Drone-Based Monitoring and Mapping for LMO Confined Field Management under the Ministry of Environment

Abstract

The objective of this study was to devise effective safety management systems for enclosed living modified organism (LMO) fields regulated by the Ministry of Environment (MOE), achieved through an assessment of the impact of LM crops on the surrounding flora. A combination of conventional survey methods and cutting-edge drone-based monitoring systems was employed, with a keen focus on their efficacy. Our investigation spans three distinct zones (forest, non-forest, and enclosed field), involving vegetation surveys, biodiversity index analyses, and drone-powered aerial observations to study topographical shifts. Over time, wild plants adjacent to the enclosed LMO field exhibited stability in terms of species composition. Nevertheless, disparities in growth patterns emerged across various areas. Predominantly, herbs thrived in enclosed and non-forest areas, while trees and shrubs flourished in forested regions. Annual plants predominantly populated the non-forest regions, whereas perennials dominated the forested areas. To this end, drones captured aerial photographs of a 31.65-hectare expanse with 40% coverage overlap, furnishing a real-time vegetation map that transcends the capacities of conventional methods. By combining vegetation surveys, drone-generated vegetation mapping, and dynamic monitoring of topographical changes, our research endeavors to facilitate the formulation of a robust safety management framework for LMO confined fields overseen by the MOE. This holistic approach aspires to prevent ecosystem contamination and establish a resilient, enduring system that averts LMO leakage, thereby safeguarding the environment.

1. Introduction

Maize, a vital staple and the world’s most cultivated grain crop, has witnessed a remarkable transformation, with over 30% now being genetically modified—a Living Modified Organism (LMO). This trend mirrors a global surge in the cultivation of LMO crops, escalating from 1.7 million hectares in 1996 to a staggering 190.4 million hectares by 2019 [1]. As novel biotechnologies continue to emerge, this expansion of LMO crop varieties seems poised to proliferate. However, utilizing LMO crops for consumption, animal feed, or industrial processes necessitates rigorous confined field trials to ascertain their efficacy and safety [2]. Introducing LMOs into ecosystems carries an inherent risk of unintended consequences, such as the inadvertent spread of genetically engineered traits among wild plant populations [3]. Effective safety management strategies for these field trials are imperative to mitigate such environmental risks.

Regulatory frameworks governing LMOs exhibit variation across countries, typically involving dedicated bodies overseeing their development, production, and application [4]. In Korea, seven ministries oversee LMOs based on usage, with the Ministry of Environment (MOE) tasked with safeguarding LMOs used in environmental remediation and the management of affected ecosystems. These regulatory bodies often establish guidelines encompassing LMO confinement in labs and fields, encompassing cultivation, environmental release, and safety assessment [5]. LMO confined field management demands multifaceted actions, including risk evaluation, leakage monitoring, contingency planning, and the formulation of standard operating procedures [6].

Monitoring environmental indicators is a pivotal facet of managing regions frequently planted with LMOs. Specifically, LMOs spread and escape into the surrounding environment [7]. Since LM crops can disseminate through pollen and seeds, meticulously evaluating their impact on neighboring plant communities is imperative [8]. The surveillance of vegetation adjacent to LMO confined fields provides vital insights into the real-world spread and persistence of LMOs [9]. In Korea, confined field experiments mandate investigating the distribution of indigenous or closely related wild species capable of hybridizing with tested LM crops. If they exist within the hybridization range, their removal becomes essential. Additionally, safety measures must be established to preclude transgene transfer or leakage from LM crops within the confined field.

Parallelly, effective safety management of LMO confined fields demands LMO leakage detection and scrutinizing changes within the surrounding flora and vegetation. Utilizing standardized biodiversity indices, like the Shannon–Weaver and Simpson diversity indices, enables a quantitative assessment of LMO impacts on plant biodiversity [10]. Modern times have witnessed the ascendancy of drone-based aerial photography as a potent tool in LMO confined field management [11]. Drone imagery analysis, which surpasses the conventional resolution of satellite images, improves agricultural productivity as well as environmental conservation by supporting crop yield prediction and invasive species detection [12,13,14,15]. In addition, drones also serve as effective tools for animal monitoring and optimization of livestock management systems, resulting in cost reductions [16,17,18,19,20,21]. Therefore, this innovative technology can solve immediate issues such as environmental preservation, marine waste cleaning, or forest fire monitoring [22,23]. Nevertheless, integrating drones into LMO confined field management constitutes a burgeoning research frontier that warrants careful exploration to delineate its prospective benefits and limitations. This study’s objective is to devise robust safety management systems for LMO confined fields, accomplished by evaluating the repercussions of LM crops on neighboring vegetation using conventional survey methods and scrutinizing the potential utility of drone-based monitoring systems.

