Introduction

Mango (Mangifera indica L.) is the third most important tropical fruit worldwide, after banana (Musa sp.) and orange (Citrus sinensis L.). México is the fifth producer worldwide and the first in exportation (^{FAO, 2016}). Oidium mangiferae Berthet, the causal agent of powdery mildew in mango, is widely distributed throughout the Pacific region. The disease has been reported in many countries of Asia, Middle East, Africa, South Africa, America, and Oceania (^{Nasir et al., 2014}). In México, the disease is widely distributed especially in the Pacific Coast and the severity has increased considerably in the last years (^{Félix-Gastélum et al., 2013}).

Powdery mildew is observed as a white powder, consisting of mycelium and conidia, on leaves, petioles, inflorescences, and fruit (^{Nelson, 2008}). Mango is the only known host for the pathogen, which can have a sporadic, but very severe incidence, causing up to 90% crop loss (^{Nelson, 2008}).

The pathogen develops in dry and cold environments, but reaches greater severity at 90% relative humidity (RH) and 20-25°C (^{Nasir et al., 2014}). In Michoacán, the maximum severity of the disease coincides with periods of temperature of 13-20°C and 22-60% RH (^{Guillén-Sánchez et al., 2003}). The control is based mainly on fungicides, such as Carbendazim, Benomyl, and Thiabendazole, to which O. mangiferae had already developed resistance (^{Mora Aguilera et al., 2002}; ^{FRAC, 2012}). In contrast, fungicide rotation with different sites of action, with molecules such as strobilurins (azoxystrobin and trifloxystrobin), anilinopyrimidines (cyprodinil), morpholines (fenpropimorph, fenipropidinspiroxamine), triazoles (myclobutanil, propiconazole, tebuconazole and Triadimefon), fenoxyquinolines (Quinoxyfen), and multisite fungicides (sulfur, dithiocarbamates and quinomethionates) can act efficiently and reduce the risk of appearance of fungal resistance (^{Margot et al., 1998}; ^{Reuveni & Reuveni, 2002}; ^{Wicks & Hitch, 2002}).

Epidemiology can aid the efficient evaluation of management practices for diseases (^{Guillén-Sánchez et al., 2003}, ²⁰⁰⁴). Epidemiological studies for specific production areas can help establish effective chemical control programs to be included in an integrated pest management program. Despite this, there are few epidemiological studies on powdery mildew on mango in México, with only one research (^{Téliz, 1998}), made in Michoacán by the Interdisciplinary Research Group on Mango (GIIM Spanish acronym). However, there is no research on the subject in Técpan de Galeana (subhumid tropic), Guerrero, which is the main producing state in México.

The objective of this work was to evaluate phytosanitary strategies for powdery mildew chemical control, based on the alternation of active ingredients with different modes and sites of action; and to explain the temporal development of the pathogen based on flowering phenology, seasonal density of inocula, and on environmental factors.

Materials and Methods

The experiment was carried in an orchard, in Las Tunas (17°16'N, 100°33'W, at 20 m altitude), Técpan de Galeana (subhumid tropic), Guerrero, México, with average annual precipitation and temperature of 1,525 mm and 27.1°C, respectively (^{SMN, 2017}).

The orchard is planted on a sandy loam red soil with 14-15 year-old 'Manila' mango, spaced at a 10x10 m arrangement. The study was conducted during the 2011-2012 cycles, within a program of agronomic management (PAM) to control mango diseases. This program was supplemented with fungicide application strategies, based on the use of different active ingredients, modes of action, and spray timing, aimed to control O. mangiferae. The PAM consisted of: sanitary pruning, soil and leaf chemical fertilization, weekly irrigations, and flowering stimulation (KNO₃, 2%).

Trees with similar trunk diameter and canopy volume were selected, treated, and evaluated. Three flowerings and vegetative flushes occurred during the experiment. Chemical control was done in the first and third flushes. The treatments were based on the rotation of fungicide active ingredients from different chemical groups and with different modes of action. Because the study was carried out in the field, in a commercial orchard, the treatments were inserted within an integrated pest management system that considered also other diseases, such as anthracnose (Colletotrichum gloeosporioides). Therefore, some tested active ingredients were not specific for control-ling powdery mildew, such as copper oxychloride.

