Remote sensing-based detection of brown spot needle blight: a comprehensive review, and future directions

Introduction

Forest disturbances, such as insect infestations, disease outbreaks, and climate change, drive pathogen spread and threaten forest health, requiring advanced monitoring and mitigation (Dudney et al., 2021; Hartmann et al., 2022; Roshani et al., 2022). Pine species (Pinus spp.) are predominantly distributed across the Northern Hemisphere and hold significant economic and ecological value. They have also been introduced to temperate and subtropical regions of the Southern Hemisphere for timber production and ornamental use (Costanza et al., 2018; Ryu et al., 2018). However, pine forests worldwide face increasing threats from needle diseases, which are caused by fungal and oomycete pathogens that directly infect, and damage conifer foliage (Wingfield et al., 2015; Drenkhan et al., 2016).

Among these, Dothistroma Needle Blight (DNB or Red Band) is one of the most destructive, affecting over 80 pine species, with Corsican Pine (P. nigra J.F. Arnold subsp. laricio (Poir.)) being particularly susceptible. This disease is caused by the hemibiotrophic fungus Dothistroma septosporum or D. pini (sexual stage: Mycosphaerella pini) (Watt et al., 2009; Mullett, Peace & Brown, 2016). Similarly, Lophodermium Needle Cast, caused by Lophodermium seditiosum, primarily affects Scots Pine (P. sylvestris) (Jansons et al., 2024). Red Needle Cast (Phytophthora pluvialis) has emerged as a significant threat to pine and Douglas fir plantations in both the Northern and Southern Hemispheres, particularly in regions with high precipitation. This disease has been increasingly reported in radiata pine (P. radiata) and Douglas fir (Pseudotsuga menziesii), where it causes needle browning, premature defoliation, and reduced growth rates (Dick et al., 2014; Gomez-Gallego et al., 2019).

Additionally, Pine Needle Rust (Coleosporium asterum) infects various pine species (P. nigra, P. banksiana, P. resinosa, P. ponderosa, P. mugo, P. sylvestris) as well as plants in the Asteraceae family (University of Minnesota Extension). Black Spot Needle Blight, caused by Pestalotiopsis neglecta, is common in P. sylvestris var. mongolica (Ma et al., 2024). Brown Spot Needle Blight (BSNB), caused by Lecanosticta acicola, primarily affects Longleaf Pine (P. palustris) and Loblolly Pine (P. taeda) (Cunningham, 2022; Meinecke et al., 2024b).

Since 2016, outbreaks of needle diseases have been increasingly reported in the southeastern United States, particularly affecting loblolly pine (P. taeda), a key timber species in the region (Meinecke et al., 2024a). Symptoms initially manifest as small, irregular yellow spots that develop into larger, dark orange to brown lesions, often resin-soaked and encircled by a yellow halo (Mullett et al., 2018; Aglietti et al., 2021). The causal agent of these outbreaks has been identified as the native fungal pathogen L. acicola (Thum.) Syd. (formerly Mycosphaerella dearnessi, syn. Scirrhia asicola), responsible for BSNB foliar disease (Meinecke et al., 2024a).

Historically, BSNB foliar disease has not posed a major threat to P. taeda; however, the increasing scale and severity of recent outbreaks raise significant concern. Initially recognized in the southeastern United States, BSNB has severely impacted P. palustris, particularly in Christmas tree farms (Skilling & Nicholls, 1974). By the 1980s, timber losses caused by BSNB exceeded 453,000 cubic meters annually in southern pine species, including P. palustris (Cordell, Anderson & Kais, 1990). Over time, BSNB has emerged as a global concern, with invasive populations reported in many countries (EPPO, 2024b). The disease exhibits a broad host range, affecting over 70 taxa, primarily within the Pinus genus, including species of Cedrus and Picea (Tubby et al., 2023).

