Effect of pollen provision on life‐history parameters of phytoseiid predators under hot and dry environmental conditions

Biological control can be severely disrupted under climate change conditions. This is the case of the spider mite Tetranychus urticae in Spanish citrus orchards, where the omnivorous phytoseiid Euseius stipulatus, the most abundant predator in the system, was highly impacted by hot and dry conditions mimicking future warmer summers. Such a situation can often be compensated by the provision of alternative food to support generalist predators. As a first step to studying whether such a technique could be applied in this case, we studied at laboratory conditions whether pollen could mitigate the negative effects of hotter and drier conditions derived of climate change on three phytoseiids with different diet specializations. In addition to E. stipulatus, these predators, which all together, are considered key for the biological control of T. urticae in citrus, are Neoseiulus californicus and Phytoseiulus persimilis. Our results confirm the extremely fine‐tuning of T. urticae to hot–dry conditions. They also provide evidence of the poor performance of E. stipulatus, especially in terms of reproduction, compared to the other two phytoseiids at these conditions, even when high‐quality pollen was available. Moreover, access to pollen in combination with T. urticae eggs enhanced survival but reduced predation and oviposition relative to a T. urticae‐only diet for N. californicus and P. persimilis. Therefore, whether the overall effect of pollen would justify its use in citrus to counteract the deleterious effects of a hotter and drier climate on the natural regulation of T. urticae is still controversial.

In the Mediterranean basin, the two-spotted spider mite, Tetranychus urticae Koch, is a key pest of clementine mandarins, Citrus clementina Tanaka (Rutaceae) (Aguilar-Fenollosa et al., 2011;Martínez-Ferrer et al., 2006;Pascual-Ruiz et al., 2014). Its main natural enemies are different phytoseiid predatory mite species (Mesostigmata: Phytoseiidae), which are present in this system and have different diet specializations (McMurtry & Croft, 1997;McMurtry et al., 2013). The most abundant phytoseiid in Spanish citrus orchards, irrespective of the citrus cultivar and management practices used, is the omnivorous Euseius stipulatus (Athias-Henriot; Abad-Moyano et al., 2009a;Aguilar-Fenollosa et al., 2011b;Vela et al., 2017). However, this phytoseiid is not the most effective predator of T. urticae. This role is played by the T. urticae-specialist Phytoseiulus persimilis (Athias-Henriot), which preys on this herbivore almost five times more frequently than E. stipulatus (Pérez-Sayas et al., 2015). The Tetranychidae-specialist Neoseiulus californicus (McGregor) is also commonly found in these citrus orchards (Abad-Moyano, Pina, Dembilio, et al., 2009;Aguilar-Fenollosa et al., 2011b;Vela et al., 2017). These specialists are consistently found in clementine orchards grown in association with a grass cover, where the abundance of E. stipulatus relative to other ground covers diminishes (Aguilar-Fenollosa et al., 2011c), resulting in enhanced biological control of T. urticae (Aguilar-Fenollosa et al., 2011b). Urbaneja-Bernat et al., (2019) showed under semi-field conditions representative of hotter and drier environmental conditions in the Mediterranean basin that the regulation of T. urticae in clementine trees provided by E. stipulatus, N. californicus and P. persimilis could be seriously disrupted. The dynamics of T. urticae in simple trophic chain modules (Bascompte & Melián, 2005) including these predators were species-specific and did not follow the same patterns in spring and summer. This study showed that these predators provided similar control levels of T. urticae when released singly in conditions mimicking spring climate change conditions. Although it is generally acknowledged that species with the highest specializations in lifestyle or habitat are typically most threatened by climate change (Aguilar-Fenollosa & Jacas, 2014), the omnivorous E. stipulatus provided no control at hotter and drier summer conditions representative of climate change, whereas the other two prey-specialized species were even more effective in summer than in spring.
These unexpected results suggest that future warmer and drier summers could result in a deficient control of T. urticae in citrus orchards because of the high impact on most abundant E. stipulatus.
However, this could be compensated by (a) a better performance of less abundant but more efficient T. urticae-specialists P. persimilis and N. californicus, which could reverse the situation (Urbaneja-Bernat et al., 2019) and/or (b) the addition of supplementary food to the system, as the importance of such a supply to support generalist predatory mite populations, like E. stipulatus, has been widely recognized (González-Fernández et al., 2009;Janssen & Sabelis, 2015;Khanamani et al., 2017;Maoz et al., 2011;McMurtry et al., 2013;Pozzebon et al., 2009). Indeed, E. stipulatus and N. californicus can persist in citrus when T. urticae is scarce, feeding on other food sources including pollen (Pina et al., 2012). Moreover, Beltrà et al. (2017) demonstrated that the provisioning of pollen and sugars in Spanish citrus orchards could boost phytoseiid natural populations in spring and fall. However, this supply had no effect from June to September. Therefore, there are doubts on whether pollen supply could be an effective measure to mitigate the effects of climate change in this system.
To challenge the hypotheses that (a) the specialist predators N. californicus and E. stipulatus can do better than the generalist omnivore E. stipulatus at hotter and drier conditions and (b) pollen supply can compensate the adverse effects of these conditions on these natural enemies, we performed a series of short-term experiments under laboratory conditions. This type of assays, which are commonly used to assess the effect of extreme climatic events such as heatwaves (Bannerman et al., 2011;Ciais et al., 2005;De Boeck et al., 2010;Gillespie et al., 2012;Jentsch et al., 2007;Sentis et al., 2013;Smith, 2011), allowed us to explore how different combinations of temperature ('T') and relative humidity ('RH'), including those typical of hotter and drier abiotic conditions associated with climate change, affect the key biological parameters (i.e. survival, oviposition and predation) of T. urticae and its predators E. stipulatus, N. californicus and P. persimilis with and without provision of pollen as a supplementary food. The results of this work should help to explain the semi-field results observed (Urbaneja-Bernat et al., 2019) and provide evidence of whether pollen supply could be a tactic allowing the conservation of these natural enemies in a rapidly approaching warmer future.

