Holst 2015;Bayir and Albayrak 2016;Holst and Meikle 2018;Sponsler et al. 2020). Measurements of hive weight collected at the time of the day when honey bees do not forage can serve to monitor the daily change in weight, and consequently, the seasonal distribution of food resources stored by honey bees (Hambleton 1925). Quantitative evaluation of the food resources collected and used by honey bees during the entire beekeeping season in various landscapes can help to identify periods during which honey bees do not collect food in the hive. The identification of gaps in plant flowering will help to enrich the habitats of bees with plant species that ensure phenological continuity of flowering and fulfilling their nutritional needs. Such an approach should result in a greater diversity of plant species, and consequently, a greater diversity of wild bee species using the same food resources. Therefore, the aim of this study was to quantitatively evaluate, by means of changes in hive weight, the seasonal distribution of food resources collected by honey bees during the Page 3 of 14 30 whole beekeeping season in various landscapes with in relation to meteorological conditions, landscape, and forest structure.

Honey bee colonies
The study was conducted from 1 April to 30 September during two beekeeping seasons (2018 and 2019) in Poland. During these two seasons, the same 104 honey bee colonies (A. mellifera L.) were monitored ( Figure A1, Appendix A). The selected colonies were similar in size and health (i.e. medium colony strength). They were housed in double-body hives that remained in the same place throughout the research period. The apiaries were located in landscapes with different proportions of the forest, ranging from 23 to 91%, and consisted of a few to over a dozen colonies. The selected colonies were placed on electronic beehive scales with remote monitoring (LIVELCO sp. z o.o., Krakow, Poland). The distribution of the selected apiaries depended mainly on voluntary registration and access to the mobile phone network.

Data collection
Three times per day (6:00 a.m., 1:00 p.m., and 9:00 p.m.), the electronic beehive scales with remote monitoring collected the following data: hive weight, interior and exterior temperature, and exterior humidity. Once a month, the beekeepers also transferred information about all beekeeping tasks that might affect a hive weight (Appendix B), using a special application designed to run on a smartphone. All data were collected on a server of the Polish State Forests. Additionally, samples of workers were collected from each monitored colony to perform morphometric measurements to determine the subspecies of the managed honey bees.
Digital maps of the landscapes surrounding all monitored honey bee colonies were prepared to evaluate the areas within a 2 km radius of each hive. Based on these maps, the landscape structures were specified (proportion of arable land, green areas, grassland, urban areas, water reservoirs, and forest). In forests, the areas of different habitats were classified into six groups by fertility and humidity (lowland poor-mesic, lowland poor-wet, lowland rich-mesic, lowland rich-wet, upland rich-mesic, and mountainous rich-mesic),   as well as total forest stand area covered by tree and bush species considered to be attractive to bees. Tree and bush species were classified into three groups including the species strongly, moderately, and slightly affecting the daily changes of hive weight (Appendix C).

Data analysis
The daily change in hive weight was calculated by subtracting the hive weight measured on a given day and a given hour from the hive weight measured on a previous day at the same time. Unusually, large daily changes in hive weight caused by a beekeeper or rainfall were corrected by replacing them with a mean weight calculated on the basis of the hive weight collected on the previous day and on the next day at the same time. The lack of a single measurement of a hive weight caused by a malfunction of the electronic beehive scale (e.g. due to problems with access to the mobile phone network, heavy rainfall, or low battery in the scale) was corrected in the same way as described above. A malfunction of a beehive scale lasting more than 2 days was treated as a lack of measurements for a hive weight. Finally, based on daily changes of hive weights collected in the evening (9:00 p.m.) after the honey bees had returned to the hive and in the morning (6:00 a.m.) before the honey bees had started leaving the hive, a mean weight was calculated.
