The efficacy of spatio-temporal predictors in forecasting the risk of Cydia nigricana infestation

doi:10.21203/rs.3.rs-2978102/v1

Download PDF

Research Article

The efficacy of spatio-temporal predictors in forecasting the risk of Cydia nigricana infestation

https://doi.org/10.21203/rs.3.rs-2978102/v1

This work is licensed under a CC BY 4.0 License

Journal Publication

published 15 Mar, 2024

Read the published version in Entomologia Experimentalis et Applicata →

Version 1

posted

You are reading this latest preprint version

The ability to estimate the risk of pest infestation can help cultivators to reduce pesticide application and provide guidance that would result in better management decisions. This study tested whether different combinations of spatial and temporal risk factors can be used to predict the damage potential of Cydia nigricana, a major pest in field pea (Pisum sativum). Over four consecutive years, the abundance of pea moth was monitored by placing pheromone traps at different field pea cultivation sites. We also assessed the phenological development stages and the percentage of damaged seeds per 100 pods collected from each growing pea field in a region of approximately 30 km in diameter. The study found the significant infestation risk indicators to be the time of flowering, the date on which male pea moths are first detected in the monitoring traps, and the minimum distance (MD) to pea fields that were planted and harvested in the previous growing season. The combination of all three factors using a general additive model (GAM) approach yielded the best results. The model proposed by this study accurately discriminated between low-infestation and high-infestation fields in 97% of cases.

Lepidoptera

Tortricidae

risk forecast

spatio-temporal risk factors

infestation pressure

Pesticide is the main option for the control of the pea moth, Cydia nigricana, a major pest in peas (Pisum sativum). The ability to estimate the risk of pest infestation can help cultivators to reduce pesticide application and provide guidance that would result in better management decisions.
In this study we tested whether different combinations of spatial and temporal risk factors can be used to predict the damage potential of Cydia nigricana, a major pest in field pea (Pisum sativum).
As the main result this study presents a preliminary combined model for the prediction of infestation risk of C. nigricana, as a function of crop phenology, pest emergence and insect movement. The model accurately discriminated between low-infestation and high-infestation fields in 97% of cases, which opens the possibility to forecast if an application of insecticides is profitable for a farmer or not.
The findings can contribute to the improvement of environmental control and facilitate good management decisions by farmers, thus reducing the use of insecticides in pea cultivation.

Within the framework of the European Union’s Farm to Fork strategy, the use of chemical pesticides and the risks they pose must be reduced by 50% by 2030 (European Comission 2022). Pesticide resistance, residue-related issues and negative environmental impacts have led to a reconsideration of alternative practices that focus on preventive damage reduction or, at least, reduced pesticide use. One such approach is integrated pest management (IPM). Pest forecasts play an important role in IPM. The predictions made by forecasts can facilitate important management decisions, such as the timing of insecticide application or the development of preventive procedures, thus optimising pest control (Prasad and Prabhakar 2012; Dent and Binks 2020)

Pesticide is the main option for the control of the pea moth, Cydia nigricana, a major pest in peas (Pisum sativum). However, the field efficacy of pesticides is variable due to the moth’s biology. The larvae feed on the pea seeds inside the pod. During this time, the larvae are protected from insecticidal sprays, reducing the efficacy of the sprays and making the direct control of the pea moth challenging. C. nigricana adults overwinter as larvae inside a cocoon in the soil of previous year pea fields and emerge as adults the next spring. Many of them mate immediately (Nicolaisen 1928; Langenbuch 1941; Stenmark 1974; Sarwar 1973). Pisum sativum is the preferred host of C. nigricana and is located by olfactory cues. Mated C. nigricana females in particular are attracted to volatiles emitted during the development of buds and flowers (Thöming et al. 2014; Thöming and Norli 2015).

Recommendations for insecticide application to control pea moth are based on threshold values. Threshold values are fundamental in IPM strategies (Stern et al. 1959; Pedigo et al. 1986). A threshold is reached when pest population pressure is considered to be economically injurious and requires direct pest control. Spraying is recommended when a value of 10 male moths per pheromone trap per day is reached (Touvinen 1982). However, the proposed number of 10 moths per trap is a coarse estimate that is not based on a great deal of scientific evidence. The number of caught male moths does not necessarily reflect the number of moths per hectare and can vary depending on weather conditions, type of trap and surrounding infestation pressure (Touvinen 1982). Identifying factors that can be used to forecast pest infestation risk could therefore contribute to a more reliable action threshold.

According to Huusela-Veistola and Jauhiainen (2006), damage to grain peas decreased exponentially when the minimum distance (MD) to the nearest pea field harvested the previous year increased. The authors also found that the pea moth abundance in the pea field increased linearly with the area under pea cropping in the previous year (continuous abundance index - CAI). The researchers therefore recommended a distance of 1.5 km between overwintering sites and new pea fields to minimize the risk of pea-moth infestation. Therefore, in addition to taking threshold levels into account, spraying only those fields that are located less than 3 km away from a previous year’s crop is being recommended (ISIP 2023; Jao 2011). However, the success of this strategy seems to depend on the time and duration of pea flowering (Thöming et al. 2011). Early-flowering and short-flowering pea varieties were particularly associated with lower C. nigricana damage (Hanson and Webster 1936; Nolte and Adam 1962). For early-sown and early-flowering green peas that are harvested when immature, an MD of only 500 m was sufficient to considerably reduce infestation pressure (Thöming et al. 2011). In contrast to Huusela-Veistola and Jauhiainen (2006), Thöming et al. (2011) did not find a relation between pea-moth damage and MD for grain peas. Thus, while the potential of distance (space) and flowering time to interfere with host-pest occurrence is known, a more holistic concept that embeds both predictors in a forecasting approach is potentially of great value. Such an approach could provide important insights into preventive risk avoidance by combining situation-adapted cropping strategies.

