The experimental model shown in Table 1 compares two SVAM used to apply fungicides with or without an adjuvant against P. infestans in processing tomato. The fixed model (SV300) involves the spraying of 300 L/ha of fungicide suspension from the first to the last application. In the proportionate spray volume (PSV) model, 300 to 750 L/ha of SV was applied. The order of application and suitable SV for specific dates is shown in Table 2. The SV for the PSV model was calculated depending on to the number of fully developed leaves per plant according to the following formula:

PSV (L/ha) = LN × SVI

where LN = average number of leaves per plant; SVI = spray volume index per leaf (11.16). SVI was calculated in the following way: SVI = 300/LN₁, where LN₁ = average number of leaves per plant on the first day of application in 2009.

It was assumed that the maximum spray volume in season would have to be maintained right until the end of the last application, despite the reduction of the number of leaves.

Azoxystrobin (methyl (E)-2-[2-[6-(2-cyanophenoxy)pyrimidin-4-yl]oxyphenyl]-3-methoxyprop-2-enoate - Amistar 250 SC; Syngenta Limited, Guildford, UK) at a dose of 125 and 250 g/ha (50% and 100% of the dose recommended, respectively) or chlorothalonil (2,4,5,6-tetrachlorobenzene-1,3-dicarbonitrile - Gwarant 500 SC; Arysta LifeScience S.A.S., Noguères, France) at a dose of 625 g/ha and 1250 g/ha (50% and 100% of the dose recommended, respectively) were applied alternatively (Table 2). Adjuvants: Slippa (Interagro (UK) Ltd., Great Notley, UK) containing polyalkyleneoxide modified heptamethyltrisiloxane (655 g/L) and Torpedo II (De Sangosse Ltd, Swaffham Bulbeck, UK) formulation of four active ingredients (alkoxylated tallow amine, 210 g/kg; alcohol alkoxylate, 380 g/kg; natural fatty acids, 75 g/kg; and polyalkylene glycol, 210 g/kg), described as multi-ingredient adjuvant, at the concentration of 1 mL/L, were added to the tank mix only when the fungicides were applied at half the dose.

Field studies were conducted for three years, from 2009 to 2011 in Poznań, Poland (16.858092 52.408529 decimal degrees). The experiment was performed using completely randomised block design, with 3 blocks in 2009 and 4 blocks in the other two years. Tomato (Solanum lycopersicum L.) cv. Polset F1 was cultivated in a 2-row system (0.5 m × 0.5 m × 1.5 m) on a sandy loam soil. There were 14 plants on the plot. The plants were planted from 2 to 9 June each year. The plantation was subject to standard fertilisation and chemical weed control procedures. Tomato growth stages (GS) were noted during vegetation. The BBCH identification keys according to Feller et al. (1995) were used to identify tomato GS.

The application of azoxystrobin and chlorothalonil commenced 5 days after the first small pale green lesion on leaf, characteristic to PLB, was observed in 2009 and after 4 days in 2011 (Table 2). In 2010 the symptoms of PLB were not observed in July, thus spraying started in August as a result of increase in accumulated risk of disease incidence due to more favourable weather condition at the beginning of this month and because of growth in leaves numbers. The spraying was conducted using a precision field knapsack sprayer equipped with horizontal spray boom with 4 nozzles, the XR 11003 (TeeJet Spraying Systems Co., Wheaton, IL, USA). A fine droplet size application was conducted at 300 kPa pressure.

The meteorological conditions were registered with a HOBO Weather Station and dedicated sensors (ONSET Computer Corporation, Bourn, USA) and are presented in Table S1 [supplement]. The following data were registered 2 m above the ground: air temperature (°C) and relative humidity (%) with a S-THA-M002 sensor, total solar radiation (MJ/m²) with a Silicon Pyranometer S-LIB-M003, wind speed (m/s) with a S-WCA-M003 sensor. The amount of rainfall (mm) was registered with a rain gauge 0.2 mm - S-RGB-M002. The meteorological dates were used to evaluate the risk of PLB development, fungicides timing and then to estimate the influence of microclimate on fungicides action.