| Plant material
Two-year-old clementine plants (Citrus clementina Tanaka cv.
Clementina de Nules (Rutaceae) grafted on citrange Carrizo) were used as a source of leaves for the assays. Fifty days before the beginning of each assay, 25 plants were defoliated and kept in a greenhouse at Universitat Jaume I (UTM: 39° 59'10.883 "N 0°3'4.769"W) set at 22 ± 2°C, 55 ± 10% relative humidity, and natural photoperiod. These plants were grown on vermiculite and peat (1:3; vol: vol) in 320 ml pots, were fertilized twice per week using a modified Hoagland's solution (Bañuls et al., 1997) and received no pesticide treatments. When necessary for the rearing of mites, bean leaves (Phaseolus vulgaris L. (Fabaceae)), lemon fruits (Citrus lemon Burm. f. (Rutaceae)) and Carpobrotus edulis (L.) (Aizoaceae) pollen (dried at 37°C, sieved and frozen until use) obtained from pesticide-free plants were used. This pollen is considered as high quality for phytoseiid mites. At laboratory conditions, it can sustain and even boost populations of N. californicus and E. stipulatus, respectively (Pina et al., 2012).

| Mite stock colonies
Four different mite species were used in our studies: the two-spotted spider mite T. urticae, and the Phytoseiidae E. stipulatus, N. californicus, and P. persimilis. These colonies were maintained in separate climatic chambers set at 25 ± 1°C, 65 ± 5% 'RH' and a 16-hr light photoperiod.
Spider mites were collected in a Clementina de Nules orchard at Les Alqueries (UTM: 39°59'15.1"N 0°3'02.0"W) in 2010. This colony has been maintained ever since using standard procedures on detached leaves of clementine mandarins (Aguilar-Fenollosa et al., 2012) and, in some cases (see below), on pesticide-free lemon fruits (Abad-Moyano et al., 2010). Spider mites were used to either feed the Phytoseiidae stock colonies or to start new cohorts for our assays. When used to feed phytoseiids, bean leaflets were infested by exposure to lemon fruit colonies. New cohorts were established by transferring 100 females to new rearing arenas on clementine leaves. Females were removed one day later, and these units containing less than 24-hr old eggs were held separately in a climatic chamber (25 ± 1°C, 65 ± 5% 'RH') and constituted the cohorts used in our assays.