This study explores if and to which extent a combination of spatial and temporal factors can predict the infestation risk of a given pea field in the region that was modelled. The following spatial risk factors are considered: (i) MD between the previous and the current pea location, and (ii) the CAI (previous pea area around the current field) The temporal risk factors were (i) the onset of flowering (BBCH 60) of pea fields, and (ii) the time at which pea moths began appearing in the traps. Given that the development of plants and insects mainly depends on temperature, this study describes the cultivation period in heat units (growing degree days). Finally, generalised additive models (gam) are used to identify decisive factor combinations.

Field experiments

Location

The data was collected over four consecutive years between 2015 and 2019 in a region of approximately 30 km in diameter in North Hesse, Germany (51°21’26.23’’N; 9°49’37.83’’O). This area is characterised by small-scale agriculture and features relatively small pea sites (0.01–15.44 ha). Furthermore, the area has a high percentage of unsprayed fields, which is a prerequisite for assessing risk without interference by pesticide application. Another reason for the selection is the high diversity of winter and summer pea varieties grown in this area. The different sowing dates offer a wide range of flowering times.

The pea fields were located using information obtained from Wirtschafts- und Infrastrukturbank Hesse and local farmers. From 2015 to 2019, we mapped all known pea fields within the study area using Quantum GIS, version 3.4. We interviewed the farmers about sowing dates and the pest management methods they use.

Measurements

From 2016 to 2019, we monitored the pea-moth flight in all 162 of the identified pea fields. From 2017 to 2019, we also monitored 36 selected winter cereal fields with peas as preceding crops as emergence sites. We used commercial sex-pheromone delta traps for this monitoring process (Trifolio-M GmbH, Lahnau, Germany). One trap per site was placed 30 m inside the edge of the field. Trap catches were recorded every third day in all the previous pea fields and once a week in all the current pea fields. Male moth flight was monitored throughout the entire flight period starting in calendar week 18, before flight start until pea harvest.

We monitored the phenological development stage of each pea field at each visit based on the BBCH scale (Meier et al. 2009). Starting in 2016, the pea-moth infestation level for every site was determined shortly before harvest. We assessed the number of infested seeds per 100 pods. The pods were randomly collected along a transect of approximately 50 m, starting after the field edge. In total we assessed 162 pea fields over four years. However, to prevent direct control measures from interfering with the results, only data from unsprayed fields was used for modelling. In addition, we only included fields in the model when all three parameters (distance, first moth occurrence and flower onset) were assessable. Consequently, we used 88 of the 162 fields for risk modelling.

Meteorological data

The meteorological data base used for the interpolation of hourly air temperature data (2 m above ground) for each monitored site was collected by automatic stations. These stations are operated by the German Meteorological Service (DWD) and Crop Protection Services (CPS) of the federal states. For the interpolation of temperature, a multiple regression analysis including the parameters ‘altitude’, ‘exposure’ and ‘slope inclination’ was used to calculate data for 1 km² grids. The interpolation process is described in detail by Zeuner (2007).

Data analysis

To choose relevant factors that affected pea-moth infestation, we first investigated the influence of the factors distance, first male moth occurrence in pheromone traps and onset of pea flowering separately. We applied two different indices to describe the spatial relation between pea field location and infestation: the CAI and the MD (Huusela-Veistola and Jauhiainen 2006). We used growing degree days with a developmental threshold of zero to describe the relation between onset of flowering, time of the first pea moth arrival and infestation level using the following equation:

$GDD= \sum _{i=1 }^{n}\frac{ \sum _{24}^{1}\left({T}_{hour}- {T}_{min}\right)}{24}$ ) (1)

with T_min = 0°C and T_hour = the temperature every full hour, with 1 January as the biofix.

CAI

The CAI displays the abundance of pea fields grown the previous season around the pea field that is currently growing. For the CAI we summed up the area of all previous year's pea fields within the range of a 0.5 km, 2 km and 4 km radius around the current pea field and divided by the total area (Huusela-Veistola and Jauhiainen 2006). We used a simple linear model to describe the relationship between CAI and infestation rate.

The minimum distance model describes C. nigricana infestation risk as a function of MD between the current pea field and the closest pea field grown the previous year. We calculated the MDs using the nearest neighbour analyses in ArcGIS, Esri version 10.6. Following the approach of Huusela-Veistola and Jauhiainen (2006) we used Eq. 2 to model the infestation risk (r₁)-minimal distance (MD₁) relation for each year and for all years.

$${r}_{1}={{a}_{1} exp}^{\left(-\frac{MD}{{b}_{1}}\right)}$$

where b₁ is the slope parameter and a₁ is the infestation risk of a nonrotated or directly adjacent field. The model was assumed to result in a statistically significant fit when the value of a₁ or b₁ was significantly different from zero.

