1. Introduction

The production of soybean (Glycine max (L.) Merrill) in Brazil occupies 3.5% of the territory and it is considered the second largest worldwide producer and exporter of whole soybeans, soy meal and soy oil ( ^{Hirakuri and Lazzarotto, 2014} ; ^{Brasil, 2016} ). Lepidoptera species are considered one of the most important groups among the insects that feed on soybean leaves causing direct defoliation. There are at least 69 species of Lepidoptera documented causing damage in soybean fields in South America (Argentina, Brazil, Chile and Uruguay). Among these species, representatives of the superfamily Noctuoidea are of particularly a concern due to its feeding behavior causing significant economic damage ( ^{Moscardi et al., 2012} ; ^{Formentini et al., 2015} ).

The identification and monitoring of the abundance of lepidopterans in agroecosystems is essential in order to establish an efficient management strategy and allow for the proper use of insecticides ( ^{Hoffmann-Campo et al., 2000} ; ^{Specht and Corseuil, 2002} ). Additionally, understanding the preference of the phenological stages by the species promotes effective pest management strategies ( ^{Moscardi et al., 2012} ; ^{Zuffo et al., 2015} ).

In an attempt to reduce the impact of key Lepidopteran pests in soybean crops, several regions of Brazil have increased the use of transgenic varieties expressing the Cry1Ac toxin. This toxin, derived from a soil bacteria B. thuringiensis (Bt), was incorporated in most important crops providing a strategy for agricultural pest management, with substantial economic and environmental benefits. Specifically, in Brazil soybean crops Cry1Ac toxin targets the velvetbean caterpillar - Anticarsia gemmatalis (Hübner, 1818), the soybean looper Chrysodeixis includens (Walker, [1858]) and species belonging to the Heliothinae subfamily. Soybean crops expressing the Cry1Ac toxin reduce insect pest population densities and minimize the use of insecticides ( ^{Roh et al., 2007} ).

To ensure the effectiveness against target insects, Bt crops need to be managed in a manner that prevents the development of resistant populations ( ^{Bates et al., 2005} ; ^{Gould, 1998} ). The strategies used to minimize insect resistance include the use of the high doses with plants expressing high levels of the toxin and the adoption of structured refuges ( ^{Gould, 1998} ). In addition, the adoption of Bt technologies can trigger changes in pest status by suppressing the populations of the main target species ( ^{Catarino et al., 2015} ).

Given the great diversity of Brazilian habitats, the community of owlet larvae is not well-known in the different soybean producing states, making it difficult to establish a sustainable management control strategy considering the different crops, expressing the Bt toxin or not. For the Cerrado Biome, there were developed soybean cultivars characterized by rusticity and grain yield potential were developed, specially adapted for areas that are more prone to high temperatures and scarcity of water.

This study is based on the hypothesis that the expression of the Cry1Ac toxin should induce a qualitative-quantitative differentiation of lepidoptera associated with non-Bt and Bt soybeans, especially on the target noctuids of the technology. The objectives of this study were to compare (A) richness, abundance and temporal variations of Noctuoids in different phenological stages on non-Bt and Bt soybeans, and (B) to correlate the meteorological variables with the results of the three most abundant species of Noctuoids.

2. Material and Methods

3. Results

In total, 1,150 larvae were collected. They were distributed into 13 species, nine genera and six subfamilies of Erebidae and Noctuidae. The majority of the species were collected from non- Bt soybeans. Agrotis ipsilon (Hufnagel, 1766), Elaphria deltoides (Möschler, 1880), Mocis latipes (Guenée, 1852), Spodoptera albula (Walker, 1857) and Rachiplusia nu (Gueneé, 1852) were not collected from Bt soybean ( Table 2 ).

The number of larvae collected from non-Bt soybeans corresponded to 92% of the individuals ( Table 2 ). This difference was significant in the V₂₀-V₂₁ to R_6, stages, when a population increase with a peak of 10 larvae per square meter in R₄ -R_5.1 ( Table 3 ) ( Figure 1 A).

Figure 1 Average number (larvae/m²) of Chrysodeixis includens , Anticarsia gemmatalis and Spodoptera spp. ( S. albula, S. cosmioides, S. eridania and S. frugiperda) collected from non-Bt -BRS Sambaíba RR (A) and Bt - BRS 9180 IPRO RR soybean (B). Between December 2015 and March 2016, in the experimental area at Embrapa Cerrados, Planaltina, Distrito Federal, Brazil.  

The most abundant species in both cultivars was C. includens (soybean looper). However, on non-Bt soybeans the average number of C. includens larvae was almost 14 times greater than on Bt soybeans. The soybean looper represented 74% of total number of larvae of all species found on non-Bt soybeans, while on Bt soybeans it accounted for 58% of the entire sample ( Table 2 ). This species was the most influential on the average numbers of larvae collected in the study. In the reproductive stage R₄-R_5.1 of the non-Bt soybeans, there was a peak population of 7.7 soybean looper larvae/m² (p <0.001) ( Figure 1 A) ( Table 3 ).