Descriptive statistics including mean, median, minimum, maximum, and interquartile range, were used to analyse daily changes in hive weight and the number of days with an increase or a decrease in hive weight. A comparison of the number of days with a daily change in the hive weight within a given range between 2018 and 2019 was conducted using one-way ANOVA. The distribution of the mean hive weight changes over 5-day observation periods (i.e. pentads), which was used in the further analysis, was prepared on the basis of daily changes in hive weight calculated from two measurements (i.e. 6:00 a.m. and 9:00 p.m.). Comparisons between the mean hive weight changes over pentads were calculated with the use of a non-parametric, twotailed Kruskal-Wallis (KW) test due to the lack of normal distribution of the data. As a result, the comparisons had to be performed separately for the two study years.
Boosted regression trees (BRT, Appendix D; Elith et al. 2008) were applied to determine the most important variables (associated with landscape structure, forest structure, and meteorological conditions) influencing seasonal changes in the weight of honey bee colonies. BRT models can select relevant variables, fit accurate functions, and automatically identify and model interactions. Each BRT model was defined with a tree complexity (number of nodes) of 7, a learning rate of 0.1, and a bag fraction of 0.7 (proportion of data selected at each step). To obtain robust results, the BRT prediction procedures, which are stochastic, were repeated 30 times, each time setting the random seed to provide a recurrence of the results. The random seed is a number used to initialise a pseudorandom number generator for a random component included in the process. Averaging of the predicted values was applied to obtain the final results.
The mean change of hive weight within a pentad was treated as a response (a dependent) variable. Predictors (explanatory variables) were included in five groups of factors: (1) periods and meteorological factors, (2) the landscape structure (see '2.2'), (3) the forest structure (see '2.2'), (4) the quality of data transferred by beekeepers, and (5) the subspecies of managed honey bees. The first group of periods and meteorological factors included the following: (a) the observation period (ranging from 1 April to 30 September, pentads), (b) the year (2018, 2019), (c) the mean air temperature within a pentad collected at 6:00 a.m., 1:00 p.m., and 9:00 p.m., respectively, (d) the standard deviation (SD) of the air temperature within a pentad collected at 6:00 a.m., 1:00 p.m., and 9:00 p.m., respectively, and (e) the standard deviation (SD) of the air humidity within a pentad.
During two beekeeping seasons (2018 and 2019), honey bee colonies were monitored for 182 days in each year. After data correction and rejection of any incorrect measurements of hive weights (based on information transferred by beekeepers), the mean number of days with measurements was 173 in 2018 and 172 in 2019, respectively. One electronic beehive scale was withdrawn due to problems with access to the mobile phone network and incomplete data in 2019. All statistical analyses (descriptive statistics, KW test, BRT analysis) were performed using Statistica, version 12.0 (Statsoft Inc. 2013).

RESULTS
In 2018, the mean number of days with an increase in the hive weight in the range of 0.30 to 0.99 kg and over 1 kg was significantly higher by about 6 and 7 days, respectively, compared with 2019 (Table I). Moreover, the mean number of days with a decrease in the hive weight below − 0.3 kg and in the range of − 0.29 to − 0.20 and − 0.19 to − 0.10 kg was significantly lower than in 2019 (Table I). The mean number of days during which daily changes of hive weight fluctuated within the range from − 0.09 to 0.09 kg was also significantly higher by about 15 days in 2018 compared to 2019 (Table I). An increase in the daily change of hive weight of more than 1 kg lasted no more than 7 and 2 days in 25% of monitored colonies according to the lower quartile and lasted more than 19 and 11 days in another 25% of monitored colonies according to the upper quartile, in 2018 and 2019, respectively ( Table I).
The amount of food collected in the hives by honey bees was significantly higher in 2018 compared to 2019 (KW test: H = 112.726, df = 35, n = 7479, p < 0.001). There were also significant differences in the amount of collected food between pentads in both beekeeping seasons (KW test: 2018, H = 1185.491, df = 35, n = 3686, p < 0.001; 2019, H = 851.872, df = 35, n = 3593, p < 0.001; Figure 1). The pentads with significantly greater increases in hive weight mainly occurred from April to June. Later, changes in hive weight were insignificant (Figure 1).