First moth occurrence

Like the MD model we used a negative exponential model to describe the time of first moth occurrence (moth) in the traps Eq. 3. For GDD calculation we applied Eq. 1

$${r}_{2}={a}_{2} {exp}^{\left(-\frac{moth}{{b}_{2}}\right)}$$

with a₂ and b₂ as the slope parameters. The model was assumed to result in a statistically significant fit when the value of a₂ or b₂ was significantly different from zero.

Blooming

We calculated the GDD until the pea fields reached the BBCH of 60 using Eq. 1. We modeled the relationship between flower time and pea moth by a general additive model (GAM) from the mgvc package. The general form of gam is presented as Eq. 4:

$${Y}_{i}= \alpha +f\left({X}_{i}\right)+{\epsilon }_{i}$$

where Y_i is the value of the response variable of sample i and X_i is the explanatory variable. The smoothing function is f(). We applied a regression spline with fixed degrees of freedom and checked for concurvity for all factors using the concurvity function Model fitness was assessed by gam.check function. We controlled for overfitting visually using visreg and vis.gam and adopted the degrees of freedom accordingly. All functions are either part of the R mgvc package (Wood 2022) or belong to the visreg package (Breheny and Burchett 2017).

Final model formulation

Gam was also used to model the relationships between infestation rate, florescence, distance and first moth occurrence. Model validation was carried out as described above. Assuming that flower time and pea moth arrival are nonlinearly correlated we included an interaction between those terms.

Model validation

For gam model validation of the two final models we used a simple linear regression and the leave-one-out cross-validation (LOOCV;(Holmes 2021). In addition, we evaluated the ability of the models to predict the infestation correctly within a certain range. The prediction was considered correct if the observed value and the predicted value ranged between two infestation level categories with 0–13% (low infestation field) or above 13% (high infestation field). These two categories were based on economic considerations, which are described as follows:

Threshold calculation

Damage caused by pea moths reduces the thousand-grain weight of damaged grain peas to 71% in comparison to non-damaged peas (Visotsching 1998). If the harvest is sold as fodder, the quality of the grains is not important – damaged grains can be sold. Starting from a selling price of 19,2 €/dt and an average yield of 34,5 dt/ha (KTBL 2022), an infestation level of 1% would result in a yield loss of 0,10 dt/ha, which amounts to 1,92 €/ha. To calculate the costs of an insecticide application in fodder peas we also used KTBL data (KTBL 2022). We included machine, labour and fuel costs (in total, 14,64 €/ha), as well as the costs for the insecticide Karate Zero® in the necessary application quantity (126,91 €/l;(BayWa 2022), resulting in 24,11 €/ha as total treatment costs. Based on a cost-benefit ratio of 1:1, one insecticide application in grain feed peas is advisable only if pea-moth infestation exceeds 13% damaged peas.

Table 1 presents an overview of the yearly mean infestation levels and the infestation range. The highest mean infestation rate, at 16.39%, and the highest individual infestation level, at 71.87% damaged grains, occurred in 2018. In 2016 and 2019, the mean infestation rate was comparatively low, at 2.62% and 4.44%, respectively (Table 1).

Table 1

Average pea-moth infestation level and infestation range from 100 randomly collected pea pods per field of organic and conventionally managed pea fields in a region 30 km in diameter in Northern Hesse (2016–2019). No pesticides were used on these fields in the respective years
Pea-moth infestation (%)	2016	2017	2018	2019
Mean (± sd) Range (min – max)	2.62 ± 1.76 0.00–8.42	11.33 ± 12.62 1.09–44.08	16.39 ± 15.88 0.00–71.87	4.44 ± 3.50 0.22–13.17
pea sites (n)	17	30	28	13

Pea-moth emergence at overwintering sites

Pea-moth emergence was assessed as the mean trap catch of male pea moths starting at 819.46 GDD. At 1202.03 GDD, 75% of the total moth counts were achieved. The mean number of moths caught per trap did not differ significantly in all three years (Table 2).

Table 2

Mean growing degree days (GDD) and moth capture at 36 emergence sites (previous pea fields) in Hesse 2017–2019
Moth eclosion (%)	Mean GDD ± se	Range
first moth in trap	819.46 ± 18.41	666–1102
10	940.54 ± 16.46	708–1146
30	1030.70 ± 15.88	807–1245
50	1092.46 ± 15.90	867–1297
75	1202.03 ± 20.21	920–1496
100	1890.38 ± 53.35	1199–2597

Effect flower onset (BBCH 60) on the infestation rate

We found a significant relationship (Table 3) between flower onset and infestation level, with the highest mean infestation of 36.25% (range: 13.06–71.87%) at sites where flowering started between 819.46 GDD (mean GDD when first moth emerged at previous year’s pea fields) and 940.54 GDD (mean GDD when 10% moth had been detected in previous years pea fields; Fig. 1, Table 2). In 2016 and 2019, at low mean infestation levels of 2.62 and 4.44 (Table 1), no flower onset occurred during this ‘period of high infestation pressure’. Pea fields flowering before 819.46 GDD and those that flowered after 940.54 GDD showed a considerably lower mean infestation (5.91%) ranging from 0–25.6%, with a smaller, second infestation peak at 1120.49 GDD in 2018 only (Fig. 1).