The southeastern United States, which accounts for 17% of global timber production is critical in national forestry (FAO Global Forest Resources Assessments, 2015). This region comprises 61% of the nation’s planted forests and 57% of its total wood volume, emphasizing the economic importance of key pine species such as loblolly pine (P. taeda), shortleaf pine (P. echinata), and slash pine (P. elliottii) (Oswalt et al., 2014; Fagan et al., 2018). States such as Alabama, Mississippi, Louisiana, and Arkansas are particularly reliant on these timber species, making them vulnerable to emerging threats such as BSNB (Datta, 2021; Meinecke et al., 2024a). In 2022, the Alabama Forestry Commission reported BSNB outbreaks in 36 of the state’s 67 counties, with wet summers and mild winters creating conditions favorable for L. acicola infection (U.S. Forest Service, 2022). This increasing prevalence highlights the need for improved strategies to mitigate its impact. As the disease progresses, it elevates stress susceptibility in trees, potentially leading to mortality with climatic factors playing a significant role in its spread (Ogris et al., 2023). Similar concerns exist globally with Dothistroma needle blight, which affects pine forests in western Canada and Fennoscandia (Drenkhan et al., 2016). The ecological and epidemiological parallels between Dothistroma species and L. acicola suggest that BSNB could expand its geographic range and pose a broader threat (Raitelaitytė et al., 2023). With changing climate conditions, proactive measures are needed to protect forestry resources on regional and global scales.

Remote sensing (RS) technologies have emerged as effective tools for identifying and monitoring needle disease symptoms such as Dothistroma needle blight. Studies using aerial and field-based spectroscopy imaging have been successfully applied to identify disease symptoms and assess severity levels (Coops et al., 2003; Watt et al., 2021; Watt et al., 2023). RS-based methods often involve the utilization of vegetation indices (VIs) derived from spectral band and different modeling techniques to detect early changes caused by the diseases (Singh et al., 2020; Marvasti-Zadeh et al., 2023). While RS applications for DNB have shown significant progress, the potential of these technologies for addressing BSNB remains largely untapped. Currently, there is a notable lack of direct studies or models developed specifically for L. acicola, limiting the ability to monitor and predict the disease’s impact. Addressing this gap by adapting RS techniques is important for enhancing detection, identifying spread pattern, and improving surveillance methodologies.

Traditional methods like on-site observations are essential but labor-intensive, time-consuming, and spatially constrained (Martinelli et al., 2015; Schnebele et al., 2015). In contrast, RS enables disease monitoring by assessing epidemic distribution, severity, and vegetation health through multi-scale and time-series analyses (Stone & Mohammed, 2017; Lausch et al., 2018; Panzavolta et al., 2021; Han et al., 2022; Lassalle & Fabre, 2022; Cao et al., 2022). While RS reduces the dependence on extensive field sampling, it complements ground-based observations to provide a more comprehensive understanding of disease dynamics (Yuan, Liu & Zhang, 2017; Pang et al., 2019; Zhang et al., 2019), overcoming the spatial and temporal limitations of discrete sampling sites.

Given the ability of L. acicola to cause significant economic and ecological damage, an important question arises: To what extent can remote sensing be utilized for the detection and monitoring of BSNB symptoms? To address this, we conducted a comprehensive bibliometric analysis using the Web of Science (WoS) database and used VOSviewer to visualize research trends based on published article keywords. The results showed a notable gap in the literature, with limited direct research on RS applications for BSNB detection and no prior bibliometric studies specifically focused on this disease. To broaden our understanding, we also examined RS applications in other pine needle diseases, such as Dothistroma Needle Blight. This analysis identified potential advantages and technological limitations to inform future approaches to BSNB detection.

This review addresses two key research questions: (1) whether RS and geospatial-based studies on BSNB have previously been published, and (2) what the advantages and limitations of RS applications in other pine needle disease research are, and whether these methodologies can provide a foundation for future BSNB investigations.

Survey methodology

This study follows a systematic bibliometric approach, consisting of three key stages: data collection, data processing, and visualization to assess research trends and gaps in BSNB detection using RS.

Data processing and visualization

Following data extraction, bibliometric analysis was conducted using VOSviewer (version 1.6.3) to identify key research trends, co-occurrence networks, and thematic clusters (Van Eck & Waltman, 2017). Bibliometric maps were generated to visualize collaborative relationships, research hotspots, and existing gaps (Zhao, Tang & Zou, 2019; Jia & Mustafa, 2022). A flowchart (Fig. 1) outlines the complete data collection and processing workflow.

