https://orcid.org/0000-0001-7946-0373
Figures
Abstract
Rice blast disease, caused by Magnaporthe oryzae, is a major and foremost constraint to rice production in Bangladesh which threaten the food security and farmer livelihoods. This study delineates season-specific hotspots of the devastating rice Neck Blast Disease (NBD) across rainfed (Aman) and irrigated (Boro) ecosystems in the Bangladesh’s most intensive rice cultivation area Cumilla District using spatio-temporal risk assessment. Multi-year field data (2017–2025) were analyzed through hierarchical clustering, Moran’s I, Local Indicators of Spatial Association (LISA), and geostatistical interpolation techniques, including ordinary and indicator kriging. Results revealed spatial heterogeneity and inter-annual variability in disease severity. Nangalkot emerged as a persistent Aman-season hotspot, with mean NBD severity reaching 45–60%, while peak Boro-season severity reached 60–75% in several Upazilas, including Cumilla Sadar Dakshin, Debidwar, Barura, Burichang, and Nangalkot. Climatic trends correlated with neck blast disease intensification, particularly during reproductive stages. Indicator kriging effectively identified high-probability risk zones (>20% severity), enabling precise mapping of localized outbreaks. Findings of this study underscore the need for location-specific, adaptive management strategies integrating resistant cultivars and predictive modeling. This research offers actionable insights for policymakers and extension services to mitigate blast epidemics and safeguard Bangladesh’s rice production under evolving climatic conditions.
Citation: Rashid MM, Rahman NMF, Haque N, Khan MAI, Mian MS, Hossain M, et al. (2026) Delineating season-specific hotspots: A spatio-temporal risk assessment of neck blast disease in rainfed and irrigated ecosystems of Cumilla District, Bangladesh. PLoS One 21(9): e0358537. https://doi.org/10.1371/journal.pone.0358537
Editor: Ahmet Uludag, Canakkale Onsekiz Mart University, TÜRKIYE
Received: February 2, 2026; Accepted: August 29, 2026; Published: September 21, 2026
Copyright: © 2026 Rashid et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The data used in this study are owned by the Bangladesh Rice Research Institute (BRRI) and cannot be made publicly available due to institutional copyright and data-sharing restrictions. The authors had permission to use the data for this study. Researchers who meet the criteria for access may request the data from BRRI through the institutional contact point: Bangladesh Rice Research Institute, Gazipur, Bangladesh; Website: www.brri.gov.bd; Email: dg@brri.gov.bd. Access will be subject to BRRI’s data-sharing and institutional approval procedures. Dataset citation: Bangladesh Rice Research Institute (BRRI). Neck Blast Disease surveillance data for Aman and Boro rice seasons in Cumilla District, Bangladesh, 2017–2025. Bangladesh Rice Research Institute, Gazipur, Bangladesh. Restricted institutional dataset.
Funding: The author(s) received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Agriculture, especially rice (Oryza sativa L.) production, is central to Bangladesh’s economy and food security [1]. Despite challenges from population growth, limited arable land, and frequent natural disasters, Bangladesh has achieved self-sufficiency in rice through advances in both crop genetics and management [2, 3]. As the staple food for over 170 million people, rice production accounts for approximately 46% of the crop sector’s gross domestic product (GDP) and provides the bulk of caloric (70%) and protein intake (56%) for the population [4, 5]. In Bangladesh, rice production mainly concentrated across three main seasons: Aus, Aman (rainfed) and Boro (irrigated) where Aman and Boro crops contribute the vast part of its total production. Given the projected population increase and the concurrent challenges posed by climate variability—driven by both internal variability and external climate forcings in the region [6–15]—maintaining and enhancing the current yield of approximately 39 million tons remains paramount [4,16–18].
However, despite the significant advances in adopting high-yielding varieties and irrigation technologies, disease of rice particularly blast disease poses the most devastating threat on production of rice [19]. The disease manifests in various stages, but infection at the panicle neck which known as Neck Blast (NBD) is particularly consequential as it directly interferes with grain filling [20, 21]. This leads to chalky kernels or complete sterility and causing immediate, catastrophic yield losses. Studies have estimated that under epidemic conditions, losses attributable to blast can range from 60% up to 100% in severely affected fields, posing a direct threat to farmer livelihoods and national stability [22].
Crucially, the confluence of high humidity, increase in temperatures, and intense monoculture during the rice plant’s vulnerable reproductive stage creates a highly congenial environment for severe Neck Blast development during the critical panicle initiation and flowering stages in Bangladesh [23–26]. Previous studies have reported that Neck Blast development is favored by moderate temperatures of approximately 25–28°C, high relative humidity above 93%, and prolonged leaf wetness of 7–14 hours, which enhance conidial germination, appressorium formation, and infection efficiency of Magnaporthe oryzae [27]. Multi-year weather analysis of the study region (Cumilla District) confirms an observable rising trend in maximum temperature across both the T. Aman and Boro seasons which indicates a climatic shift that provides a more favorable thermal environment for the pathogen [28]. While Neck Blast is known to affect both major growing seasons, disease incidence and severity are often high in the irrigated Boro season (November–May) and also demonstrate severe localized impact in the rainfed T. Aman season (July–December) [29,30].