Individuals of N. californicus were obtained from Koppert
Biological Systems (SPICAL®) to initiate a laboratory colony.
Phytoseiid stock colonies were maintained on detached leaf arenas.
These arenas consisted of single bean leaflets placed upside down on moistened filter paper placed on top of a water-saturated foam cube (3-4 cm thick) in an open plastic box half-filled with water.
Phytoseiid colonies received twice a week detached bean leaflets infested with T. urticae and C. edulis pollen as food.

| Experimental arenas
Arenas consisted of a petri dish (5 cm in diameter) with a 3 cm in diameter hole in the cover. The base of the dish was filled with bacteriological agar (2.5% weight). As soon as agar was cold and solid enough, a fully expanded clementine leaf was placed upside down on top of the agar to maintain its turgor. The cover was subsequently put in place so that the leaf substrate formed a 3 cm in diameter exposed area. The upper and lower parts of the dishes were sealed with a strip Parafilm ® (Pechiney Plastic Packaging, Menasha, WI, USA). Finally, to prevent mite escape from the arena, permanent glue (Tree Tanglefoot ® ; Grand Rapids, MI, USA) was applied along the rim of the cover hole (Guzmán et al., 2016).

| Effect of temperature and relative humidity on T. urticae performance: survival and oviposition
Less than 24 hr old presumably mated females (i.e. those reaching the adult stage immediately after the quiescent teliochrysalis stage) were selected and individually moved into a clean experimental arena. Survival (i.e. alive, dead specimens and escapees) and oviposition (number of eggs laid during the experiment) were assessed under a binocular microscope 24 hr after the onset of the assay.
Different 'T' (10-40°C in 5°C steps) and 'RH' values (30, 50 and 70%) were combined in our assays. Constant 'RH' values were obtained by using different salt solutions (Winston & Bates, 1960) in desiccators kept inside environmental chambers (Sanyo Electric Co., Ltd., Japan) set at a photoperiod of 16:8 hr L:D and the different target temperatures. We performed five replicates of six arenas per environmental condition (i.e. a total of 30 replicates per environmental condition).

| Effect of temperature and relative humidity on phytoseiid performance: survival, predation and oviposition
A fully expanded healthy clementine leaf was introduced onto a T. urticae-infested lemon stock colony. Twenty-four hours later, the infested leaf was moved into a phytoseiid colony and left there for an additional 24-hr period. Then, leaves were inspected under a binocular microscope to remove all motile stages. A separate phytoseiid colony was started with every single leaf, and they constituted the cohorts used in our assays. As this method did not work for E. stipulatus, the eggs of this species were obtained by exposing a few cotton threads to an existing colony. Twenty-four hours later, all motile forms on these threads were removed, and the remaining eggs were used to start a new cohort. Phytoseiids were reared up to the adult stage following the same procedure as for the stock colonies.
In our assays, we used gravid adult phytoseiid females at their peak oviposition rate (12-14 days from egg hatching; Aucejo-Romero et al., 2004;Janssen & Sabelis, 1992). To ensure the same level of starvation in all females tested, these were randomly selected from a cohort and individualized in plastic arenas (same as for the stock colonies but substituting the plant material by a plastic board) placed on top of a sponge in a water-containing tray where they starved for 24 hr. The edges of these plastic boards were covered with tissue paper in contact with the sponge and the water, which served as both a barrier and a water source for mites.
Experimental arenas received 15 T. urticae females, which fed, laid eggs, and produced a web for 48 hr. At that time, we removed all mobile forms of T. urticae, and only ≤ 48 hr of old eggs were left.
The mean number of T. urticae eggs per arena was ∼75. Immediately after, a starved phytoseiid female was introduced into the arena.
These units were then transferred to a desiccator, where the desired 'RH' was achieved as above. Likewise, these desiccators were introduced into an environmental chamber set at the target temperature (same 'T' and 'RH' combinations as for T. urticae). Arenas were checked 24 hr after the onset of the assay (i.e. 48 hr after the onset of the starvation period for adult phytoseiid females) under a binocular microscope. This period was selected because T. urticae eggs used in the arenas could start hatching in 48 hr, especially at high temperatures (>30°C). Survival, oviposition and predation (number of T. urticae eggs eaten) were scored. Same as with T. urticae, we performed five replicates of six arenas per environmental condition and mite species.