Results

Bibliometric analysis (research trends)

Figure 2 shows various pine diseases, their causal pathogens, and associated key terms, reflecting research focus areas based on published studies over time. The color gradient (blue to yellow) represents publication years, with blue (~2012) for older studies, green (~2014–2016) for intermediate research, and yellow (~2018) for recent studies. Strongly interconnected nodes highlight well-established research areas while emerging topics indicate growing scientific interest. BSNB, though historically less studied, has recently gained attention, particularly concerning its relationship with climate change and host susceptibility.

Figure 3 presents a co-occurrence network of key terms in BSNB research, showing relationships among major clusters. Larger nodes indicate frequently occurring keywords, while different colors represent thematic clusters. The network highlights strong associations between Brown Spot Needle Blight, pathogen dynamics, climate change, and disease resistance, reflecting active research on host range, environmental influences, and molecular factors. The presence of terms related to genetic studies, dispersal mechanisms, and host-pathogen interactions suggests ongoing efforts in epidemiology. However, remote sensing applications are notably absent, supporting the need for further exploration in this domain.

The following sections provide a detailed investigation into major aspects of BSNB research. This includes an analysis of BSNB’s global distribution, symptomatology, the impact of climate change on its spread, and existing disease detection methods.

BSNB symptomology

BSNB symptoms vary depending on the host species. They generally begin as small yellow, grey-green, or reddish-brown spots with distinct margins, which later darken to brown and may be surrounded by a yellow halo (Skilling & Nicholls, 1974) (Fig. 4). The presence of these characteristic brown spots on pine needles led (Siggers, 1932) to coin the term ‘brown spot needle blight.’ In some cases, these spots may appear resin-soaked, depending on the host species. Infected needles generally die from the tip downward and eventually fall, with the disease spreading upward from the lower branches (Skilling & Nicholls, 1974; Van Der Nest et al., 2019). As the infection progresses, lesions enlarge, leading to tissue death and premature defoliation (Adamson, Drenkhan & Hanso, 2015). In North America, BSNB infections can occur throughout the year, but L. acicola produces the most spores during the summer months, peaking in June and August (Kais, 1975).

BSNB symptoms can sometimes be confused with Dothistroma Needle Blight since both diseases cause needle discoloration, but BSNB can be positively identified with microscopy or molecular analysis. One key morphological difference is that DNB typically forms distinct red bands around the infection site, while BSNB produces brown spots. In some DNB cases, the red bands in DNB can appear dark enough to resemble brown spots, which can lead to misdiagnosis (Barnes et al., 2016). Morphological identification of BSNB-causing fungus can be used to identify these needle characteristics as well as the morphological fungal characteristics of cultured samples (Dreaden et al., 2024). Molecular identification can also provide diagnostic insights, either using conventional PCR that target known diagnostic genes (Groenewald et al., 2007; Ioos et al., 2010) or by using Loop-Mediated Isothermal Amplification (LAMP) assays, a rapid and efficient alternative to traditional PCR (Aglietti et al., 2021).

Advancements in remote sensing for disease detection

Reflectance spectroscopy is the predominant technique for in-field vegetation analysis and has been widely applied in scientific research for ecological and environmental assessments (Pandey et al., 2017; Lechner, Foody & Boyd, 2020; Singh, 2024). This method is based on the interaction between plants and light, where radiation is either reflected, absorbed, or transmitted, depending on internal and external plant dynamics such as structure, chemical composition, and water content (Zahir et al., 2024). Pathogens disrupt these dynamics by altering plant pigmentation, water content, and tissue function, which leads to structural modifications that manifest as disease symptoms (Tardieu et al., 2017). The physiological and phenological changes in plants are intricately linked to the nature of the pathogen (Dixon, 2012). These alterations produce distinct spectral patterns that facilitate spectral discrimination to differentiate between healthy and diseased leaves and canopies (Golhani et al., 2018). Therefore, remote sensing serves as a “radiodiagnosis” approach for plant disease investigation, allowing continuous monitoring over broad spatial scales (Decuyper et al., 2022; Singh, 2022a, 2022b).