The highest infestation rates corresponded with early season occurrence of the first moth in the actual currently growing pea fields in a range between 694 GDD and 831 GDD and decreased exponentially with increasing GDD (a = 7969.40, p = 0.117; b = 134.95, p < 0.001, R²=0.559; Fig. 2). Even though, in 2019, some pea fields started flowering between 605.69 and 753.33 GDD, in 2016 and 2019 we found no moths in the traps before the accumulation of 890.00 GDD, resulting in low infestation levels (Table 1, Fig. 2).

Field distance and infestation risk models

When the data of all four years was included in the analysis, the CAI did not significantly affect the infestation level. The CAI was therefore not included in the final model.

The relationship between infestation and distance calculated by the MD model, viz. Eq. 2, showed the highest fit in 2017 (R² = 0.79) with one significant parameter. A less good fit was found for 2018 (R² = 0.28) with both parameters being significant. In contrast, the results for 2016 and 2019 showed the poorest fit, as demonstrated by the R² values of R² = 0.09 and R² = 0.18, respectively (Fig. 3). The final result is an overall fit for four years of R²= 0.233 and all parameters significant (a = 23.38, b = 405.97).

Model prediction quality

We predicted the infestation rate using different combinations of the factors onset of flowering, MD’ and ‘first occurrence of the moth in trap’. All tested factors were significant in all tested models. The linear regression between predicted and observed infestation rate was significant (p < 0,001) in all models, with an R² ranging between 0.58 and 0.79 (Table 3). Model M4, which includes all three factors, had the lowest AIC of 599.85 and RSME of LOOVC-7.16 compared to the other models. However, Model M4 had a slightly higher LOOVC-MAE (4.70) compared to Model M3 (Table 3).

When the ability of the models to correctly assign the fields to infestation groups of either low or high infestation is considered, Model M4 yielded the best results. M4 classified 96.59% of the fields correctly (Table 4). Only two fields were classified as low infestation fields although they had an infestation rate of more than 13%. The forecast was less accurate when the first occurrence of moths in traps was not included in the models. Model M1, which has flowering time as the only explanatory variable, misclassified 11 fields (12,50%) as low infestation fields (Table 4).

Table 3

Model parameters and goodness of fit estimators of four gam models fitted to the pea-moth infestation data of 88 pea fields in North Hesse (2016– 2019)
model	Parameter^a	k^b	P	Predicted vs. observed R²	AIC	LOOVC MEA	LOOVC RSME
M1	Flower	10	< 0.001	0.58	642.86	6.46	9.31
M2	MD Flower	4 10	< 0.001 < 0.001	0.67	627.62	6.17	8.58
M3	moth* flower	-1	< 0.001	0.77	605.02	4.69	7.24
M4	MD moth* flower	4 -1	0.0217 < 0.001	0.79	599.85	4.70	7.16
^a Flower = onset of flowering (BBCH 60) in GDD, moth = arrival of the first moth in the current pea field in GDD, MD = minimum distance from the current pea field to the closest overwintering site of the moth. ^b simple regression spline with fixed degrees of freedom was applied

Table 4

The prediction quality of four gam models to forecast pea-moth infestation risk in 88 fields in North Hesse (2016–2019) in a range of 0–13% and above 13% damaged pea seeds
Model	overestimated (%)	predicted correctly (%)	underestimated (%)
M1 M2 M3 M4	2.27 (n = 2) 5.68 (n = 5) 1.14 (n = 1) 1.14 (n = 1)	85.23 (n = 75) 86.36 (n = 76) 95.45 (n = 84) 96.59 (n = 85)	12.50 (n = 11) 7.95 (n = 7) 3.41 (n = 3) 2.27 (n = 2)

Pest infestation levels are subject to various biotic and abiotic factors that impact on the population dynamics (Khaliq et al. 2014). The coincidence of risk factors in space and time is decisive in either raising or lowering damage potential an insect pest. This coincidence can be favorably exploited in environmental control (Teetes 1991). This study tested the combined effects of three different risk factors, (i) flowering of the host plant, (ii) arrival of first male moths in pheromone monitoring traps and (iii) crop distance, as potential predictors, defining C. nigricana pest status in peas. Low infestation was mainly associated with the co-occurrence of an asynchronous flower onset outside a certain GDD range, late arrival time of the first moths in the actual fields, and an increased distance between the previous and the currently growing pea field, as expressed by MD.

Infestation risk was significantly related to the onset of flowering (BBCH 60). Pea varieties that started flowering specifically in a GDD range of about 900 GDD showed the highest attack levels, exceeding 40% (Fig. 1). This fact also explained the low infestation in the first (2016) and final year (2019) of the study, when all monitored fields started flowering only before or only after this particular GDD timespan. This flowering pattern resulted in a significantly lower mean infestation when compared to 2017 and 2018. During flower development, pea plants are olfactorily more attractive to mated C. nigricana females than during other development stages, for example, when leaves or pods are forming (Thöming et al. 2014). In line with the preference-performance theory for phytophagous insects, female pea moths’ preference reflects offspring performance (Gripenberg et al. 2010). The arrival of pea moths in pea fields during this critical developmental stage is important because pod formation starts shortly thereafter, and pods are the nutrient source for C. nigricana larvae. The importance of flowering time and duration for potential pea-moth damage has been reported by several authors (Hanson and Webster 1936; Nolte and Adam 1962). As an environmental control measure (Pimentel and Goodman 1978; Teetes 1991), early sowings in combination with the use of an early flowering variety was the most recommended approach to prevent pea-moth damage in green vegetable peas (Wright et al. 1951; Thöming et al. 2011). Conversely, a similar effect was reported for sowing dates that were significantly late (Anonymous 1948). Thus, even when only flower onset was included in our model as a single factor, 86% of the fields in high- and low-infestation areas were classified correctly. This finding highlights the importance and value of flowering time as an infestation risk predictor (Table 5, M1).

