Temporal dynamics of sapota pest damage and Phytophthora disease: insights from time series and machine learning models

Abstract

Introduction: Sapota (Manilkara zapota L.) is a major tropical fruit crop prone to damage by bud borer (Anarsia achrasella), seed borer (Trymalitis margarias), and fruit rot caused by Phytophthora species. Climatic variability strongly influences these biotic stresses, yet long-term temporal patterns remain poorly quantified.

Methods: A decade-long dataset (2014-2022) from 21 major sapota-growing districts of Maharashtra, India, was analyzed to study pest and disease dynamics. Statistical and machine learning approaches, including ARIMA, SARIMA, and VAR time-series models, along with Random Forest feature importance analysis, were applied to quantify climatic influences and forecast severity trends. Correlation analyses were used to assess weather-pest/disease associations.

Results: Trend analysis revealed fluctuating bud and seed borer damage, while Phytophthora disease severity remained relatively stable. Bud borer incidence was positively correlated with rainfall (r = 0.69), seed borer with maximum temperature (r = 0.47), and Phytophthora with minimum temperature (r = 0.64). The ARIMA model provided accurate forecasts for bud borer (MSE = 8.03) and Phytophthora (MSE = 0.20), while the VAR model performed best for seed borer (MSE = 17.96). Random Forest analysis identified minimum temperature as the most critical driver of bud borer and Phytophthora severity, whereas relative humidity was most influential for seed borer.

Discussion: The integration of statistical and machine learning models provides robust insights into sapota pest and disease epidemiology under climatic variability. These findings highlight the importance of temperature, humidity, and rainfall in shaping pest-pathogen interactions and provide predictive tools to design timely, targeted, and climate-resilient management strategies for sapota cultivation.

Keywords: ARIMA; SARIMA; VAR; climatic factors; pest damage; phytophthora disease; random forest; sapota.

Figure 1

The chart shows the trends in pest damage (bud borer and seed borer) and disease (Phytophthora) over the years from 2014 to 2022.

Figure 2

Monthly average percent disease index (PDI) of bud borer, seed borer, and phytophthora in sapota cultivation based on 8-year data (2014–2022). The figure illustrates seasonal peaks and periods of reduced severity, supporting strategic forecasting and pest/disease management.

Figure 3

The heatmap visualizes the correlation matrix between weather variables (maximum temperature, minimum temperature, relative humidity, and rainfall) and pest/disease data (Bud Borer Damage, Seed Borer Damage, and Phytophthora Disease Severity).

Figure 4

ARIMA predictions for bud borer, seed borer, and phytophthora disease in sapota, based on averaged monthly observed data from 2014 to 2022.

Figure 5

SARIMA forecasts for pest and disease damage in sapota cultivation: The figure presents the observed and forecasted values of (a) Bud Borer Damage, (b) Phytophthora Disease Severity, and (c) Seed Borer Damage in sapota over time using SARIMA models. The solid lines represent the observed data, while the dashed lines show the forecasted trends for the next 12 months.

Figure 6

Autocorrelation function (ACF) and partial autocorrelation function (PACF) plots of residuals from SARIMA models fitted to Bud Borer, Seed Borer, and Phytophthora disease severity time series in sapota cultivation (2014–2022). The absence of significant autocorrelation in the residuals suggests that the SARIMA models adequately captured the underlying structure of the time series.

Figure 7

Vector autoregressive (VAR) model forecasts for pest and disease damage in sapota cultivation (2014-2022). The figure illustrates the observed and forecasted values for Bud Borer Damage, Phytophthora Disease Severity, and Seed Borer Damage from 2014 to 2022 using averaged monthly data. The solid lines represent the historical observed data, while the dashed lines indicate the model’s forecasted trends.

Figure 8

Combined CUSUM (cumulative sum) plots of standardized residuals for Bud Borer, Seed Borer, and Phytophthora disease severity in sapota cultivation from 2014 to 2022. The residuals for all three variables remain within control limits (± 0.5), indicating no structural breaks or instability over the 8-year period.

Figure 9

Normalized relative importance of meteorological variables in forecasting Bud Borer, Seed Borer, and Phytophthora disease severity in sapota cultivation based on 8-year averaged data (2014–2022).