Environmental Covariates for Sampling Optimization and Pest Prediction in Soybean Crops

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The article “Environmental Covariates for Sampling Optimization and Pest Prediction in Soybean Crops” describes the use of environmental covariates in the biphasic optimization of spatial sampling predictions for insect pests, using soybean crops in Brazil as an example. The authors found no spatial dependence through geostatistical methods and regression analysis. This approach can also be used to analyse pests for other crops in different climatic zones. While the work is of high quality, there are several areas that require improvement, as outlined below:

1. Lines 205-206: Did this method capture all the insects on the cloth? Could some still be present on the soybean plants?

2. How was the clay content of the soil analyzed?

3. What is the reason for the large variation in clay content across the fields (Figure 3F)?

 

4. How does the intensity of insecticide application affect the outcomes of the study? For example, efficacy may vary depending on environmental covariates.

Author Response

Author's Reply to the Review Report

Manuscript Title: Environmental Covariates for Sampling Optimization and Pest Prediction in Soybean Crops

Manuscript ID: agriengineering-3398939

Dear Reviewer,

We would like to thank you for your time and valuable feedback on our manuscript. Below are our responses to the review report:

Comments and Suggestions for Authors

The article “Environmental Covariates for Sampling Optimization and Pest Prediction in Soybean Crops” describes the use of environmental covariates in the biphasic optimization of spatial sampling predictions for insect pests, using soybean crops in Brazil as an example. The authors found no spatial dependence through geostatistical methods and regression analysis. This approach can also be used to analyse pests for other crops in different climatic zones. While the work is of high quality, there are several areas that require improvement, as outlined below:

1. Lines 205-206: Did this method capture all the insects on the cloth? Could some still be present on the soybean plants?

Authors’ response: The beat cloth method is widely used for sampling insects in crops such as soybeans and effectively captures a broad range of insects located on the upper and external parts of the plants, including leaves, stems, and flowers [1], [2]. In annual crops, the beat cloth sampling is adequate for capturing more specimens per point [2]. While it is possible that some specimens remain on the plants or that highly mobile insects may escape the cloth, the method has been shown to produce minimal errors. Furthermore, sub-sampling techniques help mitigate these errors in population counts. The method is highly efficient for pest mapping, particularly when integrated with environmental covariates in machine learning algorithms. These algorithms leverage pest-environment relationships to enhance the accuracy of predictions, ensuring robust spatial analysis and pest management strategies.

Although the beat cloth method was introduced many years ago, it remains the most widely used and efficient technique for sampling pests in annual crops.

2. How was the clay content of the soil analyzed?

Authors’ response: The soil samples were sent to a commercial laboratory for physical property analysis. The samples were dried at 40°C and sieved through a 2 mm mesh to obtain fine air-dried soil for texture determination. The soil samples were sent to a commercial laboratory for physical property analysis. The samples were dried at 40°C and sieved through a 2 mm mesh to obtain fine air-dried soil for texture determination. The particles were separated using the chemical dispersion method, in which sodium hexametaphosphate was added to the sample to break up the particle aggregates and isolate the individual fractions, according to the Brazilian Soil Classification System (SiBICS), described by Santos et al. (2018) [3]. The soil was then categorized into textural fractions: sand (> 0.053 mm, in g/kg), clay (< 0.002 mm, in g/kg) and silt (values between clay and sand). These revisions are included in lines 214-224 of the manuscript (Lines 172 to 179 in the new file without highlights).

3. What is the reason for the large variation in clay content across the fields (Figure 3F)?

Authors’ response:  The wide variation in clay content across the field, as shown in Figure 3F, can be attributed to various factors, including source material, soil formation processes, and historical land use. Among these, topography plays a key role in shaping the spatial distribution of clay. Our study area shows significant topographical variability, as seen in the slope map (Figure 3G). Topography influences water movement, sediment deposition and erosion patterns, leading to the accumulation of finer particles, such as clay, in lower areas (deposition zones) and the loss of these particles at higher altitudes (erosion zones).

4. How does the intensity of insecticide application affect the outcomes of the study? For example, efficacy may vary depending on environmental covariates.