| Effect of alternative food on phytoseiid performance: survival, predation and oviposition
In addition to arenas containing ≤ 48 hr old T. urticae eggs, two more diets were considered: (a) pollen of C. edulis, and (b) a combination of the former two. Arenas containing ≤ 48 hr of old eggs were obtained as before. Arenas containing pollen were prepared by adding C. edulis pollen ad libitum in a single point in the centre of the arena. As soon as the arenas were ready, one starved phytoseiid female was introduced. As before, the arenas were checked 24 hr later when survival, oviposition and predation were scored. In this assay, the combination of three 'T' (15, 25 and 30°C) and three 'RH' (30, 50 and 70%) was considered. We performed a total of 15 replicates per environmental condition, diet and phytoseiid species.

| Statistical methods
To study the effects of the 'T' and 'RH' on T. urticae and phytoseiid performance, we used general linear models (GLM) and separately analysed T. urticae and phytoseiids. In the case of survival, which had three different possible outputs (i.e. live and dead specimens and escapees), we used a GLM with a multinomial distribution of the error and a generalized logit link function. For predation (only for phytoseiids) and oviposition, we used a GLM with a Poisson distribution of the error and a logistic link function. The factors 'species', 'T' and 'RH' were used as fixed effects in all cases. As one of our main goals was to identify phytoseiid species-specific differences, in the case of predators, we started our analyses by considering all combinations, including 'species' as a factor. Once the statistical significance of the 'species' factor was clear, we similarly continued the analyses of survival, predation and oviposition by studying the effect of 'T', 'RH' and their interaction. We included the factor 'replicate' (1-5) as a random factor.
To study the effects of alternative food on the performance of phytoseiids, we used the same general linear models (GLM) as above.
For survival, oviposition and predation, the factors 'species', 'diet', 'T' and 'RH' were used as fixed factors. As our main goal was to identify species-specific patterns of response, same as above, we started our analyses by considering all combinations, including the 'species' factor. Then, we studied the effect of 'diet'. Eventually, we separately analysed for each species and diet, the effects of 'T', 'RH', and their interaction. The factor 'replicate' (1-3) was included in our analyses as a random factor. In both cases, when necessary, we used the Bonferroni post hoc test for mean separation at p < .05. All data were analysed using SPSS 23.0 software.

| Effect of temperature and relative humidity on T. urticae performance: survival and oviposition
The factor 'T' and the interaction 'T'*'RH' significantly affected survival and oviposition (Table 1). The absolute highest survival (i.e. the percentage of live specimens) was observed at 25°C and 30% relative humidity (100% survival) (Figure 1a). Survival was always above 60%, even at the extreme temperatures tested (10 and 40°C) during the experiment.