Anticarsia gemmatalis (velvetbean caterpillar) was the second most abundant species on both Bt and non-Bt soybeans. On non- Bt soybeans, the average number of larvae was almost 11 times greater than on Bt soybeans. Larvae of this species were collected from R₃-R₄ . Anticarsia gemmatalis represented approximately 18% of the total number of larvae for all treatments. In the reproductive stages R_5.2 and R_5.3 of the non-Bt soybean, there was a peak population of approximately 1.5 larvae/m ² (p <0.001) ( Figure 1 A, B; Table 3 ).

Larvae belonging to the four Spodoptera species [S. albula (Walker, 1857), S. cosmioides (Walker, 1858), S. eridania (Stoll, 1782) and S. frugiperda (J.E. Smith, 1797)], (n = 60) accounted for just over 5% of total larvae collected. On the non-Bt soybean, the average number of larvae was only three times greater than on the Bt soybeans. Considering all species together, individuals of Spodoptera species represented 4.2% of all larvae found on non-Bt soybeans, while their representation on Bt soybeans was 16.3%. A small number of larvae of this species occurred during the reproductive stages R₃-R₄, and it was significantly higher on Bt soybeans than non-Bt. During the two weeks that followed (R ₄-R_5.1 and R_5.2), the average number of larvae on non- Bt soybeans was significantly higher than on Bt, representing a population of 0.6 larvae/m² (p <0.001) ( Figure 1 ; Table 3 ).

The respective proportion of larvae according to size was similar between cultivars, however, different among species. The distribution by size class was as follows: Chrysodeixis includens: 22% were small, 26% were medium-sized, and 52% were large; A. gemmatalis: 11% were small, 52% were medium-sized, and 37% large; and Spodoptera spp: 15% were small, 55% were medium-sized and 30% were large.

The weekly cumulative precipitation was the only weather parameter correlated with the number of A. gemmatalis larvae (p <0.05) ( Figure 2 ; Table 4 ).

Figure 2 Relationship between average larvae/m² and weekly cumulative precipitation (mm³) in the experimental area at Embrapa Cerrados, Planaltina, Distrito Federal, Brazil. Between December 2015 and March 2016.  

According to the t - test the average yield (40.5 kg per hectare) showed no significant differences (P = 0.680) between non-Bt and Bt soybean.

4. Discussion

The low number of owlet species found in the present work (n=13) contrasts with the 31 noctuid soybean pests reported for South America ( ^{Formentini et al., 2015} ). The reduced diversity found in this study can be explained by the fact that these pest species are sporadic and limited to specific places in southern Brazil and neighboring countries (e.g. Agrotis gypaetina Guenée, 1852; Agrotis malefida Guenée, 1852; Dargida meridionalis (Hampson, 1905); Helicoverpa gelotopoeon (Dyar, 1921); Mythimna adultera (Schaus, 1894); Paracles cajetani (Rothschild, 1910); Paracles vulpina (Hübner, [1825]) and Peridroma saucia (Hübner, 1808)).

Despite this reduced diversity, two species, E. deltoides and Utetheisa ornatrix (Linnaeus, 1758), were observed feeding on soybean for the first time. Thus, there is a need for additional more surveys in different soybean regions were rotation with Crotalaria species (Fabaceae), the major hosts for U. ornatrix ( ^{Sourakov, 2015} ) occurs.

Non-Bt soybeans, represented 92% of the individuals and five species of larvae, which is certainly due to the expression of the Cry1Ac toxin, as it has already been proven to control lepidopteran pests in several countries ( ^{Walker et al., 2000} ).

The maximum density of C. includens larvae/m² was below the control level during the entire period and only one population peak was found between R ₄-R_5.1 and R_5.2, where larvae were collected mainly from the reproductive stage. This differed from several observations on early infestations (three or four weeks after soybean germination), indicating at least two population peaks per harvest ( ^{Conte and Corrêa-Ferreira, 2014} ).The small number of larvae collected on Bt soybean cultivar (with the proportion of 115 collected on non-Bt soybean) indicates the susceptibility of C. includens to the Cry1Ac toxin ( ^{Bernardi et al., 2012} ; ^{Sorgatto et al., 2015} ; ^{Yano et al., 2016} ).

