Analysis of PlanetScope Imagery-Based NDVI Changes for Monitoring Brown Planthopper Damage

Abstract

In September 2024, Imsil County, Jeollabuk-do experienced a large-scale outbreak of brown planthoppers due to abnormally high temperatures and prolonged heatwaves,with damage reported across more than 1,900 hectares. Brown planthoppers cause rapid growth deterioration and stem lodging through sucking damage, making it crucial to identify the timing and spatial distribution of damage occurrence quickly. This study conducted a retrospective analysis of brown planthopper damage using PlanetScope SuperDove imagery with 3-5 m spatial resolution. The study area encompassed rice cultivation fields in the Osu-myeon area of Imsil County, utilizing images from September 5, 10, 15, and 29, 2024.We examined time-series Normalized Difference Vegetation Index (NDVI) changes and calculated NDVI differences (ΔNDVI) between two time periods to quantify damage severity. Results showed that damaged fields exhibited a sharp NDVI decline after mid-September, clearly distinguishing them from the gradual decline pattern observed in normal areas. ΔNDVI threshold intervals were established based on drone RGB imagery captured on September 30, categorizing areas as: harvested (≤-0.27), high probability of damage (-0.27 to -0.10), minor damage or senescence (-0.10 to -0.05), and normal growth (>-0.05). These classification results aligned well with drone image interpretation, effectively reflecting the damage status. While ΔNDVI reflects complex factors, including not only brown planthopper damage but also stem lodging, making it difficult to attribute causes definitively, it serves as a useful signal for macroscopic damage detection. While ΔNDVI reflects complex factors, including not only brown planthopper damage but also stem lodging, making it difficult to attribute causes definitively, it serves as a useful signal for macroscopic damage detection. When combined with field surveys, it holds significance as an auxiliary indicator for detecting damage hotspots and improving survey efficiency. This supports the applicability of high-resolution satellite imagery for agricultural damage monitoring and is expected to provide foundational data for improving agricultural disaster response systems in the future.

1. Introduction

Recent extreme weather events, such as high temperatures, heat waves, and concentrated rainfall due to climate change, are accelerating the occurrence of pests and diseases in rice cultivation areas (Hu et al., 2015). Particularly in September 2024, the Jeollabuk-do region experienced unprecedented high temperatures around the Chuseok holiday period, leading to rapid brown planthopper reproduction and large-scale damage reports (Ministry of Agriculture, Food and Rural Affairs, 2024). Brown planthoppers are sucking pests that feed on rice sap, and when damage spreads, entire rice fields can die within a short period, with severe cases leading to stem lodging (Chaudhary et al., 2014; Xue et al., 2014). Such damage was confirmed in Osu-myeon, Imsil County, where repeated pesticide applications by farmers failed to prevent the spread (Cheonji-Ilbo, 2024). These characteristics make it difficult for farmers to respond through individual efforts alone, necessitating a scientific monitoring system capable of rapidly diagnosing damage occurrence and spatial distribution.

Since pests like brown planthoppers can rapidly expand their damage area, post-analysis comparing conditions before and after damage occurrence is important in addition to early detection. Traditional field surveys can ensure precision but have limitations in comprehensively assessing damage across thousands of hectares in spatial and temporal terms. Drone imagery provides high resolution at the tens of centimeters level, making it effective for detailed observation of damaged areas, but observation coverage is limited. In contrast, satellite imagery can repeatedly capture large areas, making it advantageous for tracking damage spread patterns over time series.

In vegetation remote sensing research, the Normalized Difference Vegetation Index (NDVI) is primarily used to observe crop growth, and for damage analysis, the method of estimating damage severity using NDVI differences (ΔNDVI) between two time periods has been widely employed (Pantaleoni et al., 2007; Ha et al., 2022; Qamer et al., 2023). In Korea, studies have been conducted using drone imagery to estimate NDVI in rice disease occurrence areas and comparing results with Sentinel-2 (10 m) and Landsat-8 (30 m) imagery (Ryu et al., 2022). However, such studies have mainly focused on diseases such as rice blast, and cases of analyzing pest damage like brown planthoppers using satellite imagery are extremely rare. Moreover, since brown planthopper damage appears sporadically and unevenly, medium to low-resolution satellites face constraints in precisely identifying damage at individual field levels.