Contact multisite-activity fungicides (C) were alternated with systemic activity fungicides (S). Protection periods of 7 and 15 days were assigned to the multisite and to systemic fungicides, respectively. The treatments were established based on these periods. In the first flush (vegetative and floral), the following treatments were used: T1, C^a + S^1a + C^a + S^2a; T2, S^1a + C^a + S^2a + C^a; T3, S^1a + C^a + C^a + S^2a; T4, C ^a + S^1a + S^2a + C^a; and T5, control; for which S^1a is azoxystrobin, S^2a is myclobutanil, C^a is sulfur + copper oxychloride, and control is distilled water. In the second flush, only contact fungicides, with multisite activity, combined with sulfur, were applied, with the following treatments: T1, Cu+S (copper oxychloride + sulfur); T2, Clo + S (chlorothalonil + sulfur); T3, Ma + S (mancozeb + sulfur); T4, Ca + S (captan + sulfur); and T5, control (distilled water). In the third flush, the following treatments were applied: T1, S^1b+C^b+S^2b+C^b; T2, C^b+S^1b+C^b+S^2b; T3, S^1b+C^b+C^b+S^2b; T4, C^b+S^1b+S^2b+C^b; and T5, control; in which S^1b is cyprodinil + fludioxonil, S^2b is quinoxyfen, and C^b is chlorothalonil + sulfur.

The application of the chemical treatments was based on the floral phenology model (^{Guillén-Sánchez et al., 2003}), which considers as the most susceptible stages those from swollen buds to fruit with diameters lower than 8 mm. The fungicides were applied according to their commercial doses: azoxystrobin, 0.25 g L^-1; myclobutanil, 0.08 g L^-1; cyprodinil+fludioxonil, 62.5 + 0.075 g L^-1; quinoxyfen, 0.05 mL L^-1; copper oxychloride, 2.5 g L^-1; mancozeb, 0.8 g L^-1; captan, 1.25 g L^-1; chlorothalonil, 2.88 g L^-1; and elemental sulfur, 1.35 mg L^-1. The applications were done with a Swissmex agricultural handbarrow, with 500 L capacity and flow of 34.2 L min^-1, at 20 bar, using 5.5-HP (4.1 kw) motor, at 3,600 rpm; with two 50 m hoses with aspersion handles.

A randomized complete block design was carried out with four replicates. In each one, five inflorescences and five vegetative shoots taken from a tree, with homogeneous growth flush, were randomly selected and labeled. A previous research had shown that there is no effect of the orientation of inflorescences and vegetative shoots on disease intensity (^{Guillén-Sánchez et al., 2003}). Two logarithmic scales (Table 1), with 7 and 10 classes (k), were developed using the software Doslog v.1.0 for Windows, with the following grades, according to the leaf area affected by the disease: 0, healthy tissue; and 6, damaged area ≥ 40%. For the inflorescence, the grades were: 0, healthy tissue; and 9, 90% panicles completely infested with the fungus. Disease severity was estimated according to these grades, on a weekly basis, from the swollen buds detection until the fruit reached a diameter of 8-15 mm, with the formula:

in which: Sev_i is the severity at time i; X_ki is the damage level k at time i; N_ki represents the inflorescences with damage level k at time I; N_i are inflorescences evaluated at time i.

The temporal analysis of the disease progress was made by adjusting growth curves with the models monomolecular, logistic, and Gompertz, taking into account the coefficients of determination (R²) and the quality of the fit, standardized with the Weibull model (b^-1). In order to compare the effect of the factors on the disease, the final severity (Y_f) and the area under the disease progress curve (AUDPC) were calculated for each treatment, by integrating the Trapezoidal method (^{Campbell & Madden, 1990}; ^{Campbell & Benson, 1994}).

The analysis of variance (Proc GLM) and the comparison of the mean (LSD test with reliability of 95%) were carried out in SAS v.9.1.3 (SAS Institute Inc, Cary, NC, USA) statistical software.

Temperature, precipitation, relative humidity, dew point, and wind direction and speed were recorded every two hours, with a Vantage Pro2 Weather Station (Davis Instruments Hayward California, USA), equipped with a Wheatherlink for Windows, v.5.9.0 software. The station was configured with a laptop, and data were exported to Microsoft Excel spreadsheet for processing.