The main objective of this study is addressing the critical knowledge gap concerning the spatial and temporal dynamics of Neck Blast disease under the distinct environmental pressures of the Aman and Boro seasons across the Cumilla District. This research seeks to: (1) quantify the annual incidence and severity of Neck Blast in major rice-growing regions and establish patterns of chronic risk; (2) identify and delineate significant spatial clustering of disease severity using Moran’s I hotspot analysis; and (3) utilize Indicator Kriging (IK) to delineate specific, high-probability risk zones (e.g., severity exceeding 20%) during the Boro season to aid in proactive disease management. The novelty of this study lies in its focus on season-specific NBD hotspots in Cumilla District, comparing rainfed Aman and irrigated Boro ecosystems using multi-year field data and geospatial techniques to identify persistent and shifting disease-risk zones. Ultimately, the findings will provide actionable, data-driven insights necessary for breeders, extension workers, and policymakers to develop targeted, sustainable management strategies, specifically through the deployment of stable, long-term resistant cultivars, to safeguard Bangladesh’s vital rice production base against this enduring and explosive phytopathological threat.
Methodology
Data: The study was conducted by gathering data from 7 Upazilas (Debidwar, Burichang, Chandina, Cumilla Adarsha Sadar, Cumilla Sadar Dakshin, Barura, and Nangalkot) in Cumilla District for the Aman season during the year 2017–2024 and from 10 Upazilas (Muradnagar, Debidwar, Burichang, Chandina, Cumilla Adarsha Sadar, Cumilla Sadar Dakshin, Barura, Laksham, Chauddagram, and Nangalkot) in Cumilla District for the Boro season during the 2018−19–2024−25 season. Follow the link to the S1 File to see the study regions broken down by year. Disease data was recorded from available sampling sites in each selected Upazila during each study year and the number of sampling sites fluctuated among Upazila and years depending on the accessibility of the field, the occurrence of diseases and survey coverage. Not all the same fields were necessarily monitored every year, but it was in the form of repeated Upazila-level field surveys. Because field selection was partly constrained by accessibility and disease occurrence, some sampling bias may have been introduced, and the surveyed fields may not fully represent all rice fields within each Upazila.
The 0–9 scale based on the Standard Evaluation System (SES) for rice was used to compute disease grading [31]. Using formula (1), the severity of neck blast was converted to a Percent Disease Index (PDI), as shown below.
Estimation of yield loss
Yield loss associated with neck blast disease was estimated from neck blast incidence using the empirical relationship described by Hossain et al. [32], following Padmanabhan [33]. Predicted grain yield was calculated as:
where Y is the predicted grain yield (lb/ha) and X is neck blast incidence (%). Predicted yield was converted to (t/ha) using a conversion factor of 0.000453592. The disease-free yield predicted by the model (approximately 0.89 t/ha was used as the reference yield, and percentage yield loss was estimated as:
The resulting yield-loss estimates were summarized by Upazila and year for comparison of the spatial and temporal effects of neck blast disease.
Cluster Analysis: To determine distances between Upazilas, an aggregative hierarchical cluster analysis was performed using the average linkage approach depending on the severity of NBD. Average linkage was selected because it provides a balanced clustering approach by considering the average pairwise distance between observations in two clusters. This method is less sensitive to chaining effects than single linkage and less restrictive than complete linkage, making it suitable for grouping Upazilas with similar Neck Blast Disease severity patterns across years. Silhouette analysis was performed to validate the hierarchical clustering results and to identify the optimal number of clusters based on average silhouette width. Data optimization and cluster analysis were carried out using R’s ‘hclust’ function (version R-4.0.3). In an average linkage hierarchical clustering, the distance (L) between two clusters (r,p) is the distance between two points, as stated by formula (2):
where X and Y represent the observations from clusters r and p, respectively.
Geostatistical analysis: To assess the spatial pattern of NBD, two major methods were used: surface interpolation and point pattern analysis. point-pattern-optimized hotspot analysis were utilized to establish cluster patterns for the occurrence of NBD. For creating spatial maps and determining possible risk zones, interpolation methods such as inverse distance weighting (IDW), ordinary kriging (OK) and indicator kriging (IK) were used. All spatial maps were prepared using the WGS 84/ UTM Zone 46N projected coordinate system (EPSG:32646), and scale bars and north arrows were added where applicable.
Point pattern analysis: Spatial autocorrelation was examined with Moran’s I and LISA statistics, tuned from the closest sampling points. Point-pattern hotspot analysis was conducted in R using the sf, spdep, and gstat packages; distance-based spatial weights were generated from Upazila coordinates, and spatial clustering was assessed using Global Moran’s I and LISA statistics at the 5% significance level. LISA helped in detecting spatial agglomeration, and significance was detected from the respective p-values in Table 1. Moran’s I index of every spatial unit i was calculated using Eq. (3).
where
xi = value at location i
= mean of variable
(Variance)
wij = weight matrix
The spatial weight matrix (wij)for Moran’s I and LISA analysis was constructed using a distance-based approach, where neighboring relationships among Upazilas were defined from their geographic coordinate. Weights were assigned by proximity, so that Upazilas that are close to each other are spatial neighbors.
Classifying each location using Moran ‘I, P-value and mean:
Spatial interpolation
Using the spatial interpolation approach, the values in the unsampled locations were predicted; for example, the NBD severities at sites (X1, X2…. Xn) are (Z1, Z2…. Zn). Spatial interpolation can be used to estimate the Z values at new point X. Using IDW and OK methods, the diseased surface area was calculated. An unsampled site’s IDW can be written using the formula (4):
Where, m represents the number of neighboring points and p indicates the parameter.
From the omitted point, using leaving one-out-cross validation the interpolated values are compared with the actual values.
The spatial correlation of the random function Z(X0) can be estimated using the interpolation method known as Kriging. Using the formula, the expected values of variable Z at unsampled point X0 are determined by formula (5).