Therefore, the objective of this study is to analyze NDVI changes before and after the occurrence of brown planthopper damage using PlanetScope satellite imagery with approximately 3–5 m spatial resolution, and to characterize damage distribution patterns based on ΔNDVI quantitatively. Through this approach, we aim to exploratively present the potential applicability of high-resolution satellites in brown planthopper damage detection. To achieve this, we evaluated changes before and after damage using PlanetScope NDVI time-series data, and classified fields by categorizing damage severity based on ΔNDVI values. Additionally, we assessed the satellite-based detection results by referencing drone imagery.

2. Materials and Methods

2.1. Study Area

The brown planthopper damage area in Jeollabuk-do, reported by September 19, 2024, totaled approximately 2,700 ha, with Imsil County accounting for about 1,900 ha (70% of the total) and recording the largest damage (MBC News, 2024). Imsil County serves as a major rice cultivation area within Jeollabuk-do and has regional representativeness in terms of damage scale. According to the Korea Meteorological Administration’s Open MET Data Portal (https://data.kma.go.kr/cmmn/main.do), the average temperature in Imsil County in August 2024 was approximately 27°C, and the September average temperature was approximately 24.5°C. Temperatures generally show a gradual decline during September (Fig. 1). However, 2024 was exceptional, with high temperatures similar to August continuing until mid-September, followed by a declining trend from mid-month onward (Fig. 1). The total number of heat wave days recorded during August and September of the same year was 19 days, corresponding to the top levels of the past 10 years (Rank 3), with September heat waves being unprecedented (Fig. 2). These abnormal weather conditions are interpreted as contributing to the increase in brown planthopper populations and damage spread.

Fig. 1. Daily mean, minimum, and maximum temperatures in Imsil (Aug–Sep 2024) compared with 1991–2020 climate normals.

Fig. 2. Monthly distribution of heat wave days in Imsil County from 2014–2024.

In this study, we selected rice cultivation fields in the Yongdu-ri area of Osu-myeon, Imsil County as the analysis target (Fig. 3). This area experienced concentrated brown planthopper damage in September 2024, and with damaged and undamaged fields intermixed, it represents an appropriate region for comparative analysis of NDVI changes before and after damage occurrence.

Fig. 3. Study area. The background map is a drone RGB image acquired on September 30, 2024, and the red boundary indicates paddy rice areas derived from Farmmap.

2.2. Data

Drone photography was conducted during field surveys on September 30, 2024, using a Mavic 3 Multispectral (M3M) platform. The flight altitude was approximately 146 m, and images were collected in blue (450 nm), green (560 nm), and red (650 nm) wavelength bands with a spatial resolution of approximately 3 cm. The collected images were orthorectified and geometrically corrected using Pix4Dmapper. In this study, RGB images captured by a drone were used to conduct a visual interpretation of damaged areas and distinguish between damaged and undamaged fields. These classified fields were used as reference data for satellite NDVI change analysis.

PlanetScope SuperDove satellite imagery provides approximately 3–5 m spatial resolution. For before-and-after damage comparison, this study utilized images from September 9, 17, 21, and 27, 2022, and September 5, 10, 15, and 29, 2024. September 2023 images were excluded from the analysis as no usable images were available due to cloud cover. The SuperDove satellite provides eight bands (Coastal Blue, Blue, Green I, Green, Yellow, Red, Red-Edge, Near-Infrared [NIR]), and this study used the Red and NIR bands for NDVI calculation (Eq. 1). Clipping was performed to match the study area boundaries, followed by NDVI calculation to analyze changes before and after the occurrence of brown planthopper damage.

\(\begin{align}NDVI = \frac{N I R - R e d}{N I R + R e d} \end{align}\)       (1)

The NDVI is defined as follows: NDVI is a representative index reflecting vegetation health, with values ranging from −1 to 1. Generally, values of 0.2 or below indicate non-vegetated areas, 0.2–0.5 represents poor growth or sparse vegetation, and values above 0.5 indicate active growth conditions. Therefore, decreases in NDVI values are closely related to crop growth decline, pest and disease damage, plant death, and other factors.

2.3. Methods

In this study, damaged fields and control groups were established by linking drone imagery with PlanetScope NDVI. Through drone RGB image interpretation, fields showing signs of damage were classified as damaged areas, undamaged areas within the same region as normal areas, and harvested fields as harvested areas. This classification was used as a basis for satellite NDVI time-series analysis and ΔNDVI-based change detection.

Time-series NDVI derived from PlanetScope SuperDove satellite imagery was used to analyze changes before and after damage occurrence. September 2022 images (9th, 17th, 21st, 27th) were used as normal-growth comparison groups, while September 2024 images (5th, 10th, 15th, 29th) were used for analysis of the damage-occurrence period. Time-series NDVI was calculated as field-level averages to compare changes across damaged, normal, and harvested areas.