The relationship between the observed and estimated values (R 2 = 0.278; n = 7479, t = 53.9, p < 0.001) indicated that based on 30 predictors (Figure 2), it was possible to explain 28 % of the variation in change of hive weight during a pentad. The observation period and the air temperature (at 6:00 a.m., 1:00 p.m., and 9:00 p.m.) affected hive weight changes most strongly, because their relative importance exceeded a value of 0.9 (Figure 2). Other predictors explaining variation in hive weight changes also belonged to meteorological features: the SD of the air temperature (collected at 6:00 a.m., 1:00 p.m., and 9:00 p.m.), the mean air humidity, and the SD of the air humidity within a pentad. Their relative importance ranged from 0.6 to 0.8 (Figure 2). The effect of the proportion of agricultural lands and forests was similar (the relative importance was at level 0.6; Figure 2). The effects of other features, including the area of forest stands classified by the presence of tree or bush species variously attractive to bees, area of other forms of land use, types of forest habitats, quality of the data transferred by beekeepers, and year, were low because their relative importance does not exceed a value of 0.6 ( Figure 2). When the same mean air temperature was assumed in each analysed term (using computer simulation), the course of hive weight changes becomes similar for both years (2018 and 2019; Figure 3). The effect of air temperature on hive weight changes varied depending on the observation period (Figure 4) in 2018. The hive weight increased when the air temperature was higher than 24 °C at 1:00 p.m.; below this temperature, the hive weight was decreasing. The strongest increase in hive weight was noted when the air temperature was about 28 °C at 1:00 p.m.
The effect of land use structure on hive weight changes concerned the following trends: a smaller increase in hive weight when the proportion of arable areas was larger and a greater increase in hive weight when the proportion of forest was larger ( Figure 5). The effects of the type of forest habitat and plant species attractive to bees, as well as subspecies of managed honey bees, were also small (their relative importance was below a value of 0.6; Figure 2). Hive weight change (kg/pentad)

Observation period (months)
year 2018 year 2019 Figure 3. The mean change of hive weight within a pentad depending on observation period in 2018 (assuming that the air temperature at 1:00 p.m. was 22.6 °C) and 2019 (assuming that the air temperature at 1:00 p.m. was 20.8 °C).

DISCUSSION
The results of the present study show that the observation period and the air temperature most strongly influenced hive weight changes. These changes mostly depended on the mean air temperature in the spring and the amount of food collected in the hives due to the limited availability of food resources in the summer. The use of food resources was differentiated among the honey bee colonies monitored and reflected the environmental conditions of the local landscapes surrounding the hives. Moreover, measuring daily changes in hive weight using electronic beehive scales proved to be, to our knowledge for the first time, effective in remote monitoring of large-scale collection and use of food resources by honey bee colonies inhabiting various landscapes during the whole beekeeping season, without disturbing them.
The distribution of hive weight changes and the number of days when the daily changes in hive weight were greater than zero kilograms show that the nutritional needs of honey bees could be met in the spring. However, the low number of days when hive weight changes were greater than 1 kg indicates that the opportunities for honey bees to collect larger amounts of food were limited. There are a number of possible reasons for these limitations in the food flow, including adverse weather conditions for foraging, too small an area covered by flowering plants, too great a distance from the hives to food sources, an inadequate structure of the honey bee colony consisting of too few active foragers, or too many foraging bees competing with each other for finite food resources (Seeley 1995;Lindström et al. 2016;Cane and Tepedino 2017;Geldmann and González-Varo 2018;Alaux et al. 2019;Renner et al. 2021). However, our results clearly indicate that honey bees were less exposed to long gaps in collecting food in the spring as soon as the weather conditions were favourable for foraging. The hive weight increased when the air temperature was higher than 24 °C at 1:00 p.m., and the effect of temperature on food collection by honey bees was the strongest at 28 °C at 1:00 p.m., which was probably additionally connected with the increased requirement of honey bee colonies for water needed to evaporative cooling of the broodnest (Winston 1991;Seeley 1995). However, during periods of adverse weather conditions, honey bees started eating their food stores. It should be noted that food stores collected by honey bees were sufficient only for a short time, and when the weather conditions were worse in May 2019 (low temperatures and prolonged rainfall), beekeepers reported colony feeding.