For OK technique, the following formula (6) has been used:
where Z represents the variable of interest at spatial coordinates Xi and Xo. n is the number of neighbors connected with the sampling point; λi is the weight associated with sampling point Xi and the ith observation point. Semivariograms are used to calculate the closest neighbor indexed based on average spatial variability and NBD severity. It is defined as follows in (7):
where γ(h) is the semivariance for the interval distance class h, N(h) is the number of data pairs of a given lag interval distance and direction, Z (xi) is the measured sample value at point i, and Z (xi + h) is the measured sample value at position I + h.
Semivariogram values are fitted with spherical, exponential, Gaussian models and Matérn (Mat) model:
such as:
Spherical model:
Exponential model:
Gaussian model:
Matérn (Mat) model:
Where, C0 = nugget
C = partial sill
a = range parameter (scale
ν>0 = smoothness parameter
Γ(ν) = Gamma function
Kν = modified Bessel function of the second kind
The Spherical, Exponential, Gaussian and Matérn are the models that have been tested, as they are the most commonly used models that represent various types of spatial dependence in geostatistical analysis. The accuracy of estimated data across models and methodologies was examined using accuracy measures such as average standard error (ASE), mean square error (MSE), and root mean square error (RMSE). The best-fitted semivariogram model was selected using leave-one-out cross-validation by comparing MSE, RMSE, and ASE, while also considering the plausibility of estimated spatial parameters, including nugget, sill, and range. Indicator kriging (IK) was utilized to identify disease-vulnerable locations when the severity of NBD exceeded 20% per field. However, for the 2021–22 season, when overall disease severity was very low, a 5% threshold was applied to capture spatial variation under low-prevalence conditions and to distinguish relatively vulnerable locations. NBD severity threshold was used as a practical operational cutoff for indicator kriging to identify areas with meaningful disease pressure and potential management concern; however, this threshold was applied as a risk-mapping criterion rather than as a universal biological threshold. Based on this, probability risk maps were constructed using the best-fitted semivariogram model. A similar method was used to create a color-coded map for ordinary kriging, with the contour symbol representing the higher risk areas of NBD in Cumilla’s various rice ecosystems.
Ethics statement
This study did not involve human participants, human data, human biological materials, or animal subjects. The research was conducted exclusively on Neck Blast Disease of rice using field-level observations collected from agricultural plots. Therefore, ethical approval from an institutional ethics committee was not applicable, and informed written or verbal consent was not required.
Result
Weather conditions
During the Aman season in Fig 1, maximum temperature exhibited a clear rising trend between 2017 and 2024. Before 2020, it fluctuated between 32 °C and 33 °C, while after 2020 it ranged between 32 °C and 34 °C. A similar upward pattern was observed in the Boro season, where values varied between 29 °C and 33 °C. Precipitation, however, showed a declining trend in the Aman season. It peaked at around 1100 mm in 2018 but, after 2022, fluctuated within the range of 500–800 mm. In contrast, the Boro season displayed an increasing rainfall trend after 2018, reaching a peak of approximately 250 mm in 2020. Minimum temperature showed a slight upward trend in both seasons, ranging between 21 °C and 27 °C in Aman and 18 °C to 23 °C in Boro. Compared to Aman, relative humidity during Boro was lower (72–80%) and more variable, peaking between 2019 and 2020, followed by a steady decline thereafter. Mann–Kendall trend analysis showed that only maximum temperature during the Aman season had a statistically significant increasing trend (tau = 0.282, p = 0.041), while detailed results for all weather variables in both seasons are provided in S2 and S3 Files.
NBD severity in different Upazila in Cumilla
Fig 2 represent the descriptive visualizations of Neck Blast Disease (NBD) severity of Aman season that varies across the study area and year. Overall, highest mean severity is documented in Nangalkot, ranging from 45% to 60%, whereas Chandina shows the lowest mean severity (0–15%) with least severity across the year 2017–2024. Following that yield loss also highest at Nangalkot and Chandina shows the least based in this study.
In Debidwar, most of the time (2017, 2018, 2020, 2022) 45% to 60% mean severity is observed where in 2021 and 2023 it observed the least severity ranging from 0% to 15%. Debidwar experienced the highest yield loss (50–100%) in the year 2022 while other times it varied between 20% to 40%. In Burichang, the mean severity fluctuated mostly between 15% and 45%, with higher intensity during 2019, 2021 and 2024. Yield loss in Burichang generally ranged from 20% to 40% except for 2017 and 2020, when yield loss dipped to the lowest range (0–15%). Cumilla Sadar Dakshin exhibited moderate to high variation. In this region, mean severity was mainly between 15% and 60%. Following by mean severity most yield loss is also experienced this year. Barura displayed variable trends across the years, with mean severity mostly between 30% and 60%, and yield loss lying within the 30% to 50% range. Highest yield loss is observed in the year of 2017, 2021 and 2022.
Overall, the spatial pattern indicates that Nangalkot and Debidwar are the most vulnerable Upazilas for NBD in Aman season, while Chandina consistently remains the least affected both in terms of severity and yield loss.
Fig 3 represents descriptive visualization of the mean severity and yield loss of Boro season in Cumilla Upazila across 2018–2025. Overall, it is evident that Muradnagar is more vulnerable and Chauddagram is considered as least for NBD in Boro season in terms of both mean severity and yield loss. It is notable that the 2020–21 and 2021–22 seasons showed the least severity during the Boro season compared to other years.
The highest severity, ranging from 60% to 75%, was observed in Cumilla Sadar Dakshin (2019–20), Debidwar (2022–23), Barura (2023–24), and both Burichang and Nangalkot (2024–25). With the exception of the 2020–2021 seasons, the least severity was recorded in Burichang (2019–20), Chandina (2022–23), and Chauddagram (2023–24). In terms of yield loss, the highest losses were reported in Burichang (2018–19), Cumilla Sadar Dakshin (2019–20), Debidwar (2022–23), Barura (2023–24), and again in both Burichang and Nangalkot (2024–25) that is highly associated with mean severity of disease.