NDVI change (ΔNDVI) between two time periods was calculated to assess vegetation vitality decline (Eq. 2).

ΔNDVI = NDVI_after - NDVI_before       (2)

Only the 2024 images were utilized for ΔNDVI analysis, and three time pairs were established reflecting the temporal characteristics around Chuseok (September 17, 2024): September 5–10, September 10–15, and September 15–29.

3. Results

3.1. NDVI Time-Series Analysis Results

Based on NDVI time-series analysis results using PlanetScope satellite imagery (Fig. 4), damaged areas in 2024 began showing a rapid NDVI decline after September 10, with a continuous decrease until September 29. This corresponds to leaf withering and plant death, as confirmed in drone RGB imagery of damaged areas.

Fig. 4. Drone RGB images acquired in 2024 (a–c) and PlanetScope-based NDVI time series for 2022 (normalyear without brown planthopper [BPH] outbreak) and 2024 (year with BPH outbreak) (d). The drone RGB images (a–c) depict the field conditions of a damaged area (a), a normal area (b), and a harvested area (c), while (d) presents the NDVI time series extracted from the corresponding areas using PlanetScope imagery.

Harvested areas in 2024 showed a very rapid decline after September 10, displaying a steeper decline curve than damaged areas. This is interpreted as a result reflecting vegetation removal due to actual harvesting operations. In contrast, normal areas in 2024 showed a gradual NDVI decline tendency during the same period. This reflects physiological changes following the grain-filling process as rice growth progressed normally.

In the 2022 comparison group, since no damage occurred, NDVI declined gradually throughout September, showing a distinct difference from the rapid decline patterns observed in the 2024 damaged and harvested areas.

3.2. NDVI Change (ΔNDVI) Analysis Results

By calculating NDVI changes (ΔNDVI) between two time periods, we were able to spatially distinguish damage occurrence and damage intensity. Damage status was confirmed based on drone RGB imagery captured on September 30, 2024, and using this as a reference, ΔNDVI threshold intervals were established in four levels: ≤−0.27, −0.27 to −0.10, −0.10 to −0.05, and >−0.05. These threshold intervals were empirically derived by comparing and analyzing the damage levels observed in drone imagery with the distribution of ΔNDVI values. The primary analysis pair was September 15–29, closest to the drone photography date, while September 5–10 and September 10–15 were used as auxiliary analysis pairs to confirm the occurrence and spread of damage.

At Site A (Fig. 5), severe damage confirmed in drone imagery was also clearly evident in September 15–29 ΔNDVI, while the preceding two time pairs showed the initial damage occurrence and spread processes. These continuous changes clearly revealed differences between damaged and normal areas, suggesting that ΔNDVI can capture the progression of brown planthopper damage in time series.

Fig. 5. Satellite-based ΔNDVI for Site A. (a) ΔNDVI between September 5 and 10, (b) ΔNDVI between September 10 and 15, and (c) ΔNDVI between September 15 and 29. The background map is a drone RGB image acquired on September 30, 2024, and the red boundary indicates paddy rice areas derived from Farmmap. ΔNDVI values represent satellite-based changes in NDVI, and the color scheme denotes relative levels of potential damage.

Site B (Fig. 6) is an area with mixed damaged and harvested fields, where damage hotspots were identified in September 15–29 ΔNDVI, consistent with the damage distribution observed in drone imagery. In the September 10–15 analysis, some fields appearing to be harvested areas showed values ≤−0.27, confirmed as areas where vegetation was removed and NDVI declined rapidly.

Fig. 6. Satellite-based ΔNDVI for Site B. (a) ΔNDVI between September 5 and 10, (b) ΔNDVI between September 10 and 15, and (c) ΔNDVI between September 15 and 29. The background map is a drone RGB image acquired on September 30, 2024, and the red boundary indicates paddy rice areas derived from Farmmap. ΔNDVI values represent satellite-based changes in NDVI, and the color scheme denotes relative levels of potential damage.

Site C (Fig. 7) also had mixed damaged and harvested fields, with damage hotspots clearly appearing in September 15–29 ΔNDVI. No distinct damage was observed in the auxiliary analysis pairs of September 5–10 and September 10–15, indicating that substantial damage occurred after September 15 at Site C.

Fig. 7. Satellite-based ΔNDVI for Site C. (a) ΔNDVI between September 5 and 10, (b) ΔNDVI between September 10 and 15, and (c) ΔNDVI between September 15 and 29. The background map is a drone RGB image acquired on September 30, 2024, and the red boundary indicates paddy rice areas derived from Farmmap. ΔNDVI values represent satellite-based changes in NDVI, and the color scheme denotes relative levels of potential damage.