| Effect of temperature and relative humidity on phytoseiid performance: survival, predation and oviposition when preying on T. urticae
We observed significant differences (p < .05) between phytoseiid species for all parameters considered (survival, predation and oviposition; Table S1). Consequently, we analysed the effect of 'T' and 'RH' for each phytoseiid species separately.
The GLM to analyse the survival 24 hr after the onset of the assay for the three predators ( Figure 2) included 'T' and 'RH'. For E. stipulatus, 'T' and its interaction with 'RH' were significant (Table 2). This species could not survive temperatures above 30°C. Below this threshold, survival usually increased with 'RH'. However, the percentage of escapees TA B L E 1 Statistics (Wald χ 2 ; df; p-value) of the different GLM models adjusted to survival and oviposition for T. urticae considering the factors temperature (T) and relative humidity (RH) as well as their interaction as explanatory variables  Figure 2).
The number of eggs preyed was significantly affected by 'T', 'RH' and their interaction for the three phytoseiid species (Table 2). In the case of E. stipulatus, the lowest predation rates were observed at 10°C irrespective of relative humidity ( Figure 2). Above this temperature and up to 25°C, predation increased, and maximum rates were usually associated with 70% relative humidity. A maximum of 15.6 ± 1.9 eggs eaten per female was observed at 15 and 25°C at this relative humidity. Beyond 30°C, there was no survival and so no predation was observed. Predation rates for N. californicus ( Figure 2) were minimal at 10°C irrespective of relative humidity. Above this temperature, they increased up to 25-30°C, then decreased at 35°C and were zero at 40°C because of no survival at this temperature.
Interestingly, at 15 and 35°C, predation was maximal at 50% relative humidity, whereas, at 25 and 30°C, the highest predation rates were associated with the highest relative humidity values tested, with a mean of 21.6 ± 1.2 T. urticae eggs eaten per female. Phytoseiulus persimilis was the most voracious mite at any of the temperatures and relative humidity combinations tested and presented a trend closely matching what we observed for N. californcius (Figure 2). In this case, maximum predation rates were reached at 30°C independent of relative humidity and at 25°C with 70% relative humidity with a mean of 39.3 ± 2.5 eggs per female. The number of eggs consumed per female decreased dramatically to 16.3 ± 2.5 eggs at 35°C, but these values were still higher than those observed at 10°C.
During the first 24 hr of the assay, the number of eggs laid was affected both by 'T' and 'RH' in N. californicus, by 'T' and the interaction of this factor with 'RH' for P. persimilis, and it was independent of these factors for E. stipulatus (Table 2). This independence could be attributed to the meagre oviposition rates observed for this phytoseiid at all combinations tested (0 to 0.2 eggs per female and day; F I G U R E 1 Fraction of T. urticae females (a) (stuck in the glue, dead or alive), and oviposition (b) (number of eggs laid per female) when exposed to temperatures in the range 10 to 40°C in combination with 30 (black bars), 50 (grey bars) and 70% (white bars) relative humidity values during the first 24-hr periods after the onset of the assay. For and oviposition, bars with the same letter are not statistically different (Bonferroni p < .05) Figure 2). The oviposition rate of N. californicus ( Figure 2) increased from about 0.1 to 2.5 eggs per female between 15 and 30°C with the absolute maximum number of eggs laid per female at 30°C and 50% relative humidity (3.2 ± 0.1 eggs). Below 20 and above 30°C, oviposition was minimal, and at 10°C, only a few eggs could be collected in the arenas kept at 70% relative humidity. Intriguingly, oviposition at 20°C and 70% relative humidity was as low as the reported minimum values. A similar trend was observed for the response of P. persimilis to temperature (Figure 2). In this case, oviposition increased from about 0.7 to around 4.1 eggs per female between 10 and 30°C.
However, in this case, the effect of relative humidity changed direc-