Hierarchical clustering
Hierarchical clustering using average linkage method is performed for two distinct seasons: Aman and Boro in Fig 4. The optimum number of clusters varied from season to season through cluster validation by silhouette analysis. For Aman season the two clusters (k = 2, average silhouette width = 0.642) showed the highest average silhouette width, which is a sign of a relatively good separation of the clusters. There was one cluster which consisted of Nangalkot, Debidwar, Cumilla Adarsha Sadar, Burichang, Barura and Cumilla Sadar Dakshin and another cluster for Chandina that has its own pattern of disease severity. For the Boro season, the maximum average silhouette width was found for three clusters k = 3, average silhouette width = 0.668, which also showed good separation between the clusters. The first cluster consisted of Chauddagram and Laksham; the second cluster comprised Chandina, Nangalkot, Cumilla Sadar Dakshin, Burichang, and Cumilla Adarsha Sadar; and the third cluster was Barura, Debidwar and Muradnagar during the Boro season. The results also suggest that the season-specific cluster structure is useful for identifying the Upazilas which have similar Neck Blast Disease severity patterns.
Spatial point pattern analysis of NBD
Local Moran’s I and p-values were used to analyze the clusters and outliers. NBD cluster patterns, which indicate scattered and aggregated clusters of severity, were found at the district level during the Aman season in 2017 and 2024 and the Boro season in 2018–2025 (Supplementary S1–S4 Fig). The majority of the Upazilas were grouped together (at I > 0) based on positive I values, indicating positive spatial autocorrelation between them. Upazilas form another cluster when I value are negative. To find out about high risk or low risk region Lisa clustering method is used.
The LISA cluster maps in Fig 5 for the Aman season (2017–2024) show clear spatial heterogeneity in the severity of neck blast disease across Upazilas. Year-wise LISA analysis identified significant High–High clusters in Debidwar in 2018 and Cumilla Adarsha Sadar in 2021, indicating localized hotspots of high Neck Blast severity. Significant Low–Low clusters were observed in Chandina and Cumilla Sadar Dakshin in 2023, indicating localized areas of lower disease severity. Some years also show high–low or low–high outliers, indicating spatial contrast between neighboring Upazilas. These results lead to shifting hotspots and necessitate continued monitoring and location-specific interventions to manage blast risk effectively.
Fig 6 LISA cluster maps for Boro (2018–2025) illustrate dynamic spatial patterns of neck blast disease risk across Upazilas. High–High clusters are prevalent in Muradnagar (2018-2019), Nangalkot (2019-20), both Debidwar and Burichang (2020-21) and, Debidwar and Chandina (2023-24) marking them as stable hotspots that indicates the high disease severity surrounding by high severity, while Low–Low clusters feature in areas of lower disease severity in Barura, Cumilla Sadar Dakshin and Nangalkot (2018-19), Barura and Cumilla Sadar Dakshin (2020-2021), Barura (2022-23), Cumilla Sadar Dakshin, Chauddagram, Nangalkot (2023-24) and Cumilla Sadar Dakshin (2024-25). Transitional clusters (High–Low or Low–High) indicate spatial heterogeneity, showing shifting disease pressure between neighboring areas. These patterns call for location-specific monitoring and adaptive interventions to mitigate risks during the Boro season.
Surface interpolation to the explicit spatial distribution
Inverse distance weighting (IDW) interpolation method is used to identify missing cell values using weighted combinations of a set of sample points. Using this method, contour map of NBD for two distinct seasons, Aman and Boro, for several Upazila of Cumilla has been shown in Fig 7 and Fig 8.
Green to Red colors indicate lower to higher disease severity points in different Upazila of Cumilla. The maps were created using R software (version R-4.0.3).
Green to Red colors indicate lower to higher disease severity points in different Upazila of Cumilla. The maps were created using R software (version R-4.0.3).
Fig 7 shows that, Nangalkot Upazila is the main points of the Cumilla Upazila, which have greater disease proportions (ranging 70% to 90%) and became a risk prone area for NBD. According to this study, the middle portion of the study region is thought to be less disease-prone, with disease severity ranging from 20% to 30%. However, more concentrations should be made in 2017 (for Debidwar and Barura), 2021 (Barura), and 2024 (Cumilla Sadar Dakshin), where roughly 50% to 60% disease risk is observed.
Fig 8 shows that disease severity risk was lowest during the Boro seasons of 2020–21 and 2021–22. In contrast, the highest risk levels were observed in Debidwar (2018–19 and 2022–23), Burichang (2023–24 and 2024–25), Cumilla Sadar Dakshin (2019–20), and Barura (2023–24).
Ordinary and indicator kriging
For ordinary and indicator kriging Spherical, exponential, and Gaussian semivariogram experimental models were used to determine the spatial patterns of NBD severity measurements. Based on cross-validation of the semivariogram results for Aman season (Table 1), which showed lower mean square error (MSE), root mean square standard error (RMSE), and average standard error (ASE) values.
From the above Table 2 it is seen that Overall, the Spherical model performed best in 2017, 2018, 2020, 2021, and 2024, showing the lowest RMSE values. In contrast, the Exponential model was selected for 2019, 2022, and 2023. Although the Gaussian model produced the lowest cross-validation error in 2023, its extremely large range and partial sill indicated an unstable and implausible fit; therefore, the Exponential model was preferred based on both predictive performance and parameter plausibility.