Authors’ response:  The intensity of insecticide application can significantly influence the results of pest-environment relationships. In areas with a high intensity of insecticide application, the suppression of pest populations can be more pronounced, masking the influence of environmental covariates on pest distribution. In our study, we conducted sampling at different phenological stages throughout the crop cycle to reduce the chances of management affecting the quality of environmental covariates selection. This allows us to capture these relationships in different scenarios, such as the environment, pest population, and spatial distribution patterns of pest insect infestations.

Thank you once again for your guidance. We look forward to your feedback and remain available to make any further adjustments if needed.

Kind regards,
Authors.

References

[1]        M. Shepard, G. R. Carner, and S. G. Turnipseed’, “Colonization and Resurgence of Insect Pests of Soybean in Response to Insecticides and Field Isolation,” 1977. doi: https://doi.org/10.1093/ee/6.4.501.

[2]        L. Storck, C. C. C. Antúnez, J. V. C. Guedes, A. Cargnelutti Filho, and J. W. R. Alvarez, “Comparison of beat cloth and entomological net methods for determining faunistic indices of soybean in Rio Grande do Sul, Brazil,” Sci Agric, vol. 73, no. 6, pp. 559–564, 2016, doi: 10.1590/0103-9016-2015-0164.

[3]        H. G. dos. Santos, Sistema brasileiro de classificação de solos. Embrapa, 2018.

 

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript investigated the spatial predictions for soybean insect pests using environmental covariates in two-phase sample optimization The results showed that the multi-objective optimized sampling design produced the most accurate predictions, suggesting that a two-phase sample optimization with prior in situ selection of environmental covariates facilitates pest predictions in agricultural systems, contributing to more efficient and sustainable IPM. The study sounds interesting as highlighting the potential of the two-phase sampling optimization for the improvement of the effectiveness of pest mapping strategies in agricultural systems. Overall, some minor points be considered by the authors before it can be recommended for publication.

1.In figure 6, the lines and labels should be given clear indication in the figure legend for a better understanding.

2. The same with Figure 7.

3. In Phase 1, 2.3.1 and 3.1.1, the subheading, e.g., vegetation indices or riparian areas should be numbered as a) Vegetation indices…or c) Riparian areas

Author Response

Author's Reply to the Review Report

Manuscript Title: Environmental Covariates for Sampling Optimization and Pest Prediction in Soybean Crops

Manuscript ID: agriengineering-3398939

Dear Reviewer,

We would like to thank you for your time and valuable feedback on our manuscript. Below are our responses to the review report:

Comments and Suggestions for Authors

The manuscript investigated the spatial predictions for soybean insect pests using environmental covariates in two-phase sample optimization The results showed that the multi-objective optimized sampling design produced the most accurate predictions, suggesting that a two-phase sample optimization with prior in situ selection of environmental covariates facilitates pest predictions in agricultural systems, contributing to more efficient and sustainable IPM. The study sounds interesting as highlighting the potential of the two-phase sampling optimization for the improvement of the effectiveness of pest mapping strategies in agricultural systems. Overall, some minor points be considered by the authors before it can be recommended for publication.

1. In figure 6, the lines and labels should be given clear indication in the figure legend for a better understanding.

Authors’ response:  Correction made.

2. The same with Figure 7.

Authors’ response: Correction made.

3. In Phase 1, 2.3.1 and 3.1.1, the subheading, e.g., vegetation indices or riparian areas should be numbered as a) Vegetation indices…or c) Riparian areas

Authors’ response:  Added.

Thank you once again for your guidance. We look forward to your feedback and remain available to make any further adjustments if needed.

Kind regards,
Authors.

Reviewer 3 Report

Comments and Suggestions for Authors

The paper under review presents investigates whether regional forecasts for soybean insect pests are improved by including environmental factors into two-phase sample optimization. In a commercial soybean field in Brazil, insect pest samples were gathered at 50 georeferenced positions during the 2021–2022 crop season. Additionally, information on environmental variables, including vegetation indices, soil characteristics, topography, and distances from riparian regions, was also gathered. Principal component analysis (PCA) and correlation are used to choose three variables. Two optimum designs are produced using the iterative Optimization of Sample Configurations using Spatial Simulated Annealing (SPSANN) method with the chosen covariates during the 2022–2023 crop season. These designs were then compared to a Regular grid.