Furthermore, the study found that the arrival of the first male moth in the currently growing pea fields turned out to be a further indicator for pest-host synchronization. The extent of pea-moth attack decreased exponentially with a later arrival time (Fig. 2). Infestation risk was high only when the male moth arrived early in the season, which occurred in 2017 and 2018. Our findings suggest that fields that have attracted male moths earlier in the season are likely to accumulate more individuals over time, resulting in higher damage. Fields that started flowering significantly early or late were characterised by the late arrival of moths in monitoring traps towards the end of the season. This scenario resulted in a low percentage of damaged grains. Therefore, infestation was highest if the susceptible development stage of the plant coincided with the beginning of the emergence period. For most crops, the time of infestation in a susceptible plant growth stage and yield loss are closely related (Awuni et al. 2015; Bardner and Fletcher 1974). The results of this study confirm that the same applies to C. nigricana: A favourable match in space and time between crop and pest-population development is a prerequisite for high damage.

However, in this study, not all the fields with a high infestation rate were flowering within the GDD of mean early pea-moth emergence. The dispersion of the insect in space is likely another important factor. In contrast to the results of Thöming et al. 2011) and Huusela-Veistola and Jauhiainen 2006), CAI was not significantly correlated with damage in this study. Therefore, CAI was not included in the study’s final model. Minimum distance significantly affected the infestation rate only in the high infestation years, 2017 and 2018 (Fig. 3). In those two years, a negative exponential decrease in infestation rate with increasing distance to previous year’s pea fields was observed. This finding is in accordance with that of Huusela-Veistola and Jauhiainen 2006). No correlation was found in low infestation years, which explains the poor fit overall.

Combining MD and flower time in one model increased the prediction quality significantly. Pea fields closely located to the overwintering sites, and thus characterized by low MDs, were substantially infested only when they were also olfactorily most attractive, which is from bud development until the end of flowering (Thöming and Knudsen 2014). This means that if a pea site is closer than another but not flowering, the directional movement of the moths would be towards the more distant but flowering pea site, resulting in a higher infestation that is independent from MD. In the case of the leaf-defoliating Colorado potato beetle, which is less dependent on a certain plant development stage for offspring survival than the pea moth, host location is less directional and less stage specific.Weisz et al.(1996) were therefore able to model the intensity of attack as a function of only the migratory distance with much greater precision than it was possible in this study. The wide range of flowering patterns in our study region and the related variable olfactorial attraction negatively affected CAI- as well as MD-only based damage predictions. Spatial risk factors might lead to better results in environments with a more synchronized crop phenology, which applies to either early vegetable pea varieties or grain pea cultivars being more synchronized at scandinavian agroclimatic conditions described in the studies ofThöming et al.(2011) and Huusela-Veistola and Jauhiainen, (2006), respectively.

Combining all three of the factors discussed above was essential and yielded the best model results. Information about flowering time, pea-moth arrival and MD makes it possible to discriminate between high- and low-infestation fields with greater precision. However, these results are preliminary and, at this stage, implementation of the model requires data monitoring of the fields, which is time intensive. Farmers do not have the required time to perform such monitoring. Thus, for the practical application of the model, the future development of digital data-acquisition processes is of major importance. In recent years, the technological means of remote sensing have expanded rapidly in agriculture. Crop classification methods using satellite data are already widely studied (Sun et al. 2018; Meng et al. 2021; Wang et al. 2022b). The automatic recognition of pea fields based on satellite imaging would allow for MD calculations without the need to involve farmers. In addition, even without including MD, the model suggested by this study correctly classified 97% of the fields. Schieler et al. (subm. unpublished) developed a temperature- and photoperiod-based model that can predict development stages, including florescence, based on sowing date and pea variety. Furthermore, in the future, the practical implementation of satellite remote sensing in predicting crop phenology can facilitate the recognition of a particular growth stage, like florescence, without requiring actual field visits (Gao and Zhang 2021).

The detection of the first male moth arrival was of great significance for the quality of the proposed model. The manual monitoring of pheromone traps would require at least weekly control of all currently growing pea sites subsequent to moth emergence in previous pea sites. A temperature-based model that predicts pea-moth emergence already exists and can help to restrict the time span needed for placing and maintaining pheromone monitoring traps in currently growing pea sites (Riemer et al. 2021). Automatic pest-counting insect traps using deep-learning techniques (Bjerge et al. 2021; Hong et al. 2021tő 2021; Wang et al. 2022a) would be a substantial improvement on these traps, but they are not market ready yet.

Limitations and recommendation for future research

This study faced certain limitations. Although weather data was interpolated from a height 2 m, reproducible results were achieved. However, further microclimatic refinements would produce greater precision. Data collected over only four years was included in the model, and the number of fields was reduced by almost half due to the exclusion of fields that received insecticidal treatment. The limited data can influence the precision of the model, which therefore requires validation by independent datasets of successive years. This study’s data is also restricted to the local conditions of one model region. Therefore, the data requires validation before generalizing it to other agroecological pea-cultivation areas. Furthermore, this preliminary model was developed with data from grain peas; the prediction needs to be refined and expanded to other grain-pea and vegetable-pea varieties.