RS technology involves the use of different platforms and sensors (Fig. 5). Spaceborne remote sensing enables large-scale disease detection and monitoring by capturing extensive spatial data (Massey et al., 2023). Airborne platforms, equipped with advanced RS capabilities, facilitate forest monitoring at finer spatial resolutions. These systems offer operational flexibility and the ability to generate high-resolution imagery. UAVs, in particular, provide low-altitude flight capabilities and very high-resolution imaging, making them particularly effective for the early detection of disease outbreaks (Berie & Burud, 2018; Singh, 2022a). Their adaptability enhances responsiveness in disease monitoring and helps overcome limitations associated with satellite-based observations (Li et al., 2020).

The fundamental RS sensor is an optical imaging system, similar to a conventional camera, capable of collecting data beyond the visible spectrum, extending into the infrared and thermal regions (Reddy, 2018). These sensors come in various configurations: multispectral sensors capture a limited number of broad bands, whereas hyperspectral sensors acquire thousands of narrow spectral bands, providing a more detailed spectral profile. Radar-based RS, an active system operating in the microwave region, is particularly effective under cloudy conditions due to minimal atmospheric absorption. Both lidar (Light Detection and Ranging) and SAR (Synthetic Aperture Radar) are active RS technologies that emit energy pulses and record reflected signals. Lidar operates within the visible to near-infrared spectrum and, as an active sensor, functions independently of external light sources, making it suitable for nighttime data acquisition (Lohar et al., 2021). SAR, in contrast, possesses unique capabilities, such as detecting surface roughness and land cover characteristics (Ouchi, 2013). Unlike lidar, SAR can penetrate dense cloud cover and operate effectively in darkness, whereas lidar provides high-resolution spatial data under clear sky conditions or at night. These active sensors provide distinct advantages for data collection under various environmental conditions. The utilization of multiple remote sensing platforms and sensors holds the potential to detect and monitor various forest diseases.

Assessing disease symptoms using multi-source remote sensing: needle disease case studies

For the second objective of this article, which explores RS applications in pine needle disease research, this section reviews key studies. RS has been extensively utilized for disease detection and progression monitoring in pine forests (Fig. 6). Techniques such as hyperspectral imaging, multispectral, lidar, and satellite-based observations have played a key role in identifying early symptoms, tracking disease dynamics, and developing predictive models. These approaches enable large-scale disease surveillance and support proactive management strategies.

RS-based research on pine needle diseases has evolved significantly, with early studies laying the foundation for disease assessment using airborne and field-based spectral analysis. Coops et al. (2003) used airborne hyperspectral imagery (CASI-2) to assess Dothistroma Needle Blight in P. radiata plantations in Australia, finding strong correlations between crown reflectance and ground-based severity estimates. This study laid the foundation for canopy health assessment using hyperspectral remote sensing. In the same year, Stone, Chisholm & McDonald (2003) employed portable chlorophyll fluorometry to analyze DNB-affected P. radiata needles, identifying the 709/691 nm reflectance ratio as a key spectral indicator of disease severity.

Beyond hyperspectral approaches, thermal and lidar-based techniques have also been used to detect pine needle diseases. Smigaj et al. (2019a) used UAV-borne thermal imaging and lidar to detect Red Band Needle Blight (RBNB) in P. sylvestris, finding significant correlations between crown temperature and disease severity (R² = 0.27–0.41). In a related study, Smigaj et al. (2019b) combined hyperspectral and lidar data to classify infected trees, achieving 80.9% accuracy, which improved to 96.7% using stepwise discriminant function analysis. Expanding to large-scale monitoring, Watt et al. (2021) analyzed 6,276 observations of DNB severity in New Zealand’s P. radiata plantations over 37 years, comparing parametric and non-parametric models to develop a fine-scale disease severity map. Watt et al. (2023) further refined this approach using UAV-collected hyperspectral data, integrating 3D radiative transfer models (PRO4SAIL) with Random Forest algorithms, improving prediction accuracy from R² = 0.52 to R² = 0.85.