Disease symptoms were evaluated visually on 10 plants per plot. The modified Horsfall-Barrat rating scale of 1 to 12 (1=0%, 12=100% disease severity) was used to evaluate PLB severity on leaves (Berger, 1980). The number of fruits with PLB symptoms was calculated as percentage, 20 fruits per plant were observed. The estimations were conducted on the last day of the application and 2 weeks later.

The accumulated value of the risk of the disease was calculated using the NegFry model according to the procedure proposed by Ullrich & Schroeder (1966). The daily risk value was calculated based on the hourly relative humidity, precipitation and the average temperature within the field of the crop. Accumulated risk value over the period of time is the result of summing daily risk values starting at the day of tomato planting.

Samples for study of residues were collected in the amount of no less than 1 kg from the plot, 3 days after the last application of azoxystrobin and 7 days after the last application of chlorothalonil in 2009 and 2011 or 2 and 5 days after the application in 2010, respectively (Table 2).

The QuEChERS pesticide multiresidue method was validated for extraction of two fungicides: azoxystrobin and chlorothalonil, according to the Health & Consumers Directorate-General of the European Commission (SANCO) requirements [SANCO 12571/2013]. The validation of this method was carried out using six pesticide-free tomato samples for each fortification level. The homogenized samples (10 g) were spiked with a small volume (<0.5 mL) of an appropriate standard mixture solution and shaken with acetonitrile (10 mL) (Payá et al., 2007). As a slight modification, the obtained mixtures were frozen for two hours to save the volatile pesticides before the addition of salt mixture consisting of anhydrous magnesium sulphate (4 g), sodium chloride (1 g), disodium hydrogen citrate sesquihydrate (0.5 g), and trisodium citrate dehydrate (1 g). After centrifugation an aliquot of the acetonitrile phase was transferred into a centrifugation tube containing a mixture of anhydrous magnesium sulphate and primary secondary amine (PSA) sorbent (0.9 g). After purification with PSA, the extracts were re-acidified by adding 50 μL of the 50 mL/L formic acid solution. The extracts were transferred to the glass tubes and carefully evaporated to dryness. The last step was to add an acetone/hexane solution (Lehotay et al., 2010). A mixture of azoxystrobin and chlorothalonil, amenable to gas chromatography (GC), was quantitatively recovered from spiked tomato samples and determined using gas chromatography with a nitrogen phosphorus detector and electron capture detector (GC–NPD/ECD 6890 – N Agilent Technologies, USA; column: DB-5, 30 m × 0.53 mm × 0.88 µm, Agilent J&W Scientific, USA) The spiking levels for the recovery experiments were: 0.02, 0.2 and 2.0 mg/kg for GC–NDP/ECD analyses. The method accuracy and precision were evaluated by performing recovery studies. The precision was expressed as the relative standard deviations (RSD). Accuracy was measured by analysing samples with known concentrations and comparing the measured values with the true values. A default expanded measurement uncertainty of 50% was applied according to results reported by laboratories participating in a number of EU proficiency tests [SANCO 12571/2013].

The experiment analysed 8 treatments giving incomplete three-factor structure and zero control (Table 1), therefore the study was regarded as a one-factor experiment. As a result of the Bartlett test, in case of some of the variables, the logarithmic transformation or the non-parametric Kruskal-Wallis test was necessary. A multivariate analysis of variance was conducted using the Hotelling-Lawley test to analyse the difference between the average values of the multivariate variables (R stats package).

An appropriate one-way variance analysis (ANOVA) with division into group contrasts between the treatments (so called basic contrasts) was also conducted. In order to determine the coefficients of the contrasts (normalisation and orthogonalisation) presented in Table 3, the R gmodels package was used (Warnes et al., 2015). Tukey’s post hoc test was used for comparison of the studied treatments.

Diverse influence of the studied treatments on values of azoxystrobin and chlorothalonil residues implies the use of exploration techniques of analysis of those multidimensional data. To group treatments, taking into consideration average values of residues in all 3 years of studies, a cluster analysis with the use of Ward’s hierarchical clustering and Euclidean distance was performed. The package agricolae (Mendiburu, 2013) for program R (3.0.2) was used for the calculations.