From Table 3, spherical model can be selected as optimal model for the season of 2018−19, 2019−20, 2022−23 and exponential outperform than others in 2021−22, 2023−24 and 2024−25 where gaussian model is perform best in 2020−21 based on minimum MSE values.
Using ordinary kriging interpolation in Fig 9, it is observed that in Aman season Nangalkot is more vulnerable, almost greater than 70% risk, to NBD across all study year. The district’s northeast and southeast regions (Debidwar, Barura, and Nangalkot) were more risky (>80%) in 2017 than others. The severity of the disease decreases between 2018 and 2020, but in 2021, Barura and Cumilla Sadar Dakshin demonstrate a risk of over 60% for NBD, which is prevalent in the majority of the study region in 2022. While Nagalkot continues to be a risk-prone area, the northern section of the district has relatively low disease severity in 2023. In the last year, the Upazila exhibits a fairly consistent moderate severity; nevertheless, Nangalkot exhibits a risk of over 70% for NBD, indicating that this is a high-risk location for NBD during this season.
Green to red-color coded surfaces depicts lower to higher disease severe points. The maps were created using R software (version R-4.0.3).
Fig 10 shows that the risk of NBD was significant prior to the 2021–22 Boro season. In 2018–19, the northern region had a disease severity of between 40% and 60% higher than the southern region. Cumilla Adarsha Sadar was the most dangerous of these, with a mean severity of about 60%. Cumilla Sadar Dakshin, which is in the center of the district, was identified as the NBD high-risk region the following year. Season 2021–2022 had the lowest danger by comparison. Risk has risen since this season. The majority of Cumilla District’s Upazilas can be regarded as high-risk areas in recent years. During this season, the risk of NBD is greater than 70% in Nangalkot and Burichang.
Green to red-color coded surfaces depicts lower to higher disease severe points. The maps were created using R software (version R-4.0.3).
Fig 11 uses the indicator kriging approach to examine the spatial probability of Neck Blast Disease severity above 20% across the study Upazilas during the Aman seasons from 2017 to 2024. It is clear from this graph that in 2017, northern part of Debidwar and Nangalkot, and southern part of Barura were more susceptible to NBD. Between 50% and 80% was the range. In 2018, severe NBD (>90%) affected the northern parts of Debidwar and Cumilla Sadar and the southern half of Nangalkot, while mild NBD (10%–20%) affected the rest of the district. All Upazilas had a 50% to 70% risk in 2019, with the exception of Barura, where a 20% risk was noted throughout this period. While other study regions had a 10%–20% chance of disease in 2020, central Debidwar, Cumilla Adarsha Sadar, and Nangalkot had 60%–80% disease. While other Upazilas had risk levels below 30% in 2021, Barura, Cumilla Sadar Dakshin, and Burichang faced 60% to 80% risk. The northern portion of the district was at extremely high risk (>90%) for neck blast disease in 2022, while all Upazilas were at high risk. In 2023, every Upazila has a lower risk (<20%) with the exception of Nangalkot and the central region of Cumilla Sadar Dakshin. The danger of NBD increased in 2024, and Debidwar, Cumilla Sadar Dakshin, and Nangalkot may be considered high-risk areas. Nangalkot showed >90% probability of exceeding the specified NBD severity threshold during the period.
Green to red-colored points depict lower to higher levels of risk prone areas of NBD. The maps were created using R software (version R-4.0.3).
Fig 12 uses the indicator kriging approach to examine the spatial probability of Neck Blast Disease severity above 20% across the study Upazilas during the Boro seasons from 2018 to 2025 except 2021 and 2022 where above 5% has been consider for lower prevalence of disease. In 2018–2019, most areas of northern part show the high risk (70%–90%) while other region shows relatively lower risk (40%−60%) than these. In 2019–2020, disease probability intensified sharply, with extremely high risk (>90%) concentrated in central Cumilla Sadar Dakshin and southern Nangalkot and Debidwar and Barura while north part of Cumilla Adarsha Sadar faced less than 20% risk of disease. In 2020–2021, elevated risk persisted in Debidwar, Barura, and some part of central of the district (60%–80%), whereas other regions remained moderate (20%–40%). In 2021–22, overall disease severity was very low; therefore, this season was mapped using a > 5% severity threshold because of the very low overall disease severity and is therefore not directly comparable with years using the > 20% threshold. In 2022–2023, high probability zones (60%–80%) reappeared in Debidwar and Barura, while surrounding Upazilas experienced moderate levels. In 2023–2024, most areas showed relatively lower probability (<40%), except localized hotspots in Debidwar and Barura (70%–90%). The 2024–2025 season again showed strong resurgence, with Burichang, Cumilla Adarsha Sadar, and southern Nangalkot emerging as high-risk zones (>80%), indicating a return of severe blast pressure in these locations.
A > 20% severity threshold was used except in 2021–22, when >5% was applied due to low disease severity; therefore, 2021–22 is not directly comparable with other years. Green to red indicates lower to higher probability. Maps were generated using R (version 4.0.3).
Discussion
The spatio-temporal analysis of Neck Blast Disease (NBD) severity revealed a highly heterogeneous distribution across the study area. It underscores the necessity of location-specific, adaptive management strategies. The application of geostatistical techniques such as Local Indicators of Spatial Association (LISA) and Kriging effectively identified persistent disease hotspots and quantified the probability of high-severity outbreaks.