As Sentinel-2 satellite imagery has proven effective for forest disease monitoring, its 13 spectral bands, particularly the red-edge band (705–740 nm), have been widely used to detect chlorophyll content changes, water stress, and early symptoms of needle diseases (Wang et al., 2018; Zhang et al., 2018). Meneghini et al. (2022) demonstrated its utility for detecting White Pine Needle Damage (WPND) in P. strobus forests, identifying NDII (Normalized Difference Infrared Index) as the most effective index for tracking defoliation, with early July as the optimal detection window. While severe WPND was detectable, light to moderate damage remained challenging, with classification accuracies not exceeding 75%. The study emphasized the need for higher spatial resolution sensors (e.g., PlanetScope, WorldView) and plot-based biochemical sampling to improve detection accuracy.

More recently, Timalsina et al. (2024) explored hyperspectral RS and foliar traits for WPND detection. Their study demonstrated that a Random Forest (RF) model based solely on spectral vegetation indices (SVIs) achieved 87% accuracy and a Kappa coefficient of 0.68, effectively classifying trees as asymptomatic or symptomatic. When field-measured foliar traits were combined with RS data, accuracy dropped to 77%, with a lower Kappa coefficient (0.46). These findings highlight the potential of hyperspectral data and machine learning models for timely WPND detection, aiding forest management. The study suggests that higher-resolution hyperspectral data from UAVs could improve local-scale assessments, while Sentinel-2’s red-edge bands could enable broader landscape-level WPND monitoring. Building on these advancements, Watt et al. (2024) developed a novel methodology integrating Sentinel-2 satellite imagery and climatic data to predict Red Needle Cast (RNC) outbreaks in P. radiata forests in New Zealand 7–8 months before peak disease expression. The study utilized the Red/Green Index Difference (R/G) to classify areas as disease-free or showing RNC symptoms, with solar radiation, relative humidity, rainfall, and maximum air temperature identified as key climatic predictors. Using 1,976 plots, a RF model was developed, achieving 89% accuracy and an F1 score of 0.89 in predicting RNC incidence. This approach enables large-scale, early disease detection, allowing for targeted monitoring and treatment to mitigate the disease’s impact before its peak expression in spring.

While Pine Wilt Disease (PWD) is a vascular wilt disease rather than a needle disease, it has been extensively studied in remote sensing-based detection using ML and DL models (Fig. 6). Its frequent inclusion in RS applications highlights its value as a reference system for automated disease monitoring in pine forests. The well-established methodologies used for PWD detection provide a foundation for developing similar approaches tailored to pine needle diseases. However, its inclusion in this study is based solely on its relevance to RS-based methodologies, rather than any direct biological comparability to BSNB or other needle diseases.

PWD, caused by the pinewood nematode (Bursaphelenchus xylophilus), has been widely monitored using a combination of UAV-based multispectral imaging, high-resolution satellite data, and ML/DL-driven classification models. Mantas et al. (2022) demonstrated that Sentinel-2 imagery could detect PWD with 95% accuracy, while Yu et al. (2021a) applied UAV-based multispectral imaging, achieving classification accuracies between 60.98–66.7%. The integration of ML/DL models has significantly enhanced PWD detection capabilities, improving accuracy, efficiency, and scalability. Li et al. (2021) demonstrated the effectiveness of CNN-based models such as YOLOv4-Tiny-3Layers, achieving 84.88% precision for airborne edge computing. UAV-based Faster R-CNN models further improved classification accuracy, reaching 60.98–66.7% (Wu et al., 2021), while Zhang et al. (2022b) achieved 95.24% accuracy using SVM and Genetic Algorithm models on hyperspectral UAV data, showcasing the potential of advanced classification techniques. Beijing-2 satellite data combined with CNN models achieved 99.4% accuracy in rapid tree identification (Zhou et al., 2022b), further demonstrating the potential for scalable disease monitoring. The application of deep learning methods continues to enhance PWD detection, with models such as YOLOv5-PWD integrating UAV and satellite imagery for large-scale surveillance (Cai et al., 2023). Additionally, hybrid approaches leveraging UAV-based hyperspectral data and ML models, such as RF and SVM, have shown high accuracy, with some studies achieving over 95% precision in monitoring PWD infection stages (Zhang et al., 2022b). The growing reliance on ML/DL for PWD detection highlights its crucial role in precision forestry, offering cost-effective, scalable, and highly accurate solutions for disease monitoring. Table 2 provides a comprehensive summary of ML/DL models applied to PWD detection, highlighting the various methodologies, accuracies, and research gaps.