Risk of PLB. The infestation of plants with PLB differed throughout the years, with its peak observed in 2011. Natural incidence of P. infestans was observed in 2009 and 2011 on 16 and 8 July, respectively, while in 2010 it was 1 September. However, according to the NegFry model, the accumulated risk of PLB in July was the highest in 2009, followed by 2011 and 2010 (Fig. 1). Heavy rainfalls in June and July 2009 and in July 2011 favoured the incidence of PLB in July in both years (Table S1 [supplement]). However, the lower average air temperature in July 2011 and rich foliage of tomatoes (Tables 2 and S1 [supplement]) increased the risk of PLB incidence more than in other years. Within 14 days after the last application of azoxystrobin the average daily air temperature was higher and relative air humidity was lower in 2011 than in 2009 and 2010 and it reduced the risk of new infection with P. infestans in 2011 (Table S1 [supplement]).

Fungicide dose response. Each year applications of fungicides reduced tomato infestation with PLB (Table 4). However, there were not always significant differences between particular treatments and control (Table 5). The dose of fungicide applied without adjuvants influenced tomato infestation with PLB in 2011. Two weeks after last application in 2011 a full dose of the fungicide resulted in lower infestation of tomato than half the dose (contrast value –22.25 for leaves and – 41.5 for fruits) (Table 4). While the mean severity of disease on leaves was 55.5% for full and 66.7% for half the dose, fruits were infested in 33.3% and 54.1%, respectively. Thus the corresponding reduction of plant infestation after application of fungicides was approximately 16.7% for leaves and 38.5% for fruits (Table 5). On the day of last spraying of azoxystrobin in 2011 the reduction in PLB incidence between full and half a dose was 24.1% for leaves and 27.6 for fruits. Unfortunately, contrast values for leaves reached –10.43 and –24 for fruits and were insignificant (Table 4). Similarly, contrast analysis for 2009 and 2010 showed lack of significant differences in results. However, plant infestation on suitable treatments in these years could confirm higher potential of full than half a fungicide dose for tomato protection against PLB.

SV response. The SVAM influenced the long-term infestation of tomato (Table 4). Two weeks after last spraying in 2011 the PSV resulted in lower leaves infestation of 8% (contrast value = 21.25) and fruits of 17.4% (contrast value = 38) than SV300. The corresponding values of mean infestation for PSV were 61.5% on leaves and 45% on fruits, while for SV300 66.8% on leaves and 54.5% on fruits. The post-hoc test proved that PSV system decreased tomato infestation with PLB when half of the recommended dose of the fungicide was applied without adjuvant (Table 5). The corresponding reduction of plant infestation on suitable treatments was 19.9% for leaves and 37% for fruits. However, the results in 2009 and 2010 did not differ between PSV and fixed system.

Adjuvants and SV response. The adjuvants under study did not increase tomato protection against PLB in comparison with the fungicide applied without adjuvant (Table 4). The multi-ingredient adjuvant even increased the infestation of tomato leaves by mean of 18% but only with the PSV system in 2011 (Table 5). The contrast analysis for 2011 showed that regardless of SVAM polyalkyleneoxide modified heptamethyltrisiloxane decreased leaves infestation by 10.2% to multi-ingredient adjuvant (Table 4).

Validation of analytical method of fungicide residue determination. The mean recoveries were above 78.5% and 64% for azoxystrobin and chlorthalonil, respectively. The relative standard deviations were below 8.5% for azoxystrobin and 11.9% for chlorothalonil (Table 6). The detection limit was 0.02 mg/kg.

Dose response. The contrast analysis showed that the full dose of the fungicide resulted in higher residues of azoxystrobin and chlorothalonil than half the dose (Table 7). The mean residue of azoxystrobin after spraying with half the dose (0.065 mg/kg) was significantly reduced in comparison to full dose (0.1 mg/kg) only in 2011. The corresponding reduction was 35.1%. The chlorothalonil residues were significantly lower with full compared to half the dose in 2009 and 2010 by mean of 45.2% and 78.7%, respectively. The mean residues were as follows: 0.42 mg/kg with full and 0.23 mg/kg with half the dose in 2009 while in 2010 they were 0.45 mg/kg and 0.095 mg/kg, respectively. However, the post-hoc test indicated that only chlorothalonil residues can be regarded as significantly lower after application with half the dose (2009) (Table 8). The reduction was 17.8% for SV300 and even 76.9% for PSV. The cluster analysis showed that combined residues of fungicides applied at a full dose were higher than at half the dose (Fig. 2). It is noteworthy that the azoxystrobin residues in 2010 were higher than in other years because of the necessary repetition of the last application due to local short heavy rainfall (Table 2).