Regarding the factor ‘first male moth trapped’, two aspects must be pointed out: First, even detecting just the male fraction of the pest population resulted in reproducible predictions. Nevertheless, predictions could be improved in the future by addressing the fact of specifically mated females being more attracted to the floral odor of flowering peas than males (Thöming and Knudsen 2014). Monitoring the females with kairomone-baited monitoring traps could improve the prediction results, in analogy to Cydia pomonella monitoring (Knight 2010; Knight and Light 2005; Light et al. 2001). when this study was conducted, such traps containing pea-plant volatiles (kairomones) were not commercially available yet. Second, Pisum sativum is the preferred and most abundant host plant within the model region in terms of acreage. However, other cultivated and non-cultivated grain legumes, such as Vicia faba, Vicia sepium, Laturus odoratus, Orobus sp. and others, have also been reported as potential host plants (Wright and Geering 1948; Hoffmann and Schmutterer 1999; Kruess and Tscharntke 2000). Although the reproductive potential in alternate hosts can be regarded as comparably low in general (Thöming and Norli 2015), the extent to which alternate host plants may affect the forecasting value of first moth arrival in monitoring traps and the factor MD remains unknown.

The adverse effects of pesticides on the environment and human health are numerous and widely documented. Nevertheless, in Germany, the application of pes insecticides in the past 25 years has not decreased (BVL 2021). For the effective use of insecticides, pest-host interactions should be studied extensively. However, such studies are often lacking. A lack of confidence in the outcome can lead to the unnecessary application of insecticides, which is both costly and harmful to the environment. This study presents a preliminary combined model for the prediction of infestation risk as a function of crop phenology, pest emergence and insect movement. The findings can contribute to the improvement of environmental controls and facilitate good management decisions by farmers, thus reducing the use of insecticides in pea cultivation.

Authors Contribution Statement

Natalia Riemer and Helmut Saucke mainly designed the study. Natalia Riemer carried out the experiments. The data was statistically analysed by Natalia Riemer with contribution of Manuela Schieler and interpreted by Natalia Riemer, with great support from Helmut Saucke. Natalia Riemer wrote the first draft of the manuscript. Helmut Saucke secured the funding. All authors read and approved the manuscript.

Acknowledgments

The work is supported through funding from the Federal Ministry of Food and Agriculture (BMEL), based on a decision of the parliament of the Federal Republic of Germany via the Federal Agency for Agriculture and Food (BLE) under the Federal Program for Ecological Farming and other forms of Sustainable Agriculture (BÖL) (project no. 2812OE034, CYDNIGPRO). We would like to thank Markus Hammer-Weis, Janina Ridthaler, Fabian Stampa, Lena Franke and Sandra Carnaghi for sample collection.

Funding

Competing interests

The authors have no relevant financial or non-financial interests to disclose.

Author contributions

Helmut Saucke and Natalia Riemer mainly designed the study. Natalia Riemer carried out the experiments. The data was statistically analysed by Natalia Riemer with contribution of Manuela Schieler and interpreted by Natalia Riemer, with great support from Helmut Saucke. Natalia Riemer wrote the first draft of the manuscript. Helmut Saucke secured the funding. All authors read and approved the manuscript.

Data availability

The datasets generated during and analysed during the current study are available from the corresponding author on reasonable request.