In order to estimate the density of O. mangiferae conidia in the air, volumetric spore traps were designed, constructed, and installed in the center of the experimental area, at 1.5 m above ground level. The air suction was performed with a blower motor for industrial use, with 12V, and a volumetric capacity of 2.3-3.5 L min^-1. Inside the trap, the spores were attached onto an adhesive tape placed around a rotating plastic disk, with a digital mechanism scheduled for seven days. Each tape had 1.5x39.5 cm, taken from rolls of transparent self-adhesive polypropylene (Contac). The tapes were tightly attached to a strip sheet of a hygrothermograph (as a template), around a plastic disk. This helped to differentiate individual days and hours of spore trapping. The tapes were removed weekly, and each 39.5-mm section of the template corresponded to a trapping interval of 24 hours. They were examined under a compound microscope, at 40X, for daily and weekly quantification of the spore density in the air.

Pearson correlation coefficients (r) were calculated between disease intensity and conidia density, and climatic conditions of temperature, relative humidity, speed and direction of the wind, precipitation, and dew point.

Results and Discussion

Flowering (Fl1, Fl2, and Fl3) and vegetative flushes (Veg1, Veg2, and Veg3) were registered and, in the experimental site, with subhumid tropic climatic conditions, they occurred in the periods shown in Figure 1, which depicts the severity of the disease in the field accordingly to the scale used, and allowed of the temporal development estimation of the epidemics. The first symptoms of the disease were observed, on average, 30 days after the appearance of the swollen buds. The most susceptible phenological stages were: full flowering, with 30% severity fruit set (3-5 mm diameter); and fruit with 8-15 mm, with 35% severity. ^{Joubert (1991)}, ^{Schoeman et al. (1995)}, and ^{Guillén-Sánchez et al. (2003)} reported similar results. The maximum severity in inflorescences (Fl1, Y_f = 6.55; Fl2 Y_f = 6.90; and Fl3 Y_f = 6.15) and in vegetative shoots (Veg2, Y_f = 1.80; and Veg3, Y_f = 1.65) was recorded 40 days after initial symptoms (Tables 2 and 3).

Figure 1. Temporal progress curves for disease severity during the vegetative (A) and flowering (B) stages of mango with epidemics of Oidium mangiferae during three growth flushes, according to five chemical treatments, in a commercial mango orchard (Mangifera indica 'Manila') in Las Tunas. C^a, sulfur + copper oxychloride; S^1a, azoxystrobin; S^2a, myclobutanil; C^b, chlorothalonil + sulfur; S^1b, cyprodinil + fludioxonil; S^2b, quinoxyfen; Cu+S, copper oxychloride + sulfur; chlo+S, chlorothalonil + sulfur; Ma+S, mancozeb + sulfur; and Ca+S, captan + sulfur. 

According to ^{Singh (2000)}, warm humid weather, low-night temperatures, and high-wind velocity favor disease spread and development. These climatic conditions are common in the subhumid tropic region evaluated, and explain why the disease was found during all three growth flushes. The climatic conditions impacted the disease severity, with correlation coefficients ranging from 0.82 to 0.93 (Table 4). Epidemic relevance of the disease in foliage is significant, since it provides a source of inocula that can infect the inflorescences, both in the same production cycle and in the following growing seasons (^{Akhtar & Alam, 2002}; ^{Guillén-Sánchez et al., 2003}). However, future studies are needed to estimate their impact on tree vigor.

The treatment with the best result in the inflorecences was T1 [S^1a+C^a+S^2a+C^a = azoxystrobin, sulfur + copper oxychloride + myclobutanil + (sulfur + copper oxychloride)], that showed Y_f lower than 4.95 and AUDPC of 145.75, in the first flowering flush. In vegetative shoots, however, the best results were obtained with T3 (S^1a+C^a+C^a+S^2a), that showed Y_f= 1.00, and AUDPC=26.5.

In the second growth flush, when systemic fungicides were not used, the best combination of ingredients were chlorothalonil + sulfur, with AUDPC=178.75 in the flowering; and copper oxychloride + sulfur, with AUDPC=11.75 in the vegetative shoots. This last result suggests a synergistic effect of the combination with sulfur.

In the third and final growth flush, the best treatment was T1 [S^1b+C^b+S^2b+C^b = quinoxyfen + (chlorothalonil + sulfur) + (cyprodinil + fludioxonil) + (chlorothalonil + sulfur)], both for inflorescences - that showed Y_f= 4.85, and AUDPC=181.50 - and foliage, with Y_f= 0.60, and AUDPC=11.50 (Tables 2 and 3). The control treatment recorded the highest intensity of disease in all flower and vegetative stages evaluated.