SV response. The SVAM significantly influenced only chlorothalonil residues in 2009 by mean of 35.9% (contrast value –0.36) (Table 7). The mean residues were as follows: 0.165 mg/kg with PSV and 0.258 mg/kg with SV300. The post-hoc test proved that PSV system significantly decreased residue only when half of the fungicide dose was applied. The cluster analysis showed that after the application of fungicides at half the dose using the PSV model the residues created a separate cluster from SV300 and were located closer to the control (Fig. 2).

Adjuvants and SV response. The statistical analysis proved that adjuvants affected residues of chlorothalonil in 2009 (Table 8) and azoxystrobin in 2010. The residues of chlorothalonil were reduced with SV300 from 0.37 mg/kg to 0.15 mg/kg in case of polyalkyleneoxide modified heptamethyltrisiloxane and even to 0.06 mg/kg for multi-ingredient adjuvant. Despite polyalkyleneoxide modified heptamethyltrisiloxane increased chlorothalonil residue of 22.2% with PSV in 2009. Moreover, chlorothalonil residue after spraying with PSV system and using last one adjuvant amounted to 0.11 mg/kg and was significantly higher than 0.07 mg/kg achieved with multi-ingredient adjuvant. The azoxystrobin residue was significantly reduced with SV300 from 0.54 mg/kg to 0.29 mg/kg only after the application of multi-ingredient adjuvant (2010).

The cluster analysis showed that combined residues of fungicides with adjuvants were roughly lower using PSV relative to SV300 system (Fig. 2). The adjuvants contributed to the reduction of fungicide residue in case of SV300 system.

A full dose of fungicide has higher potential to increase the fungicide efficiency (Bain et al., 2014). This fact was confirmed by azoxystrobin and chlorothalonil residues in tomato plants under study (Table 7, Fig. 2). However, in low disease pressure circumstances a reduced dose of fungicide can even ensure similar level of plant protection. It could be seen in the results obtained in 2009 and 2010. The study showed that tomato canopy characteristics can influence the fungicide rate. In view of the fact that the tomato plants had more leaves per plant at the beginning of application in 2011 (46.7) than in 2010 (33.4) and in 2009 (26.9) (Table 2), the average fungicide rate per unit of leaf surface area should have been lower in 2011 than in 2010 and 2009. Between 2009 and 2011 the delay in the leaves number was over 2 weeks (Table 2). We assumed that lower long-term fungicide rate after spraying may result in higher ultimate tomato infestation with PLB. However, this study cannot confirm directly this thesis. Studies on other crops (Walklate et al., 2003; Dammer et al., 2008; Llorens et al., 2011), including potato (Cooke et al., 2011) indicate that crop and canopy characteristics modify the fungicide rate per leaf surface area unit and could affect fungicide efficiency.