Anonymous (1948) Biology and Control of the Pea Moth. Nature 162:309–310. https://doi.org/10.1038/162309b0
Awuni GA, Gore J, Cook D, Musser F, Catchot A, Dobbins C (2015) Impact of Oebalus pugnax (Hemiptera: Pentatomidae) Infestation Timing on Rice Yields and Quality. J. Econ. Entomol. 108:1739–1747. https://doi.org/10.1093/jee/tov123
Bardner R, Fletcher KE (1974) Insect infestations and their effects on the growth and yield of field crops: a review. BER 64:141–160. https://doi.org/10.1017/S0007485300027061
BayWa (2022) Syngenta KARATE® ZEON. https://www.baywa.de/de/pflanzenbau/getreide/pflanzenschutz/insektizide/syngenta-karate-zeon-/pp-6875718/. Accessed 24 May 2022
Bjerge K, Nielsen JB, Sepstrup MV, Helsing-Nielsen F, Høye TT (2021) An Automated Light Trap to Monitor Moths (Lepidoptera) Using Computer Vision-Based Tracking and Deep Learning. Sensors 21:343. https://doi.org/10.3390/s21020343
Breheny P, Burchett W (2017) Visualization of regression models using visreg. R J 9:56–71
BVL (2021) Absatz an Pflanzenschutzmitteln in der Bundesrepublik Deutschland: Ergebnisse der Meldungen gemäß § 64 Pflanzenschutzgesetz für das Jahr 2021. www.bvl.bund.de/psmstatistiken. Accessed 28 April 2023
Dent D, Binks RH (2020) Insect pest management. CABI Publishing, [S.l.]
European Comission (2022) Farm to Frok - Progress. https://ec.europa.eu/food/plants/pesticides/sustainable-use-pesticides/farm-fork-targets-progress_en. Accessed 23 May 2022
Gao F, Zhang X (2021) Mapping Crop Phenology in Near Real-Time Using Satellite Remote Sensing: Challenges and Opportunities. J. Remote Sens. 2021:1–14. https://doi.org/10.34133/2021/8379391
Gripenberg S, Mayhew PJ, Parnell M, Roslin T (2010) A meta-analysis of preference-performance relationships in phytophagous insects. Ecol. Lett. 13:383–393. https://doi.org/10.1111/j.1461-0248.2009.01433.x
Hanson AJ, Webster RL (1936) The pea moth: Laspeyresia nigricana Steph. J. agric. res. Wash. D.C. 327:1–22
Hoffmann GM, Schmutterer HJ (1999) Parasitäre Krankheiten und Schädlinge an landwirtschaftlichen Kulturpflanzen: 14 Tabellen, 2nd edn. Ulmer, Stuttgart
Holmes E (2021) loogam: Leave one out cross-validation on gam or lm object. https://rdrr.io/github/eeholmes/SardineForecast/man/loogam.html. Accessed 20 May 2022
Hong S-J, Nam I, Kim S-Y, Kim E, Lee C-H, Ahn S, Park I-K, Kim G (2021) Automatic Pest Counting from Pheromone Trap Images Using Deep Learning Object Detectors for Matsucoccus thunbergianae Monitoring. Insects 12:342. https://doi.org/10.3390/insects12040342
Huusela-Veistola E, Jauhiainen L (2006) Expansion of pea cropping increases the risk of pea moth (Cydia nigricana; Lep., Tortricidae) infestation. J. Appl. Entomology 130:142–149. https://doi.org/10.1111/j.1439-0418.2006.01047.x
ISIP (2023) Erbsenwickler. https://www.isip.de/isip/servlet/isip-de/infothek/leguminosen/futtererbsen/erbsenwickler#:~:text=Eine%20gezielte%20Bek%C3%A4mpfung%2C%20die%20gegen,BBCH%2069%2F71)%20gegeben. Accessed 6 January 2023
Jao C (ed) (2011) Efficient decision support systems - practice and challenges from current to future. InTech, London
Khaliq A, Javed M, Sohail M, Sagheer M (2014) Environmental effects on insects and their population dynamics. J. Entomol. Zool. Stud 2:1–7
Knight A (2010) Improved monitoring of female codling moth (Lepidoptera: Tortricidae) with pear ester plus acetic acid in sex pheromone-treated orchards. Environ. Entomol. 39:1283–1290. https://doi.org/10.1603/EN10034
Knight AL, Light DM (2005) Seasonal flight patterns of codling moth (Lepidoptera: Tortricidae) monitored with pear ester and codlemone baited traps in sex pheromone-treated apple orchards. Environmental Entomology 34:1028–1035
Kruess A, Tscharntke T (2000) Effects of habitat fragmentation on plant-insect communities. In: Ekbom B, Irwin ME, Robert Y (eds) Interchanges of insects between agricultural and surrounding landscapes. Springer Netherlands, Dordrecht, pp 53–70
KTBL (2022) Leistungs-Kostenrechnung Pflanzenbau. https://daten.ktbl.de/dslkrpflanze/postHv.html#Ergebnis. Accessed 24 May 2022
Langenbuch R (1941) Zur Biologie des Erbsenwicklers Grapholitha nigricana Steph. Arbeiten zur Physiologischen und Angewandten Entomologie:216–247
Light DM, Knight AL, Henrick CA (2001) A pear-derived kairomone with pheromonal potency that attracts male and female codling moth, Cydia pomonella (L.). Naturwissenschaften 88:333–338
Meier U, Bleiholder H, Buhr L, Feller C, Hack H, Heß M, Lancashire PD, Schnock U, Stauß R, van den Boom T, Weber E, Zwerger P (2009) The BBCH system to coding the phenological growth stages of plants – history and publications –. J. Kulturpflanzen 61:41–52. https://doi.org/10.5073/JFK.2009.02.01
Meng S, Wang X, Hu X, Luo C, Zhong Y (2021) Deep learning-based crop mapping in the cloudy season using one-shot hyperspectral satellite imagery. Comput. Electron. Agric. 186:106188. https://doi.org/10.1016/j.compag.2021.106188
Nicolaisen W (1928) Der Erbsenwickler, Grapholita (Cydia, Laspeyresia) sp., sein Schaden und seine Bekämpfung unter Berücksichtigung der Anfälligkeit verschiedener Erbsensorten. Kühn Archiv:196–256
Nolte H-W, Adam H (1962) Über das Verhalten des Erbsenwicklers gegenüber Erbsensorten und Erbsen-Neuzuchtstämmen. Der Züchter 32:175–179