Challenges and future outlook

Remote sensing has significantly improved forest disease detection and monitoring, yet several challenges and limitations persist. A primary issue is distinguishing biotic and abiotic stressors due to overlapping spectral characteristics. Multispectral and hyperspectral imaging allow vegetation index calculations, such as normalized difference vegetation index (NDVI), linked to chlorophyll degradation and early stress responses. Thermal sensing detects canopy temperature anomalies from altered transpiration, serving as a potential early indicator of infection. Lidar enables individual tree assessments by detecting defoliation patterns and canopy thinning, essential for tracking disease progression. Integrating these sensors through multi-source data fusion holds strong potential for improving BSNB detection accuracy at both stand and landscape scales.

While correlation analyses and spectral indices aid differentiation, biochemical validation remains essential for improving RS-based classification accuracy (Girma et al., 2005; Fahey et al., 2020). Ground truthing ensures remote sensing data reliability by confirming stressors through field studies and laboratory analyses. Integrating remote sensing with ground truthing improves disease identification and quantifies forest areas affected by BSNB, other needle pathogens (needle cast), and abiotic stressors. A combined RS-based and field approach strengthens BSNB detection and improves understanding of its ecological impact. Despite similarities to Dothistroma Needle Blight, BSNB remains underexplored in RS-based studies. Hyperspectral imaging effectively detects foliar diseases like DNB (Coops et al., 2003) but requires validation to differentiate BSNB from abiotic stressors and co-occurring pathogens. A major limitation in RS-based disease monitoring is the inherent trade-off between spatial and temporal resolution. While high-resolution data improve detection accuracy, operational implementation is often constrained by acquisition costs and revisit frequency, limiting feasibility for large-scale applications (Samadzadegan, Toosi & Dadrass Javan, 2024).

Although Pine Wilt Disease is distinct from needle diseases, remote sensing techniques, including UAV-based multispectral imaging and high-resolution satellite data, have been effectively used for its detection. ML/DL-driven classification models have demonstrated high accuracy in identifying symptomatic trees at various infection stages, supporting early disease management. While PWD affects the vascular system, needle diseases primarily impact foliage, progressing gradually. However, advancements in remote sensing methodologies developed for PWD can be adapted to improve BSNB detection, provided that the differences in disease progression and symptom presentation are carefully considered. ML/DL models trained on spectral signatures and vegetation indices can enhance early symptom identification, while multi-sensor data fusion can strengthen large-scale BSNB monitoring.

Geospatial climate models play a key role in assessing BSNB risk. Temperature, humidity, and precipitation influence disease progression by affecting pathogen sporulation, dispersal, and host susceptibility. Integrating RS-derived disease indicators with climate models can improve risk assessments and support the development of early warning systems. While RS has been applied to monitor various needle diseases, further research is needed to fully integrate it with climate modeling for BSNB risk prediction. Combining UAV and satellite-based RS with climate modeling can provide a more comprehensive understanding of BSNB dynamics. Developing predictive models and risk maps that incorporate both RS and climatic data would enhance early detection and disease management strategies.

Future advancements require optimizing multi-sensor data fusion, improving ML/DL adaptability, and strengthening climate-RS integration. High-temporal-resolution remote sensing data can bridge monitoring gaps, while UAV-based high-resolution imaging remains essential for localized assessments. Collaboration between remote sensing experts, forest pathologists, and climate scientists will refine predictive models and improve large-scale disease monitoring. Despite challenges, continued advancements in sensor technology, data analytics, and computational modeling will support scalable, proactive BSNB detection and management strategies.