The PSV model resulted in lower long-term infestation of fruits and leaves than 300 L/ha in 2011 when half of the fungicide dose was applied. According to Grinstein et al. (1997), Washington (1997) and Prokop & Veverka (2006), higher surface coverage by the contact fungicide can be a factor favouring the decrease in infestation of the leaves by disease. The interception of the fungicide and spray suspension into the canopy may confirm this dependence and can clarify the tomato infestation in 2011 both with PSV and SV300. As was proved, the spray deposit into the canopy is proportional to the SV applied (Walklate & Cross, 2013). On the other hand, this deposit into the canopy is inversely proportional to the LAI (Zhu et al., 2004). Thus, the estimated spray depos it should be higher with PSV than with SV of 300 L/ha. At the end of the spraying season 754 L/ha of SV was applied in 2009, 597 L/ha in 2010 and 686 L/ha in 2011 with the PSV adjustment model. The fungicide residues can also explain the spray interception into the canopy and thus, biological results. If there had been no differences in fungicide residues between both SV adjustment systems under study, the spray deposit into the canopy could have been higher in the case of higher SV. The contact fungicide residue could have verified this better than the systemic fungicide. In 2011 the residues of chlorothalonil with half the dose of fungicide ranged 0.05-0.07 mg/kg for SV300 and 0.08-0.12 mg/kg for PSV (Table 8). Therefore, after the last application of 686 L/ha in 2011, the spray deposit on fruits was higher than after the application of 300 L/ha. The spray deposit on leaves at the same level above the ground could also have been higher. Hence, the higher spray deposit into the canopy could be considered as the reason of better long-term protection of tomato with the same fungicide dose per hectare in 2011. In 2009 and 2010 the residues of chlorothalonil were poorly coherent with result of 2011. PSV system gave significantly lower chlorothalonil residues than the fixed system (Table 7). The mean residue after spraying with PSV system was 0.17 mg/kg while 0.27 mg/kg with SV300. The corresponding residues in 2010 were: 0.16 mg/kg for PSV and 0.25 mg/kg for fixed system (Table 8). However, these results may be connected with tomato canopy shape. Tomato GS ≥ 85 appeared during the last spray application in 2009 (and also in 2010). At this time the canopy is flattened and much more transparent to spray droplets than at earlier growth stages. Thus, the plant coverage may not differ so much between both SV application systems. Moreover the SV of 300 L/ha resulted in higher concentration of the fungicide suspension on the plant surface. A high fungicide residue on the exposed fruits is in accordance with a high coverage of external part of the potato canopy by chlorothalonil (Bruhn & Fry, 1982a; Hamm & Clough, 1999). According to literature data, the results of biological action of contact fungicides against PLB were satisfactory when an SV of 160 L/ha was applied (Jensen & Nielsen, 2008). The alternate application of azoxystrobin and chlorothalonil presented in this study gave a chance to increase the probability of satisfactory biological results of PLB control using a spray volume of 300 L/ha. A systemic fungicide such as azoxystrobin may be able to maintain high performance using a mid- or low-volume spraying (Ratajkiewicz et al., 2009; Wise et al., 2010).

In spite of the positive potential of PSV in tomato protection against PLB, this system was not able to increase biological results significantly when fungicides were applied at a full dose. On the other hand, a high pressure of P. infestans and leaf biomass (2011) caused high infestation of tomato with PLB despite the PSV system.

Adjuvants did not have positive effect on the biological efficacy of the PSV application system or the fixed (Table 5). However, without adjuvants the PSV system was more efficient in tomato protection against PLB than the fixed system (2011). On the other hand, the cluster analysis shows that combined residues of fungicides applied at half of the recommended dose with adjuvants could be inversely associated with the SV. Hence, residues tend to be higher with the fixed rather than with the PSV system (Fig. 2). However, as was proved, the growth stage of tomato and canopy shape during last spraying had some influence on the residue in particular years. According to literature data (Schönherr et al., 1999; Ramsdale & Messersmith, 2001; Menéndez & Bastida, 2004), an appropriately selected adjuvant can enhance the biological action of pesticides and this effect is stronger with low rather than with high SV. The higher ratio between the adjuvant and pesticide and between the adjuvant and water can justify the stronger biological effect (Ramsdale & Messersmith, 2001; Menéndez & Bastida, 2004). In our study the same concentration of the adjuvant in water was used with both SV application systems. The ratio between the adjuvant and fungicide concentration in the spray suspension increased with increasing of SV in the PSV system, meanwhile the concentration of fungicide decreasing. Thus, this contraction could diminish the possible positive effect of adjuvant applied with the PSV system.