Spatial and temporal dynamics of NBD vulnerability
The analysis demonstrated distinct vulnerability areas between the T. Aman and Boro rice seasons. In the Aman season (2017–2024), Nangalkot was consistently identified as the most vulnerable Upazila, characterized by the highest mean severity (approximately 45%–60%) and explicated maximum yield loss. Neck blast is the most devastating disease, capable of inflicting catastrophic yield losses ranging from 70% to 80% [34]. Kriging models consistently assigned Nangalkot a greater than 70% risk probability, increasing in a 90% chance of infection in the final year of study. This persistent vulnerability may reflect unmeasured local factors, such as favorable micro-environmental conditions, susceptible rice varieties, cropping practices, or disease-management behavior; however, these factors were not directly measured in the present study [19,35,36]. Conversely, Chandina exhibited low mean severity (0–15%), serving as a control area. The emergence of High–High clusters in Debidwar and Cumilla Adarsha Sadar using LISA indicates significant localized, self-reinforcing disease aggregation points that demand targeted, intensive control measures.
The Boro season (2018–2025) presented a more dynamic epidemiological landscape. While Muradnagar was identified as the overall most vulnerable (45%−60%) Upazila, the spatial distribution of risk zones shifted temporally. High-risk areas, indicated by High–High clusters, moved between Muradnagar, Nangalkot, Debidwar, and Burichang across different years. This spatial non-stationarity indicates that Boro-season Neck Blast risk varied considerably across years and locations [37]. The recent resurgence of high probability zones (>80%) in Burichang, Cumilla Adarsha Sadar, and Nangalkot (2024–2025) warrants immediate risk mitigation efforts in the short term.
To the best of our knowledge, there was no previous study which was done in spatial explicit mode to identify hotspot of Neck Blast Disease in any season in Cumilla District. The results do agree with global research on rice blast. Rice blast is one of the most devastating diseases of rice globally, and it is a major threat to the world food security, according to Asibi et al. [19]. Kihoro et al. showed that rice blast has socioeconomic impacts on the farmers’ livelihood in the Greater Mwea region in Kenya [22]. Similarly, Simkhada and Thapa reported that the weather factors that favor the blast outbreaks are cloudy weather, high relative humidity, low night temperature and long dew duration [34]. A spatial epidemiology study by Amoghavarsha et al. classified the areas of Karnataka, India, as being at a high risk of rice blast in various rice ecosystems and showed how spatial analysis tools like autocorrelation analysis and interpolation methods can be used to detect disease-prone areas [38]. The current results are consistent with the previous studies indicating the complex spatial variability in Neck Blast risk and the need for spatial monitoring and management to mitigate the disease impact.
Influence of climatic trends
Observed climatic trends between 2017 and 2025 provide context for the observed disease patterns. Both Aman and Boro seasons exhibited an upward trend in both maximum and minimum temperatures. Together with the declining rainfall trend in Aman, these climatic shifts were associated with changes in NBD severity and may have contributed to conditions favorable for disease development.
A notable observation was the lowest NBD severity period during the Boro seasons of 2020–21 and 2021–22. This period coincides with a decline in Relative Humidity (RH) following a peak in 2019–2020. Given that high RH and temperature are critical factors for blast infection, the period of lower Boro RH coincided with reduced NBD severity, suggesting a possible association between humidity conditions and disease development. In a study of environmental factor for rice blast disease, Simkhada K. et al. (2022) demonstrated that the most conducive conditions for disease outbreaks are cloudy weather, high relative humidity, low night temperatures, and prolonged dew duration [34]. However, the subsequent rebound in severity confirms the complex, non-linear relationship between climatic variables and disease expression which suggests that other factors, such as varietal shifts or inoculum pressure, may override marginal climatic suppression. These data emphasize the need for robust predictive modeling that integrates weather variables with spatial context.
Performance of geostatistical modeling
The performance of the geostatistical models validated their utility in high-resolution risk mapping. Indicator Kriging proved particularly effective by mapping the spatial probability of severity exceeding a critical 20% yield-loss threshold, offering a clear, actionable metric for stakeholders. The variability in the best-fit semivariogram models, Spherical dominance in Aman and a shift to Exponential in later Boro years, confirms the distinct spatial dependence characteristics of the disease in each season. This result is coincided with the study of Amoghavarsha C, et al., (2022) which took place in Karantaka [38]. The spherical model, indicative of a continuous, patch-like distribution, was appropriate for the more established Aman crop cycle, while the exponential model suggests faster diminishing spatial correlation, aligning with the more sporadic and climate-sensitive nature of Boro outbreaks. This finding is critical: effective spatial epidemiology requires season- and year-specific model parameterization to achieve accurate risk assessment.
In summary, the study successfully delineated stable and transient high-risk zones for NBD. Nangalkot and Muradnagar are designated as primary management targets. The dynamic nature of the Boro epidemic, in contrast to the more persistent Aman risk, underscores the critical need for an integrated disease management framework that leverages high-resolution, geostatistical risk maps, informed by seasonal climatic forecasting, to deploy fungicides and resistant varieties strategically.
The interpretation of this study’s results is constrained by several limitations. The most significant limitation is the lack of regional or high-resolution climatic data for the study sub-regions, which prevented precise assessment of the localized effects of rainfall, temperature, and humidity on disease prevalence and intensity. Because the analysis was conducted at the Upazila level, finer field-level variability in NBD severity may not have been fully captured. therefore, the risk maps should be interpreted as Upazila-level patterns intended for regional planning and surveillance rather than as predictions for individual fields. The spatial analysis also did not incorporate cultivar-specific information, crop-management practices, or fungicide-use data, as these were not consistently available across Upazilas and years. Another limitation is the small number of spatial units (7 Upazilas in Aman and 10 in Boro), which may reduce the statistical power and stability of Moran’s I, LISA, clustering, semivariogram fitting, and Kriging predictions. Because no multiple-comparison correction was applied to repeated LISA analyses, significant local clusters should be interpreted cautiously as exploratory spatial patterns. Nevertheless, the multi-year dataset provides valuable insights into persistent and changing NBD patterns across seasons and years. Although cross-validation metrics were used to assess model performance, spatial prediction uncertainty was not mapped explicitly; therefore, the interpolated risk surfaces should be interpreted cautiously, particularly in areas with sparse observations. Future studies should incorporate cultivar-specific resistance profiles, crop-management information, fungicide-use data, finer-resolution spatial units, and high-resolution climatic variables to better explain the spatial and temporal variability of NBD hotspots.