The reduced amount of A. gemmatalis larvae (maximum 1.63/m² ), restricted to the reproductive stages of the soybean, differ with most publications, which document high rates of infestation and defoliation by this species ( ^{Walker et al., 2000} ). The proportion of A. gemmatalis larvae collected on Bt soybeans versus those collected from non-Bt were less than 110 , which certainly illustrates the species susceptibility to the Cry1Ac toxin ( ^{Walker et al., 2000} ). The lower incidence of A. gemmatalis compared to C. includens showed here may support that this species is no longer the most important lepidopteran pest of soybean as suggested by ^{Moscardi et al. (2012)} .

Four of the five species of Spodoptera whose larvae feed on soybean in South America were collected ( ^{Formentini et al., 2015} ; ^{Montezano et al., 2015} ). Spodoptera accounted for only 5% of the total sample, differing from other studies that suggested the increase of this genus in soybean crops in Brazil, both on leaves and pods ( ^{Moscardi et al., 2012} ). More than 50% of the Spodoptera larvae belong to Spodoptera eridania (Stoll, 1782), the most important pests of soybeans and cotton in different agricultural systems ( ^{Santos et al., 2005} ). It is important to keep in mind that Spodoptera are not targeted pests of the Bt technology, which explains the lower susceptibility of Spodoptera to the Cry1Ac toxin ( ^{Bernardi et al., 2014} ; ^{Sorgatto et al., 2015} ).

The larvae sizes differed from the expected, even though the control level was not reached. Medium and large larvae were not expected in a large number for A. gemmatalis and C. inlcudens species. The small size larvae (<1.5 cm) represented the lowest proportion of individuals collected of all species, contrasting with studies on soybeans and other crops ( ^{Specht and Corseuil, 2002} ).

Considering the atypical climate, the low number of individuals collected can be attributed to the El Niño phenomenon that for standardization is characterized by a positive Oceanic Niño Index (ONI) greater than or equal to +0.5 °C for at least three consecutive months. The last episode extended from October 2014 to April 2016, with ONI values above 2.0 between November 2015 to January 2016 ( ^{Null, 2016} ). In the Brazilian Savanna this phenomena determines high temperatures and irregular precipitation with volumes much lower than expected. In the 2015-2016 crop season, 855.2 mm of precipitation was recorded, representing only 61.81% of the 1,383.7 mm expected in the climatological normal ( ^{Silva et al., 2014} ). This event led a decreased yield of grains and cotton in all biome ^{(Brasil, 2016)} . For decades, it has been known that climate has an effect on insects, including Lepidoptera ( ^{Pardikes et al., 2015} ). With contrast there is a lack of knowledge of the consequences of El Niño in different Brazilian regions and its impact on food production and pest insects ( ^{Anderson et al., 2017} ).

The negative correlation between population density of A. gemmatalis and rainfall reinforces the idea that years with low or irregular rainfall can be associated with increased population densities ( ^{Sujii et al., 2002} ). The identification of a correlation between climatic variables and the occurrence of Lepidoptera would be desirable to forecast the occurrence of each species based on current and future weather conditions.

This study documented the efficacy of the Bt technology and the Cry1Ac toxin for the management of target species, such as C. includens and A. gemmatalis, corroborating the reports of different authors after laboratory and field studies ( ^{Walker et al., 2000} ; ^{Bernardi et al., 2014} ; ^{Sorgatto et al., 2015} ; ^{Yano et al., 2016} ).

Nevertheless it is recommended that comparative studies should be conducted to test the effectiveness of different, non-suppressive, sampling methods compared to absolute sampling. ( ^{Didonet et al., 2003} ; Conte and Correa-Ferreira, 2014; ^{Cabral Antúnez et al., 2017} ).

Despite the differences detected between the numbers of larvae collected in both cultivars ( Tables 1 and 2 , Figures 1 and 2 ), the productivity of non-Bt and Bt soybean was statistically equal, even without the use of any insecticide during the entire soybean development cycle. Even with 10 times more larvae per square meter found in non-Bt soybean, no differences in productivity were observed, indicating that the total number of larvae did not reach the level of economic damage ( ^{Hoffmann-Campo et al., 2000} ; ^{Conte and Corrêa-Ferreira, 2014} ). This finding reinforces the need to monitor the number of larvae throughout the soybean to rationalize the use of insecticides.

In conclusion, this study shows the differential abundance/occurrence and susceptibility of the main representatives of Noctuoidea associated with soybeans demonstrating the importance of field monitoring to assist on decision-making, and to assess the effectiveness of the technology. In addition U. ornatrix and E. deltoides were identified for the first time on soybean ( ^{Formentini et al., 2015} ). Overall, this type of study allows for the assessment of the status of each species in the agro-ecosystem. The evaluation of target pests, which are kept under control by the Bt technology, and the monitoring of secondary pests or other species that may use the available resource, can assist in the early detection of outbreaks. Therefore, under the particular conditions of this study, the negative correlation between population density of A. gemmatalis and rainfall was found.