Population Dynamics of Agonoscena pistaciae and Comparison of the Artificial Neural Network and Neural-Genetic Models for Predicting the Pest Population

پژوهش های حفاظت گیاهان ایران

Article 5, Volume 33, Issue 2 - Serial Number 44, June 2019, Pages 171-185   

Document Type: Research Article

DOI: 10.22067/jpp.v33i2.79972

Authors

S. Gholami Moghaddam; L. Naderloo

Razi University, Kermanshah

Abstract

Introduction: The common pistachio psylla, Agonoscena pistaciae (Hemiptera: Aph­alaridae), is a key pest of pistachio trees. Nymphs and adults suck sap from leaves resulting in defoliating, falling flower buds, stopping tree growth and finally yield loss. The population dynamics of insects is influenced by several physical and biological factors such as temperature and natural enemies. Identifying the key factor in population dynamics is difficult due to the potential interactions between biological and environmental factors. Non-linear analysis methods such as artificial neural networks (ANNs) are suited to be applied in the ecosystem with non-linear and complex ecological data. These methods have been widely used as a robust information-processing instrument in many research fields, especially in predicting pest occurrence. For example, a neural model is used to predict bionomic variables related to the nutritional dynamics of blowflies.
In the present investigation, the seasonal abundance of A. pistaciae in a pistachio orchard was evaluated for two years. This study aimed to assess the performance of ANN in representing nonlinear dynamic data for common pistachio psyllid populations. To this end, back propagation ANN was implemented to evaluate the relationship between the pest occurrence and influential factors.
 Materials and Methods: The population density of ‎psyllids was monitored weekly by the yellow sticky trap for the adult and direct counting for ‎the nymph. After collecting the related data, the curves of the seasonal dynamic population of adults and nymphs were drawn. Then, the number of generations and duration activity of psyllid in each generation were determined. Multi-layer perceptron neural network ‎‎(MLP), genetic algorithm (GA) and multi-linear regression (MLR) were used to determine the relative significance of biotic (natural enemies) and ‎abiotic (weather variables) factors for predicting A. pistaciae density. An ANN model was designed by using the inputs (average temperature, average rainfall, average relative humidity, wind speed and direction,  and population of natural enemies), hidden layer (the number of neurons in the hidden layers determined by trial and error), and one neuron in the output layer (the occurrence amount for predicting the population). The Levenberg–Marquardt algorithm was used as the learning algorithm. The root mean square error (RMSE) and coefficient of determination (R2) were statistics, calculated for both the training and testing set for each iteration.
Results and Discussion: The population fluctuations of A. pistaciae on Akbari pistachio cultivar during 2015 and 2016 indicated that the psyllid populations in the field had five apparent peaks from late March to October. Agonoscena pistaciae in Rafsanjan county had six complete and one incomplete generation in 2007 and 2008 (9). The general population trends were similar over time within two years, but population densities of adults and nymphs were higher in 2016. Statistical comparison of weather variables between two years showed no significant difference. 
Several topologies were examined and the best result was obtained with 15 and 9 neurons in the first and second hidden layer for both adult and nymph in MLP method, respectively. In genetic algorithm, a hidden layer with 14 neurons for adult and 16 neurons for nymph was employed. The R2 values of MLR, MLP and GA methods (at test phase) were 0.32, 0.61, 0.73, respectively and the RMSE values were 31.79, 0.223 and 0.083, respectively for adult. In the prediction of the population density of the nymph by MLR, MLP and GA, the R2 values were obtained to be 0.22, 0.84, 0.88, respectively, and the RMSE values were 48.03, 0.051 and 0.051, respectively.
Conclusion: The R2 and RMSE values showed reliable performance of ANN and GA. The ANNs also modeled the numbers of the psyllid with high accuracy. In addition, the higher R2 and lower RMSE were obtained for MLP and GA methods relative to MLR. It has been reported in the related literature that the ANN consistently outperformed the statistical models. The ANN as a nonlinear predictor exhibited a high accuracy in predicting the richness of aquatic insect species in running waters by a set of four environmental variables (21). Based on the principal components analysis and back propagation artificial neural methods to analyze historical data on the population occurrence of Scirpophaga incertulas, the new model could improve the prediction accuracy, compared with other methods (27). It is worth noting that in regression models, the weak correlation between dependent and independent variables does always not imply that these variables are not associated, as they may have a nonlinear correlation.

Keywords

Common pistachio psylla‎; Genetic algorithm‎; Multi layer perceptron neural network‎; Population dynamic; Weather variables‎

References

1- Assar M. 2001. Biology of pistachio psylla Agonoscena pistaciae (Hom. : Psyllidae) in Dameghan and effect of Releasing Chrysoperla carnea (Stephens) (Neu.: Chrysopidae) egg integrated with using sticky yellow traps in reducing psylla populations in different generations. M. Sc. dissertation, Islamic Azad University, Science & Research Branch, Tehran, Iran. (In Persian)

2- Bianconi A., Von Zuben C.J., Serapião A.B.S., and Govone, J. 2009. Artificial neural networks: A novel approach to analysing the nutritional ecology of a blowfly species, Chrysomya megacephala. Journal of Insect Science 10: 1-18.

3- Chon T.S., Park Y.S., Kim J.M., Lee B.Y., Chung Y.J., and Kim Y. 2000. Use of an artificial neural network to predict population dynamics of the forest–pest pine needle gall midge (Diptera: Cecidomyiida), Environmental Entomolology 29: 1208-1215.