The results also indicate that honey bees were vulnerable to starvation caused by their inability to collect more food stores in the hive and an insufficient amount of food in the environment starting in the last 10 days of June. It is likely that honey bees did not find enough food in your environment. In July and in August, the decrease in hive weight was not balanced by the amount of nectar and pollen collected by foragers in the hive. Despite favourable weather conditions and a large number of foragers, honey bees were forced to use the colony's food stores most likely due to deficiencies in food resources in the surrounding area. In the last decade of July, a honey bee colony starts rearing a generation of workers, which need to survive the winter. Shortages of food resources starting in the last 10 days of June to the end of the growing season do not promote the development of the honey bee colony because foragers put more effort into finding food sources, and as a result, they may reduce the amount of reared brood (Mattila and Otis 2007). In July and in August, long gaps in the flow of food into the nest may cause malnutrition of honey bees, and consequently, health problems including deterioration of immune function or higher vulnerability to pathogens and parasites (Alaux et al. 2010;Di Pasquale et al. 2013). As a result, the honey bee colony may not survive the winter.
In practice, estimating threshold values of colony weight changes indicating an imbalance between the supply and demand of food can be helpful in assessing the nutritional potential of local landscapes. Hive weight changes in the range of 0.0 to 0.3 kg are usually associated with the functioning of the honey bee colony and mainly result from changes in the amount of reared brood and adult bees, the comb weight connected with its construction, capping of the cells, and the accumulation of cocoons in the cells, the amount of propolis and water brought to the hive (Seeley 1995). An increase in hive weight changes above 0.3 kg usually results from the gain of honey and beebread stores and is disproportionally high compared with changes resulting from the functioning of the honey bee colony (Winston 1991;Seeley 1995). Therefore, it can be expected that hive weight changes under 0.3 kg occurring for long periods of time are the first sign of a limited flow of food to the nest. This may indicate that the balance between the number of bees and the availability of food resources in the environment is disturbed. Based on data obtained from continuous monitoring of hive weight, the number of managed honey bee colonies can be controlled and limited if necessary or a given area can be enriched with flowering plants important to bees. An increase in the proportion of areas covered by floral resources can ensure a continuous flowering phenology particularly late in the growing season, and it can reverse some of the negative effects caused by current land use and its composition (Odoux et al. 2012;Requier et al. 2015;Dolezal et al. 2019;Clair et al. 2022).
A marked decrease in food resources collected by honey bees was observed starting in the last 10 days of June in both years and in all monitored landscapes irrespective of their structure, proportion of different habitats, or dominant plant species, and irrespective of differences in temperature distribution, which affects the floral resource phenology. This indicates that food deficiencies in the environment are common. The development of bees at the individual, colony, and population levels is severely constrained and even stopped when nectar and pollen are unavailable or only partially accessible. Nectar deficiency, a source of carbohydrates, limits the flight activity of bees needed to find and collect food. In turn, pollen shortages, a rich source of nutrients (proteins, fats, vitamins, and minerals), can disturb individual bee development and brood rearing (Brodschneider and Crailsheim 2010;Wright et al. 2018). It should be noted that honey bees store a limited amount of pollen in the nest, under 1 kg (Jeffree and Allen 1957;Camazine 1993). Therefore, continuous monitoring of hive weight can help in a simple way, and on a large scale, to quantitatively determine the food resources collected by bees. This knowledge can be useful for enriching and diversifying plant communities to ensure a continuous flowering phenology during the whole growing season, particularly in landscapes changed due to anthropogenic effects.