In summary, the ΔNDVI ≤−0.27 interval mainly corresponds to harvested areas, while the −0.27 to −0.10 interval is interpreted as reflecting growth decline due to brown planthopper damage or stem lodging. The −0.10 to −0.05 interval was classified as minor damage or physiological senescence, and the >−0.05 interval as normal growth. Applying these intervals resulted in a clear distinction between damaged and normal fields, with fields appearing as hotspots, also consistent with field surveys. However, since the ΔNDVI threshold intervals in this study are empirically set based on specific time periods and regions, values may differ under other conditions. Therefore, ΔNDVI can be utilized as an auxiliary indicator for assessing relative damage intensity in conjunction with field surveys and supplementary data.

4. Discussion and Conclusions

This study analyzed brown planthopper damage that occurred in Imsil County in 2024 by integrating PlanetScope SuperDove satellite imagery with drone imagery. The NDVI time-series analysis revealed that damaged fields showed a sharp decline after mid-September, which was distinctly differentiated from the change patterns of normal growing areas and harvested fields.

ΔNDVI analysis enabled spatial differentiation of damage occurrence and damage intensity. The ΔNDVI threshold intervals (−0.27 or below, −0.27 to −0.10, −0.10 to −0.05, above −0.05) established with reference to drone RGB imagery were interpreted as harvested areas, severe growth decline, minor damage or physiological senescence, and normal growth, respectively. Analysis results from Sites A, B, and C showed that damage hotspots identified in the September 15–29 ΔNDVI exhibited high concordance with actual damage distribution observed in drone imagery.

Particularly noteworthy is the pattern observed in the lower-left field of Site C in Fig. 7(c). While this field was identified as normal in the drone imagery, ΔNDVI decreased uniformly across the entire field in the range of −0.10 to −0.05. This contrasts with the non-homogeneous decrease pattern observed in brown planthopper-damaged areas and is interpreted as natural decline due to physiological senescence. This difference suggests the possibility of distinguishing damage causes through ΔNDVI pattern analysis.

However, ΔNDVI analysis can reflect various growth-inhibiting factors, including brown planthopper damage, lodging, and other diseases such as rice blast, making it difficult to definitively determine the cause of damage based on values alone. Additionally, the ΔNDVI threshold intervals established in this study are empirical settings based on the specific temporal and environmental conditions of Imsil in 2024, and may vary when applied to different regions or time periods. Therefore, future research should accumulate cases from various regions and time periods to establish a more generalized threshold system, or develop automated detection algorithms based on machine learning.

Furthermore, while this study applied the widely used NDVI as a pioneering attempt at satellite-based brown planthopper damage detection, detection accuracy could be improved by examining various vegetation indices utilizing the Red-edge and Green bands provided by PlanetScope imagery (e.g., GNDVI, Red-edge NDVI). The development of deep learning-based detection models combining multi-spectral information and temporal patterns will enable rapid and objective damage assessment across large-scale areas.

Nevertheless, ΔNDVI is meaningful as a signal for detecting damage occurrence at a macroscopic level, enabling priority identification of areas with high damage potential. Therefore, ΔNDVI can be utilized as a supplementary indicator in conjunction with field surveys and auxiliary data to evaluate relative damage intensity, narrow the investigation scope, and enhance efficiency in damage hotspot detection and area estimation.

In conclusion, PlanetScope satellite imagery demonstrates significant potential as a means to macroscopically detect spatial characteristics of agricultural damage and complement field investigations, and can be utilized as foundational data for future agricultural damage monitoring research. The utility of high-resolution satellite imagery has been particularly confirmed for early detection and damage extent estimation of sudden and widespread pest damage such as brown planthopper infestations.

Author Contributions

Conceptualization: Lee S.-J.; Data curation: Lee S.-J., Ryu JH, Lee KD; Methodology, Formal analysis, Validation: Lee S.-J.; Project administration: Lee S.-J.; Funding acquisition, Supervision: Lee KD; Writing–original draft: Lee S.-J.; Writing–review & editing: All authors.

Conflicts of Interest

No potential conflict of interest relevant to this article was reported.

Funding

This research was funded by the Rural Development Administration (Grant No. PJ017651).

Data Availability Statement

The raw drone imagery and satellite data used in this study contain sensitive information about farm locations and cannot be publicly shared. The processed/analyzed data that support the findings of this study are available from the corresponding author upon reasonable request.