Other factors may also influence the fungicide action with adjuvants. The spreader adjuvants under study can increase droplet spreading and thus the coverage of the plant surface with a fungicide suspension. Therefore, the disappearance of fungicides due to volatilisation may also increase (Bedos et al., 2002), especially when a higher SV is applied. The photodegradation may also increase with increasing coverage. According to vapor pressure of chlorothalonil (0.076 mPa at 25 °C) and azoxystrobin (1.1 × 10^–7 mPa at 20 °C) the potential risk of volatilisation was higher for the first fungicide (Tomlin, 2000). However in field trial on potato the fate of chlorothalonil due to volatilization was only 5% after 7.6 days (Leistra & Van Den Berg, 2007). Furthermore, Monadjemi et al. (2011) estimated half-life of chlorothalonil due to photodegradation at 5.3 days. There are studies indicating high azoxystrobin stability on tomato leaves within 1–4 days after spraying (Szpyrka & Sadło, 2008) and even after 7-10 days on fruits (Garau et al., 2002). Thus chlorothalonil is more susceptible to physical interactions on leaf surfaces than azoxystrobin. On the other hand, the polyalkyleneoxide-modified heptamethyltrisiloxane could have facilitated the infiltration of fungicides into cuticles (Stevens et al., 1991). It was shown that cuticular waxes may affect photodegradation rates and photoproducts (Ter Halle et al., 2006). Thus, this adjuvant has a potential to diminish the disappearance of fungicides from the plant surface. We can conclude that the key to achieve high biological effectiveness of a spray mixture is the appropriate adjustment of the adjuvant properties with the fungicide. Mitani et al. (2001) showed that the addition of a super-spreading adjuvant to a spray liquid containing cyazofamid of stable residual activity and rainfastness properties is beneficial and necessary for the fungicide to be effective against PLB.

The number of leaves gives valuable information about the growth of tomato. Thus, the SV proposed in the PSV model simply reflects abundance of tomato foliage. However, we do not know how much of spray deposit is intercepted on particular stratum of canopy or on leaves. According to Bruhn & Fry (1982a) and Hamm & Clough (1999) we can assume that interception of spray into tomato plant decreased proportionally from top to bottom of canopy. However, in the end of tomato development the shape of canopy changes considerably due to stems starting to drop down and it is difficult to describe. Similarly, the interception of spray and fungicide into the canopy can change considerably. It would be valuable to optimise the PSV model according to changes in the canopy shape during fructification. Measuring the LAI, normalized difference vegetation index and fractional canopy cover would give valuable information about the tomato canopy and crop (Trout et al, 2008; Čereković et al., 2010; Fortes et al., 2015). A model based on a few plant and crop structure parameters would give better results of spray volume adjustment during the whole vegetation period than a PSV model based on one parameter.

However, let us hypothesize that the SV given by the proposed PSV model could be stopped or decreased from then on, as the plant canopy drops down and flattens. This thesis may be confirmed by residues of the contact fungicide (chlorothalonil). The plants at the GS 85-88 development stage (94 days after planting) (2009) had higher mean chlorothalonil residue in fruits after the application of SV300 (0.26 mg/kg) than PSV (0.17 mg/kg). In contrast, on younger plants with many numerous erect stems (GS 75-82, 70 days after planting) in 2011 the residues of chlorothalonil were opposite of that and equalled 0.07 mg/kg for SV300 and 0.12 mg/kg for PSV. On the other hand, the residues of systemic azoxystrobin were generally independent of SV applied in examined range. The mean residue of azoxystrobin in 2009 after the application with SV300 was 0.08 mg/kg and with PSV 0.09 mg/kg while in 2011 0.16 mg/kg and 0.14 mg/kg, respectively. The fungicide residue influences its biological action so it would appear to be justified to stopped or even decreased SV calculated for chlorothalonil and azoxystrobin according to PSV model as the plant canopy flattens.

On the other hand, it is worthy of attention that the residue of azoxystrobin in the end of withholding period (3 days) after spraying with SV calculated according to PSV and SV300 were substantially lower than the 3 mg/kg of maximum residue level in tomato fruits (EU, 2012a). While withholding period for chlorothalonil is also 3 days, the residues were evaluated 7 days (or 5 days in 2010) after spraying. These residues were also substantially lower than the 2 mg/kg of maximum residue level (EU, 2012b). Hence, both SVAM guaranteed food safety and consumer security.

Only one system of plant spacing and tomato variety was used in this study. The leaf number index used to calculate the SV may have varied due to different plant spacing and varieties. However, the number of leaves per plant can be easily translated into the number of leaves per hectare, thus diminishing the influence of plant spacing. Despite this, leaf morphology is influenced by tomato variety types. The proposed SVI may work properly with a tomato variety similar to Polset F1. However, according to plant morphology, this variety is classified as the most commonly used variety type in processing tomato production. In conclusion, the simplicity of PSV model can make it a useful tool in adjustment of SV in processing tomato crop.