| Effect of alternative food on phytoseiid performance: survival, predation and oviposition
There were significant differences between the three phytoseiids for survival (Table S2). Consequently, we further analysed the influence of the factor 'diet' for each species separately (Table S3). We found that this factor was significant (p < .001) in all cases. Consequently, these results led us to eventually analyse the influence of 'T', 'RH', and their interaction for each mite species and diet separately.  In the case of the tetranychid-specialist predator N. californicus, survival was higher when the mite had access to T. urticae F I G U R E 2 Survival (stuck in the glue (grey bars), dead (black bars) and alive (white bars)), predation (number of T. urticae eggs eaten per female), and oviposition (number of eggs laid per female) for Euseius stipulatus, Neoseiulus californicus and Phytoseiulus persimilis when exposed to temperatures in the range 10 to 40°C in combination with 30 (black bars), 50 (grey bars) and 70% (white bars) relative humidity values during the first 24-hr period after the onset of the assay. For each phytoseiid species, predation and oviposition bars with the same letter are not statistically different (Bonferroni p < .05) eggs, either alone (86.5%) or combined with pollen (85.9%), and decreased when pollen was the only food source available (31.6%) ( Figure 3). However, when we analysed the survival for each diet, we observed that the GLM model provided a good fit (p < .05) only in the case of pollen alone ( and 'RH' in both diets (Table 4). The highest predation rates were observed when E. stipulatus was offered a mixture of pollen and T. urticae eggs (average of 11.4 ± 1.4 eggs eaten versus 7.9 ± 1.6 for the T. urticae eggs only diet) (Figure 4). However, we observed similar predation rates at 15 and 25°C and 70% relative humidity when T. urticae eggs constituted the only food source available. The predation rate of N. californicus also was significantly affected by 'T' and 'RH' for the two diets considered (Table 4). In general, highest predation rates were observed at 30°C irrespective of relative humidity and, within each temperature, at 70% relative humidity ( Figure 4). The number of eggs preyed by P. persimilis (Figure 4) was affected by 'T', 'RH' their interaction in both diets (Table 4). Pollen provision affected predation rates because the number of eggs preyed decreased when offered a mixed diet (20.6 ± 1.6) compared with T. urticae eggs only (28.5 ± 2.1). For both diets, predation significantly decreased at 15°C independently of relative humidity, and there were no differences between 25 and 30°C.
The number of eggs laid by E. stipulatus at the different temperature and relative humidity combinations was significantly different in diets with either T. urticae or pollen alone (Table 5).
However, when we offered a mixture of T. urticae eggs and pollen, these differences disappeared (Table 5). These results should probably be attributed to this species' extremely low fecundity in our assays, which was always below one egg per female and day ( Figure 5).
For N. californicus, oviposition was affected by 'T' and 'RH' (