Conclusion
The present study successfully employed indicator kriging to demarcate the spatio-temporal variation of the risk of Neck Blast Disease (NBD) over the study area, with high spatial heterogeneity and extreme inter-annual variation being evidenced. Spatially, the research identified recurring susceptible areas, with particular emphasis on demarcating Nangalkot as a chronic hotspot area where the highest maximum mean NBD severity was recorded during the Aman season. Temporally, Boro season had complex epidemic cycles, namely sudden intensification (e.g., 2019–2020, in which risk probability was over 90% in various central and southern Upazilas) followed by sudden declines (e.g., 2021–2022). The recurring and very localized re-appearance of high-probability areas, particularly in Debidwar and Barura, confirms that NBD risk is not homogeneous. These findings call for the application of highly location-specific and adaptive monitoring and management strategies to counter these high-probability zones in a bid to avert devastating losses in yields.
Supporting information
S1 File. Year-wise list of study Upazilas by rice season.
This table presents the Upazilas included in the study by year and season for Aman and Boro rice ecosystems in Cumilla District.
https://doi.org/10.1371/journal.pone.0358537.s001
(PDF)
S2 File. Mann–Kendall trend analysis for Aman-season weather variables.
This table presents Kendall’s tau, p-values, and trend interpretations for maximum temperature, minimum temperature, rainfall, and relative humidity during the Aman season.
https://doi.org/10.1371/journal.pone.0358537.s002
(PDF)
S3 File. Mann–Kendall trend analysis for Boro-season weather variables.
This table presents Kendall’s tau, p-values, and trend interpretations for maximum temperature, minimum temperature, rainfall, and relative humidity during the Boro season.
https://doi.org/10.1371/journal.pone.0358537.s003
(PDF)
S1 Fig. Moran’s I values for the Aman season.
This figure presents year-wise Moran’s I values used to assess spatial autocorrelation of Neck Blast Disease severity during the Aman season.
https://doi.org/10.1371/journal.pone.0358537.s004
(TIFF)
S2 Fig. Moran’s I p-values for the Aman season.
This figure presents the corresponding year-wise p-values for Moran’s I analysis during the Aman season.
https://doi.org/10.1371/journal.pone.0358537.s005
(TIFF)
S3 Fig. Moran’s I values for the Boro season.
This figure presents year-wise Moran’s I values used to assess spatial autocorrelation of Neck Blast Disease severity during the Boro season.
https://doi.org/10.1371/journal.pone.0358537.s006
(TIFF)
S4 Fig. Moran’s I p-values for the Boro season.
This figure presents the corresponding year-wise p-values for Moran’s I analysis during the Boro season.
https://doi.org/10.1371/journal.pone.0358537.s007
(TIFF)
Acknowledgments
The authors gratefully acknowledge the BRRI Agromet Lab for its significant contribution to this research, as well as the BRRI Regional Station, Cumilla, and the Plant Pathology Division of the Bangladesh Rice Research Institute (BRRI) for their valuable technical and institutional support.
References
- 1. Hossain M, Fischer KS. Rice research for food security and sustainable agricultural development in Asia: achievements and future challenges. GeoJournal. 1995;35(3):286–98.
- 2. Bishwajit G, Sarker S, Kpoghomou M-A, Gao H, Jun L, Yin D, et al. Self-sufficiency in rice and food security: a South Asian perspective. Agric & Food Secur. 2013;2(1).
- 3. Saha I, Durand-Morat A, Nalley LL, Alam MJ, Nayga R. Rice quality and its impacts on food security and sustainability in Bangladesh. PLoS One. 2021;16(12):e0261118. pmid:34972112
- 4. Jamal MR, Kristiansen P, Kabir MJ, Lobry de Bruyn L. Challenges and adaptations for resilient rice production under changing environments in Bangladesh. Land. 2023;12(6):1217.
- 5.
Alam MJ, Mazadul M. Food security and food safety in Bangladesh: Way forward. 2024.
- 6. Dogar MM, Kucharski F, Azharuddin S. Study of the global and regional climatic impacts of ENSO magnitude using SPEEDY AGCM. J Earth Syst Sci. 2017;126(2).
- 7. Dogar MM, Stenchikov G, Osipov S, Wyman B, Zhao M. Sensitivity of the regional climate in the Middle East and North Africa to volcanic perturbations. JGR Atmospheres. 2017;122(15):7922–48.
- 8. Dogar MM, Kucharski F, Sato T, Mehmood S, Ali S, Gong Z, et al. Towards understanding the global and regional climatic impacts of Modoki magnitude. Global and Planetary Change. 2019;172:223–41.
- 9. Dogar MM, Hermanson L, Scaife AA, Visioni D, Zhao M, Hoteit I, et al. A review of El Niño southern oscillation linkage to strong volcanic eruptions and post-volcanic winter warming. Earth Syst Environ. 2022;7(1):15–42.
- 10. Dogar MM, Fujiwara M, Zhao M, Ohba M, Kosaka Y. ENSO and NAO linkage to strong volcanism and associated post‐volcanic high‐latitude winter warming. Geophysical Research Letters. 2024;51(1).