The Weibull model described well (^{Madden et al., 1981}; ^{Sinha & Varma, 2004}) most of the registered epidemics, with R² higher than 0.85 (Figure 1). However, there were differences in apparent infection rates, due to the treatments, only in the vegetative flushes 1 and 3 (Table 3). There were no expressive differences of treatment effects on the development of the disease in vegetative or flowering flushes (Tables 2 and 3). Fungicides from different chemical groups and modes of action, or with multisite activity, are commonly reported as controlling powdery mildew in mango (^{Hollomon et al., 1997}; ^{Reuveni et al., 1998}; ^{Wicks & Hitch, 2002}; ^{Guillén-Sánchez et al., 2004}; ^{Nelson, 2008}). These reports were confirmed in the present study, where azoxystrobin, myclobutanil, and copper oxychloride + sulfur were the most effective fungicides.

However, the present study, not only evaluated specific authorized fungicides (^{MRL, 2010}; ^{Sagarpa, 2010}) for powdery mildew in mango, but also their alternated use and the sprayed order, aiming at the preservation of their effectiveness with a rational phytosanitary approach (^{Halleen & Holz, 2001}). Overall, the best strategy was to initiate chemical control using a systemic fungicide, followed by a contact one with multisite activity. In addition, the application time should consider the availability of susceptible tissues (full flowering, fruit set of 3-5 mm, and fruit with 8-15 mm diameter), regardless of the number and intensity of the growth flushes (^{Schoeman et al., 1995}; ^{Halleen & Holz, 2001}). This strategy should be combined with cultural practices to reduce environmental humidity in the orchard, such as ventilation pruning, elimination of inoculum sources, avoiding excessive irrigation, and improving drainage and the timely control of weeds (^{Saifullah et al., 2007}). Moreover, periodical foliar phosphate fertilization (^{Reuveni & Reuveni, 2002}) can also help to improve the control of the disease. These practices may preserve long-term effectiveness of systemic molecules, thus limiting the risk of generating resistance (^{Halleen & Holz, 2001}; ^{FRAC, 2012}).

The periodic occurrence of cold winds (≤ 20°C) in subhumid tropical regions near the sea, as occurred in this study, has an important impact on the promotion of the floral initiation processes (^{Pérez & Puche, 2003}), but it can also increase the susceptibility of mango plants to powdery mildew, as observed in Técpan de Galeana (subhumid tropic), Guerrero. In areas with more than one epidemic (more than one growth flush) per cycle, the chemical control should be carried out more often. Therefore, it is important to alternate the active ingredients, with different combinations and order of application (^{FRAC, 2012}), as ^{Tahir et al. (2009)} reported for O. mangiferae, and ^{Halleen & Holz (2001)} for Uncinula necator on grapevines.

In the present study, the severity in flowering and vegetative flushes were associated with temperatures higher than 30°C (r=0.92 and 0.94, respectively), relative humidity ≥90% (r=0.80 and 0.93), and density of air sporulation (r=0.88 and 0.96). In addition, spore density was correlated with temperatures between 20 and 26.9°C (r=0.65), relative humidity between 60 and 89.9% (r=0.65), dew point between 15 and 19.9°C (r=0.63), and wind speed between 1-1.9 km h^-1 (r=0.70) (with WN direction, r=0.88).

Flowering and vegetative severity of the disease were correlated with temperature higher than 30°C and relative humidity higher than 90%. Therefore, these parameters were considered determinant for the development of the disease. Inocula were present throughout the studied period (Figure 2). Therefore, the severity will depend only on the proportion of susceptible tissues and on favorable environmental conditions. ^{Singh (2000)} reported that warm humid weather (64-72%), low-night temperatures, and high-wind velocity favor the spread of powdery mildew. Moreover, it is reported that maximum sporulation occurs between 20 and 26.9°C, with relative humidity higher than 80%, and dew point between 20-24.9°C (^{Makowski, 1993}; ^{Akhtar & Alam, 2000}; ^{Halleen & Holz, 2001}; ^{Sinha et al., 2001}). It has also been documented that temperatures above 35°C, humidity between 50 and 70%, and wind exceeding 1.0 km h^-1 favor conidial dispersal (^{Schoeman et al., 1995}; ^{Akhtar & Alam, 2000}; ^{Singh, 2000}).

Figure 2. Oidium mangiferae spores monitoring (A), and their relationship with disease severity on inflorescences (B), during three flowering flushes, in a commercial 'Manila' mango orchard (Mangifera indica), in Las Tunas. 

Conclusions