| D ISCUSS I ON
We aimed to determine whether, in agreement with previous semi-  values observed for N. californicus and P. persimilis, especially for predation and oviposition. Because T. urticae could survive temperatures of 40°C and even achieve maximum oviposition rates at 35°C (Figure 1), a temperature that only N. californicus and P. persimilis could survive with limited reproduction (Figure 2), our results confirm that T. urticae outbreaks in citrus could become increasingly more frequent in the future.
We hypothesized that climate change could differentially impact second and third trophic levels of the mite community established around T. urticae in clementines. Our results demonstrate that the different parameters studied (survival, oviposition and predation) depend on both temperature and relative humidity and are affected by the available food source. Moreover, the magnitude of the impact was species-specific. Mean temperatures above 25°C, which can be taken as a proxy of summer climate change conditions in Spanish citrus-growing areas (Urbaneja-Bernat et al., 2019), were more detrimental to phytoseiids than to T. urticae, which presents maximum survival and oviposition at temperatures about 5°C higher than bestadapted phytoseiids, independently of relative humidity (Figure 1).
Probably because of the experimental setup used, where escapees could not seek shelter but instead died stuck in the glue, the highest impact of hot and dry conditions among phytoseiids was observed for the omnivore E. stipulatus. Theory would not predict, the highest impact of climate change on generalist instead of specialist predators, but this is what we were expecting based on previous semi-field assays (Urbaneja-Bernat et al., 2019). The survival of T. urticae in our experimental conditions was always above 60%, and it even reached 100% at 25°C and 30% relative humidity (Figure 1).
This means that the strain we worked with, originally collected in a clementine orchard of La Plana Region, is quite tolerant to hot and dry conditions. Other authors had also reported high survival rates for T. urticae, at either constant or fluctuating temperatures, similar to ours (Gotoh et al., 2004;Vangansbeke et al., 2013). Likewise, the maximal oviposition rate (8.7 eggs) obtained at 35°C independently of relative humidity falls into the range of what other authors had previously reported (7.1-9.1 eggs) (Bounfour & Tanigoshi, 2001;Vangansbeke et al., 2013). These values contrast with what was observed for the three phytoseiids. Only N. californicus presented a survival comparable to T. urticae at high temperatures (except when pollen was the only food source available) ( Figure 3). This species was initially purchased from a commercial producer, and this may explain the tolerance to a wide range of temperature and relative humidity conditions representative of the new environments where it may be released. For the other two phytoseiids, survival barely exceeded 60% in the best case, which was always at temperatures below 25°C, except in the case of E. stipulatus when fed pollen only.
In this case, survival also exceeded 60% irrespective of temperature and relative humidity ( Figure 3).
Although N. californicus and, particularly, P. persimilis could increase their predation rates on T. urticae eggs up to 30°C, and probably effectively regulate the herbivore below this temperature (i.e. predation rate > T. urticae oviposition rate) (Figures 1 and 2 (Ferragut et al., 1987) do not consider E. stipulatus as a suitable biological control agent for T. urticae. However, both field (Pérez-Sayas et al., 2015) and semi-field (Grafton-Cardwell et al., 1997) assays in citrus point at the important role of this phytoseiid, which should be attributed to its higher abundance relative to other co-occurring phytoseiids rather than to its effectiveness, in the regulation of T. urticae populations in clementines. This predator is known to be poorly adapted to prey on T. urticae because it cannot invade the web produced by this spider mite (Abad-Moyano et al., 2010;Ferragut et al., 1992). However, we observed a non-negligible predation rate at low temperatures and high relative humidity regimes (15.6 eggs at 15-20°C and 70% relative humidity), which improved when E. stipulatus had access to pollen ( Figure 4). As this is an omnivorous phytoseiid, considered a specialized pollen feeder (Ferragut et al., 1987;González-Fernández et al., 2009;Guzmán et al., 2016;Pina, Argolo, Urbaneja, & Jacas, 2012), However, this has not always been the case (Castagnoli & Simoni, 1999;Castagnoli et al., 2001;Croft et al., 1998;Ghazy et al., 2014;Nguyen et al., 2015), and this may be partly attributed to the use of different strains which may differ in their tolerance to harsh conditions. As this species was also the most tolerant to climate change conditions in our laboratory assays, its good performance in the field (Urbaneja-Bernat et al., 2019) was not a surprise, and it indeed performed better than what could be deduced based on these laboratory assays. This enhanced performance could be the result of two factors acting synergistically. On the one hand, in the laboratory, we worked at constant temperature regimes, which can be taken as a worst-case scenario that does not allow the mite to recover, most likely at night, from maximum temperatures attained during the day at field conditions. It is known that fluctuating temperatures have usually a lower impact on arthropod physiology and behaviour than a constant temperature equivalent to their mean (Gotoh et al., 2014;Nguyen & Amano, 2010;Vangansbeke et al., 2013;Bayu et al., 2017). On the other, we observed that the number of phytoseiid escapees in our assays was more extensive than that of dead specimens (Figures 2 and 3), and this may be taken as indicative that in the real world, these individuals would have been able to survive in refuges (i.e. crevices or cracks in branches). This behaviour, which may impact predator fitness in terms of lost foraging time and reproduction opportunities when looking for shelter (Gillespie et al., 2012), may increase its survival under field conditions. These two factors could also apply to the third predator considered in this study, P. persimilis, as the number of escapees for this mite was even higher than observed for N. californicus (Figure 1b,c). Indeed, P. persimilis spends relatively more time searching and moving around the leaf than the other predatory mite species (Gontijo et al., 2012;Sabelis & Dicke, 1985). However, Skirvin and Fenlon (2003) showed that the mobility of P. conditions. Importantly, our results also show that for the other two phytoseiids, access to pollen in combination with T. urticae eggs decreased the number of specimens found dead in the arenas and reduced predation and oviposition relative to the T. urticae-only diet.
Further implications of this supplementary food on interspecific relationships between these predatory species, as E. stipulatus is considered a superior intraguild competitor (Abad-Moyano et al., 2010), could shed light on whether pollen supply in this particular system could be advisable or not.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.