Pedigo LP, Hutchins SH, Higley LG (1986) Economic Injury Levels in Theory and Practice. Annu. Rev. Entomol. 31:341–368. https://doi.org/10.1146/annurev.en.31.010186.002013
Pimentel D, Goodman N (1978) Ecological basis for the management of insect populations. Oikos 30:422–437. https://doi.org/10.2307/3543337
Prasad YG, Prabhakar M (2012) Pest monitoring and forecasting. In: Abrol DP, Shankar U (eds) Integrated pest management: principles and practice. CABI, Wallingford, pp 41–57
Riemer N, Schieler M, Racca P, Saucke H (2021) Modelling of post-diapause development and spring emergence of Cydia nigricana (Lepidoptera: Tortricidae). Bull. Entomol. Res. 111:402–410. https://doi.org/10.1017/S0007485320000772
Sarwar SL (1973) Untersuchung zur Biologie und Bekämpfung des Erbsenwicklers Laspeyresia nigricana Steph. (Lepidoptera:Tortricidae). doctoral thesis, Karl-Marx-Universität Leipzig
Stenmark A (1974) Studies on the pea moth (Laspeyresia nigricana Steph.) in Central Sweden. Statens Växtskyddsamstalt Meddelanden 15:451-472
Stern V, Smith R, van den Bosch R, Hagen K (1959) The integration of chemical and biological control of the spotted alfalfa aphid: The integrated control concept. Hilgardia 29:81–101
Sun Z, Di Wang, Zhong G (2018) A Review of Crop Classification Using Satellite-Based Polarimetric SAR Imagery. In: 2018 7th International Conference on Agro-geoinformatics (Agro-geoinformatics). IEEE
Sütő J (2021) Embedded System-Based Sticky Paper Trap with Deep Learning-Based Insect-Counting Algorithm. Electronics 10:1754. https://doi.org/10.3390/electronics10151754
Teetes GL (1991) The environmental control of insects using planting time and plant spacing. In: Pimentel D (ed) CRC handbook of pest management in agriculture, 2nd edn. CRC Pr, Boca Raton u.a.
Thöming G, Knudsen GK (2014) Attraction of pea moth Cydia nigricana to pea flower volatiles. Phytochem. 100:66–75. https://doi.org/10.1016/j.phytochem.2014.01.005
Thöming G, Norli HR (2015) Olfactory cues from different plant species in host selection by female pea moths. J. Agric. Food. Chem. 63:2127–2136. https://doi.org/10.1021/jf505934q
Thöming G, Pölitz B, Kühne A, Saucke H (2011) Risk assessment of pea moth Cydia nigricana infestation in organic green peas based on spatio-temporal distribution and phenology of the host plant. Agric. For. Entomol. (Agricultural and Forest Entomology) 13:121–130. https://doi.org/10.1111/j.1461-9563.2010.00507.x
Thöming G, Norli HR, Saucke H, Knudsen GK (2014) Pea plant volatiles guide host location behaviour in the pea moth. Arthropod Plant Interact. 8:109–122. https://doi.org/10.1007/s11829-014-9292-5
Touvinen T (1982) Prognosis of injury by Cydia nigricana (Lepedoptera, Tortricidae) using "the oecos pea moth monitoring system": A preliminary report. Acta. Entomol. Fenn. 40:35–38
Visotsching W (1998) Praxiserhebungen und Feldversuche zur Verbreitung und vorbeugenden Regulierung des Erbsenkäfers (Bruchus pisorum) und des Erbsenwicklers (Cydia nigricana) in biologisch bewirtschafteten Körnererbsenschlägen in Österreich. Diplomathesis, Universität für Bodenkultur Wien
Wang R, Li R, Chen T, Zhang J, Xie C, Qiu K, Chen P, Du J, Chen H, Shao F, Hu H, Liu H (2022a) An automatic system for pest recognition and forecasting. Pest. Manag. Sci. 78:711–721. https://doi.org/10.1002/ps.6684
Wang X, Zhang J, Xun L, Wang J, Wu Z, Henchiri M, Zhang S, Zhang S, Bai Y, Yang S, Li S, Yu X (2022b) Evaluating the Effectiveness of Machine Learning and Deep Learning Models Combined Time-Series Satellite Data for Multiple Crop Types Classification over a Large-Scale Region. Remote Sens. 14:2341. https://doi.org/10.3390/rs14102341
Weisz R, Smilowitz Z, Fleischer S (1996) Evaluating risk of Colorado Potato Beetle (Coleoptera: Chrysomelidae) infestation as a function of migratory distance. J. Econ. Entomol. 89:435–441. https://doi.org/10.1093/jee/89.2.435
Wood S (2022) Mixed GAM computation vehicle with automatic smoothness estimation. https://cran.r-project.org/web/packages/mgcv/mgcv.pdf. Accessed 7 December 2022
Wright DW, Geering QA (1948) The biology and control of the pea moth, Laspeyresia nigricana, Steph. Bull. Entomol. Res. 39:57–84. https://doi.org/10.1017/S0007485300024299
Wright DW, Geering QA, Dunn JA (1951) Varietal differences in the susceptibility of peas to attack by the pea moth, Laspeyresia nigricana (Steph.). BER 41:663. https://doi.org/10.1017/S0007485300027929
Zeuner T (2007) Landwirtschaftliche Schaderregerprognose mit Hilfe von Geographischen Informationssystemen. Doctoral Thesis, Johannes Gutenberg - Universität Mainz

No competing interests reported.

Download PDF

Journal Publication

published 15 Mar, 2024

Read the published version in Entomologia Experimentalis et Applicata →

Version 1

posted

You are reading this latest preprint version

The efficacy of spatio-temporal predictors in forecasting the risk of Cydia nigricana infestation

Status:

Journal Publication

Version 1

Abstract

Figures

Key Message

Introduction

Material and methods

Data analysis

Results

Discussion

Conclusion

Declarations

Authors Contribution Statement

Acknowledgments

Funding

Competing interests

Author contributions

Data availability

References

Additional Declarations

Status:

Journal Publication

Version 1