The flight activity of honey bee foragers depends on the temperature and increases along with an increase of temperature in the range of 15 to 25 °C (Abou-Shaara et al. 2017). The results of the present study indicate that temperature strongly affected changes in hive weight, but a significant impact on hive weight was seen at an air temperature of 28 °C (at 1:00 p.m.). However, despite the optimal temperature, not all monitored colonies collected food. The mean change in hive weight during a pentad was subjected to a random factor and/or probably was affected by presently unknown features that were not included in this analysis (e.g. individual features of each honey bee colony, other meteorological or landscape factors). Surprisingly, land use only slightly affected hive weight changes which can be explained by the flowering time of crops overlapping with the flowering time of wild plants in the spring. In the summer, the proportion of flowering crops is small, whereas forests can be a source of honeydew. It is also likely that an insignificant influence of land use on hive weight changes resulted from properties in these areas which were not typical agricultural lands because more than half of the monitored colonies were located in places where agricultural lands do not cover more than 30% of the area. Moreover, the season, but not landscape diversity, shaped the amount and diversity of food collected by honey bees Danner et al. 2017;Sponsler et al. 2017;Malagnini et al. 2022). It is possible that more attention should be paid to other factors such as solar radiation which along with temperature most strongly affected the foraging activity of honey bees (Clarke and Robert 2018). It is also likely that some colonies were weak due to having too few, less active foragers, swarming, or infestations by Varroa destructor mites or Nosema ceranae microsporidia (a gut parasite). Honey bee colonies with access to adequate amounts of high-quality pollen have lower pathogen loads and are less vulnerable to the negative effects of infestation than those that are poorly nourished (Di Pasquale et al. 2013;Dolezal and Toth 2018). Therefore, continuous monitoring of hive weight can be used to assess the nutritional state of the honey bee colony, and indirectly its health status.
It is worth noting that managed honey bees as floral generalists can be used as an indicator of food availability in the environment for wild bee species because some of them visit the same plant species as honey bees to collect nectar and/ or pollen, particularly bumble bees (Goulson and Sparrow 2009;Martins et al. 2018). Moreover, collectively foraging honey bees continuously search for food source patches within a vast area (1-3 km) around their nest (Visscher and Seeley 1982;Seeley et al. 1991). Therefore, the number of managed honey bee colonies should be strictly controlled in the environment in order not to lead to food competition with wild bees (Hudewenz and Klein 2015;Lindström et al. 2016;Cane and Tepedino 2017;Mallinger et al. 2017;Geldmann and González-Varo 2018;Wojcik et al. 2018;Alaux et al. 2019;Renner et al. 2021). Based on the results of the present study, it can be expected that the nutritional needs of wild honey bees and solitary bees will be met in the spring, but they can be exposed to a substantial food shortage in July and August, just as managed honey bees. Solitary bee species, which do not collect food stores, could be at risk of starvation, especially late spring species ending their life cycle in July and summer species preparing to overwinter (Michener 2007;Leach and Drummond 2018). Furthermore, solitary bees usually forage close to their nests, at a distance of no more than 300 m; only a few of them are able to cover greater (more than 1000 m) foraging distances (Greenleaf et al. 2007;Zurbuchen et al. 2010). Increased foraging distances forced by the lack of food result in a higher cost of foraging flights and a lower number of provisioned brood cells in solitary bees (Zurbuchen et al. 2010).

CONCLUSIONS
Continuous monitoring of hive weight provided important data about the amount of food collected by honey bees in the hive, and indirectly, about the availability of food resources in the environment and their seasonal distribution. This method applied on a large scale allowed us to study different environmental and in-hive factors affecting the development of honey bee colonies and to determine if the balance between supply and demand of food is maintained. Future research should focus more on the evaluation of food resource distribution in the environment during the growing season in order to restore continuity in plant flowering and to meet the nutritional needs of bees.
Restoring the biodiversity of forest habitats may promote the conservation of wild bees without limiting the number of managed honey bee colonies.
Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.