 Geostatistical techniques, regression analysis (pest abundance), and classification (pest presence or absence) using the Random Forest algorithm are used to assess the data from the three sample designs, which totaled 50 points. The most accurate predictions, however, came from a multi-objective optimized sampling design that was specifically designed to improve configurations for locating and estimating variograms and geographical trends that are crucial for spatial interpolation. Pest predictions in agricultural systems are improved by a two-phase sample optimization process that includes prior in situ selection of environmental variables. This leads to more sustainable and effective agricultural management. Overall this study is interesting and new. The paper is also written well. So, I recommend acceptance of the paper. The following are some suggestions:

1. I suggest revising Figure 5 so that the problem of text overlapping with arrows be solved. 

 

2. In Section 3.2.2, I suggest considering to use non-linear regression to see if the results may be improved. 

 

3. I suggest adding a few sentence on how vegetation indices approach may be improved.

Author Response

Author's Reply to the Review Report

Manuscript Title: Environmental Covariates for Sampling Optimization and Pest Prediction in Soybean Crops

Manuscript ID: agriengineering-3398939

Dear Reviewer,

We would like to thank you for your time and valuable feedback on our manuscript. Below are our responses to the review report:

Comments and Suggestions for Authors

The paper under review presents investigates whether regional forecasts for soybean insect pests are improved by including environmental factors into two-phase sample optimization. In a commercial soybean field in Brazil, insect pest samples were gathered at 50 georeferenced positions during the 2021–2022 crop season. Additionally, information on environmental variables, including vegetation indices, soil characteristics, topography, and distances from riparian regions, was also gathered. Principal component analysis (PCA) and correlation are used to choose three variables. Two optimum designs are produced using the iterative Optimization of Sample Configurations using Spatial Simulated Annealing (SPSANN) method with the chosen covariates during the 2022–2023 crop season. These designs were then compared to a Regular grid.

Geostatistical techniques, regression analysis (pest abundance), and classification (pest presence or absence) using the Random Forest algorithm are used to assess the data from the three sample designs, which totaled 50 points. The most accurate predictions, however, came from a multi-objective optimized sampling design that was specifically designed to improve configurations for locating and estimating variograms and geographical trends that are crucial for spatial interpolation. Pest predictions in agricultural systems are improved by a two-phase sample optimization process that includes prior in situ selection of environmental variables. This leads to more sustainable and effective agricultural management. Overall this study is interesting and new. The paper is also written well. So, I recommend acceptance of the paper. The following are some suggestions:

1. I suggest revising Figure 5 so that the problem of text overlapping with arrows be solved. 

Authors’ response:  The correction was made.

2. In Section 3.2.2, I suggest considering to use non-linear regression to see if the results may be improved. 

Authors’ response: Thank you for the suggestion to explore non-linear regression. In this study, we employed the Random Forest regression model, which is inherently non-linear and capable of capturing complex relationships between predictors and the response variable. Random Forest is a robust and flexible machine-learning algorithm that handles interactions and non-linearities effectively without requiring prior specification of the functional form. We believe this approach is well-suited for our dataset and the objectives of our study, as it maximizes predictive accuracy and interprets the pest-environment relationships.

3. I suggest adding a few sentence on how vegetation indices approach may be improved.

Authors’ response: Thank you for your suggestion. We tested several vegetation indices (VIs) using data collected during each sampling event and throughout the crop cycle. This approach allowed us to capture the relationships between pests and VIs under various scenarios, including changes in environmental conditions, pest population dynamics, and spatial infestation patterns, providing a representative VI selection. While vegetation indices provide valuable insights into vegetation biomass, some have limitations, such as susceptibility to saturation under high biomass conditions, as observed with NDVI in certain studies. However, no saturation issues were detected in our analysis. Additionally, the suitability of a specific VI can vary depending on the crop type and the study's objectives, highlighting the importance of testing multiple indices before selecting the most appropriate one. Therefore, we believe the methodology employed in this study is the most suitable for selecting vegetation index to enable two-phase optimized sampling and spatial prediction of pest infestations.  

Thank you once again for your guidance. We look forward to your feedback and remain available to make any further adjustments if needed.

Kind regards,
Authors.