4- Dezianian A., and Sahragard A. 2004. Biology and natural enemies of pistachio psyllid, Agonoscena pistaciae in Dameghan region, Journal of Agricultural Science 1(5): 83-92. (In Persian)

5- Dustiy Z., Moeini Naghadae N., Zamani A.A., and Naderloo L. 2017. Modeling the population changes of sunn pest with environmental variables using artificial neural network and comparison with the linear regression model in Chadegan County. Iranian Journal of Plant Protection Science 47(2): 307-315. (In Persian with English abstract)

6- Food and Agriculture Organization. 2016. FAO statistical databases. From http://www.faostat.fao.org.

7- Freeman J., and Sakura D. 2005. Neural Networks: Algorithms, Applications, and Programming Techniques, Addison-Wesley, Berlin.

8- Han J., Pei J., and Kamber M. 2011. Data mining: concepts and techniques, Elsevier, Waltham, USA, 673p.

9- Hassani M.R., Nouri-Ghanbalani Gh., Eizadi H., and Shojaei M. 2010. Population fluctuations of pistachio psylla, Agonoscena pistaciae (Hemiptera: Psyllidae), in Rafsanjan region. Iranial Journal of Plant Protection Science 40(2): 93-98. (In Persian)

10- Heung B., Chak H., Zhang J., Knudby A., Bulmer C.E., and Schmidt, M.G. 2016. An overview and comparison of machine-learning techniques for classification purposes in digital soil mapping, Geoderma 265: 62-77.

11- Jalaeian M., and Karimi Malati A. 2013. Comparison of common pistachio psyllid (Agonoscena pistaciae) population on main native and non-native pistachio varieties in Khorasan Razavi Province, Plant Pest Research 2(4): 45-54. (In Persian with English abstract)

12- Karamouz M., and Araghinejad S. 2014. Advanced Hydrology. 3rd edition. Amirkabir University of Technology Press, Tehran. (In Persian)

13- Kouchakzadeh M., and Bahmani A. 2006. Assessment of artificial neural networks revenue in reducing required parameters for estimation of reference evapotranspiration. Journal of Agriculture Science 11(4): 87-97.

14- Lankin-Vega G., Worner S.P., and Teulon D.A.J. 2008. An ensemble model for predicting Rhopalosiphum padi abundance. Entomologia Experimentalis et Applicata 129: 308–315.

15- Mehrnejad M.R. 2008. Seasonal biology and abundance of Psyllaephagus pistaciae (Hymenoptera: Encyrtidae), a biocontrol agent of the common pistachio psylla Agonoscena pistaciae (Hemiptera: Psylloidea). Biocontrol Science and Technology 18: 409-417.

16- Mehrnejad M.R. 2010. Potential biological control agents of the common pistachio psylla, Agonoscena pistaciae, a review. Entomofauna 18: 249-272.

17- Mehrnejad M.R., and Emami S.Y. 2005. Parasitoids associated with the common pistachio psylla, Agonoscena pistaciae in Iran, Biological Control 32: 385-90.

18- MenhajM. 2005. Computational intelligence. Amirkabir University of Technology Press, Tehran. (In Persian)

19- Moeini- NaghadehN. 2007. Ecological approach on insect pest management. Razi University Press, Kermanshah. (In Persian)

20- Montgomer D.C., Peck E.A., and Vining, G.G. 2015. Introduction to Linear Regression Analysis, John Wiley & Sons.

21- Park Y.S., Cereghino R., Compin A., and Lek S. 2003. Applications of artificial neural networks for patterning and predicting aquatic insect species richness in running waters, Ecological Modelling 160: 265-280.

22- Samih M.A., Alizadeh A., and Saberi Riseh R. 2005. Pistachio pests and diseases in Iran and their IPM, Organization of Jihad-e-University, Tehran.

23- Schaap M.G., Leij F.J., and van Genuchten M.T. 1998. Neural network analysis for hierarchical prediction of soil hydraulic properties. Soil Science of Society of America Journal 62: 847-855.

24- Souliotes C., Markoyiannaki-Printziou D., and Lefkaditis F. 2002. The problems and prospects of integrated control of Agonoscena pistaciae Burck. and Laut. (Hom., Sternorrhyncha) in Greece. Journal of Applied Entomology 126: 384-388.

25- Tonnang H.E.Z., Nedorezov L.V., Owino J.O., Ochanda H., and Lohr B. 2010. Host–parasitoid population density prediction using artificial neural networks: diamondback moth and its natural enemies. Agricultural and Forest Entomology 12: 233-242.

26- Worner S.P., and Gevrey M. 2006. Modelling global insect pest species assemblages to determine risk of invasion. Journal of Applied Ecology 43: 858-867.

27- Yang L.N., Peng L., Zhang L.M., Zhang L.L., and Yang S.S. 2009. A prediction model for population occurrence of paddy stem borer (Scirpophaga incertulas), based on back propagation artificial neural network and principal components analysis. Computers and Electronics in Agriculture 68: 200–206.

28- Yanik E., and Unlu L. 2015. Initial study of rearing and release of Anthocoris minki Dohrn (Hemiptera: Anthocoridae) for biological control of Agonoscena pistaciae Burckhardt and Lauterer (Hemiptera: Psyllidae) in pistachio orchards. Agriculture & Forestry 61: 347-358.

29- Zhang W., Zhoung X.Q., and Liu G.H. 2008. Recognizing spatial distribution patterns of grassland insects: neural network approaches. Stochastic Environmental Research and Risk Assessment 22:207–216.