- 11. Dogar MM, Fujiwara M, Ohba M, Almazroui M. Investigating post-eruption amplified winter cooling in the Middle East and North Africa—unraveling ENSO and NAO dynamics. Environ Res Commun. 2025;7(4):045012.
- 12. Dogar MM, Watanabe S, Fujiwara M, Abid MA, Habeebullah TM, Khan B. Tropical volcanic impacts on MENA climate via ENSO and NAO dynamics in a high-top model. npj Clim Atmos Sci. 2025;8(1).
- 13. Dogar MMA, Almazroui M. Revisiting the strong and weak ENSO teleconnection impacts using a high-resolution atmospheric model. Atmospheric Environment. 2022;270:118866.
- 14. Dogar MM, Sato T. Analysis of climate trends and leading modes of climate variability for MENA Region. JGR Atmospheres. 2018;123(23).
- 15. Dogar MM, Sato T. Regional climate response of middle eastern, african, and south asian monsoon regions to explosive volcanism and ENSO forcing. JGR Atmospheres. 2019;124(14):7580–98.
- 16. Bell AR, Bryan E, Ringler C, Ahmed A. Rice productivity in Bangladesh: What are the benefits of irrigation?. Land Use Policy. 2015;48:1–12.
- 17. Das LC, Mohiul Islam ASM, Aktar S, Sultana F. Climate change challenges in Bangladesh. Int J Big Data Mini Glob Warm. 2024;06(01).
- 18. Hussain AM, Bari E. Economic Valuation of Loss and Damage in the Crop Sector of Bangladesh. 2024. https://ssrn.com/abstract=5274407
- 19. Asibi AE, Chai Q, Coulter JA. Rice blast: a disease with implications for global food security. Agronomy. 2019;9(8):451.
- 20. Tasfia M. Survey on neck blast disease of rice caused by Magnaporthe oryzae and its in-vitro management (Doctoral dissertation, Department of Plant Pathology, Sher-e-Bangla Agricultural University).
- 21.
Prasanna GS. Chapter-3 Rice Blast. Tamil Nadu Agricultural University.
- 22. Kihoro J, Bosco NJ, Murage H, Ateka E, Makihara D. Investigating the impact of rice blast disease on the livelihood of the local farmers in greater Mwea region of Kenya. Springerplus. 2013;2(1):308. pmid:23888278
- 23. Verma R, Bisht AS, Roy B, Kumar S. Unveiling the Dynamics of rice blast: insights into pathogenesis, epidemiology, and management. Detection, Diagnosis and Management of Air-Borne Diseases in Agricultural Crops. Springer Nature Singapore. 2025. p. 143–59.
- 24.
Kotyal K. Development and characterization of doubled haploids generated from F1 anthers of drought adaptive traits pyramided parents in rice (Oryza sativa L.). Bangalore: University of Agricultural Sciences.
- 25. Hossain MM, Ahmed S, Alam MS, Hossain A. Adverse effects of heat shock in rice (Oryza sativa L.) and approaches to mitigate it for sustainable rice production under the changing climate: A comprehensive review. Heliyon. 2024;10(24):e41072.
- 26.
Mohapatra PK, Sahu BB. Panicle architecture of rice and its relationship with grain filling. Springer International Publishing. 2022.
- 27. Devanna BN, Jain P, Solanke AU, Das A, Thakur S, Singh PK, et al. Understanding the dynamics of blast resistance in rice-magnaporthe oryzae interactions. J Fungi (Basel). 2022;8(6):584. pmid:35736067
- 28. Sarker MAR, Alam K, Gow J. Performance of rain-fed Aman rice yield in Bangladesh in the presence of climate change. Renew Agric Food Syst. 2017;34(04):304–12.
- 29. Ghatak A, Willocquet L, Savary S, Kumar J. Variability in aggressiveness of rice blast (Magnaporthe oryzae) isolates originating from rice leaves and necks: a case of pathogen specialization?. PLoS One. 2013;8(6):e66180. pmid:23776629
- 30. Dey RK, Toma SR, Rahman N, Khatun MS, Begum MS, Hoque M. Phenology and yield performances of HYV and local rice cultivars during boro and aman seasons in haor ecology of sylhet. J Bangladesh Agric Univ. 2025;23(2):129–40.
- 31.
International Rice Research Institute. Standard evaluation system (SES). Manila, Philippines: International Rice Research Institute. 2013.
- 32. Hossain M, Ali MA, Hossain MD. Occurrence of blast disease in rice in Bangladesh. American Journal of Agricultural Science. 2017;4(4):74–80.
- 33.
Padmanabhan SY. Estimating losses from rice blast in India. The Rice Blast Diseases. Baltimore, Maryland: Johns Hopkins. 1965.
- 34. Simkhada K, Thapa R. Rice blast, a major threat to the rice production and its various management techniques. Turkish JAF SciTech. 2022;10(2):147–57.
- 35. Suzuki H. Meteorological factors in the epidemiology of rice blast. Annu Rev Phytopathol. 1975;13(1):239–56.
- 36. Luo Y, TeBeest DO, Teng PS, Fabellar NG. Simulation studies on risk analysis of rice leaf blast epidemics associated with global climate change in several Asian countries. Journal of Biogeography. 1995;22(4/5):673.
- 37. Singh O, Bathula J, Singh DK. Rice blast modeling and forecasting. Int J Chem Stud. 2019;7(6):2788–99.
- 38. Amoghavarsha C, Pramesh D, Sridhara S, Patil B, Shil S, Naik GR, et al. Spatial distribution and identification of potential risk regions to rice blast disease in different rice ecosystems of Karnataka. Scientific Reports. 2022;12(1):7403.