2. Materials and Methods

2.1. Study Area

The study was conducted at the LMO confined field of the National Institute of Ecology (NIE), Seocheon, South Korea (36°01′43.4″ N, 126°43′21.8″ E), constructed in 2018 for LMO risk assessment purposes. The study area encompasses a radius of 100 m around the LMO confined field, which occupies an area of 9873 m² and is comprised of cutting slopes, test fields, and collection wells. The LMO confined field is irregular and bordered by buildings, waterways, playgrounds, and forested areas. The LMO confined field was designed and currently operated to assess and test risks associated with releasing LMO plants into natural environments. However, no LMO plants have been released, and the natural environment has been partitioned into observation fields and artificially managed areas.

2.2. Vegetation Survey

To monitor native plants around the LMO confined field, the survey site was divided into three areas—forest, non-forest, and confined fields—centered on the field boundary. All plant species growing within 100 m of the field were recorded according to the season and year. For the vegetation survey, 27 plots were established at 0, 25, 50, and 100 m from the field. The plot sizes were 5 m × 5 m in the forest, 1 m × 1 m in the non-forest area, and 2 m × 2 m in the confined field. The survey was conducted in three temporal zones: spring (May 2020, April 2021, and May 2022); summer (August 2020, 2021, and 2022); and autumn (October 2020, 2021, and 2022). All surveyed species were identified and classified using the National Species Database of Korea [24] and other botanical illustrations. The importance values of the species in each survey area were calculated by converting the dominance class to the median of the dominance range [25] and using the average relative frequency [26].

2.3. Diversity Statistics and Data Analysis

Plant biodiversity and dominance in each survey area were analyzed (by season and year) using the Shannon–Weaver diversity index [27], Simpson diversity index [28], Margalef species richness index [29], and Pielou’s species evenness index [30].

Shannon–Weaver diversity index (H′) $=-\sum \left(P_i\times \log \left(P_i\right)\right)$;

Simpson diversity index (1-Lamda′) $=1-\sum (N_i\times \left(N_i-1\right)÷\left(N\times \left(N-1\right)\right)$;

Species evenness index (J’: evenness) $=H^{\prime }÷\log \left(s\right)$;

Species richness index (d: richness) $=\left(S-1\right)÷\log \left(N\right)$;

P_i: Proportional abundance of species I;

N_i: The number of species I;

N: Density of species;

S: Total number of species.

All analyses were performed using the SAS Studio (version 3.8; SAS Institute Inc., Cary, NC, USA). The data were subjected to analysis of variance (ANOVA) at a significance level of 5%. If the ANOVA results indicated significant differences between means, Tukey’s honest significant difference (HSD) test was used to determine specific differences between means. A General Linear Model (GLM) was used to determine the effects of the year, location, and season on plant diversity.

2.4. Image Acquisition Using Aerial Drone

Drone-based aerial photography was used to monitor the facilities in the LMO field and to observe changes in internal (experimental) and external (wild) plants. A DJI INSPIRE 2 unmanned aerial vehicle (UAV) (DJI, Guangdong, China) and a ZENMUSE X4S system were used to acquire aerial imagery from the field. Aerial images were acquired between 13:00 and 15:00, and between 2020 and 2022. The frequency of flights was approximately once per month between March and November 2020 and 2022, depending on weather conditions. Images were only acquired when the weather was cloudy or sunny. UAV flights were performed 150 m above the ground, and a camera lens (AF-S 8.8 mm; aperture size, f/5.6; focal length, 9 mm; diaphragm, 2.97; ISO, 100; exposure time, 1/400 s) was maintained perpendicular to the ground. More than 470 images (image dimensions: 5472 × 3078 pixels) were acquired and used in this study.

2.5. Data Processing

Aerial images captured by drones were used to produce an accurate vegetation map. Images were processed and edited using photogrammetry software such as PhotoScan Professional (Agisoft LLC, St. Petersburg, Russia). Because each drone image-capturing point has GPS coordinates and captures overlapping images, a single image can be generated by editing. Additionally, an orthomosaic image can be created to produce a digital elevation model (DEM) and contour lines, which can be used to determine the dividing lines of the actual terrain. These processed images were then used to create the vegetation maps using QGIS software (QGIS Development Team, version 3.22.9-Białowieża, Poland).

3. Results and Discussion

3.1. Exploring Spatiotemporal Vegetation Dynamics Surrounding LMO Confined Fields: Insights from Comprehensive Wild Plant Surveys

According to the Transboundary Movement, Etc. of LMOs Act (henceforth “the LMOs Act”), seven government ministries have managed LMOs in South Korea. The MOE performs the risk review and safety management of LMOs for environmental remediation and consultative review on the risk posed to the natural ecosystem by LMOs for other purposes. Since 2018, the MOE has been operating a confined field for LMO research at the National Institute of Ecology (NIE), an institute of LMO risk assessment to promote scientific and systematic safety management by standardizing the evaluation criteria for LMOs under the jurisdiction of the MOE.

The pivotal aim of this study was to forge innovative safety management systems tailored for LMO confined fields under the regulatory purview of the MOE. This objective was realized through meticulously evaluating the interplay between LM crops and the surrounding plant ecosystem. A hybrid approach, fusing traditional survey techniques with cutting-edge drone-based monitoring systems, was judiciously employed, casting a critical spotlight on their efficacy (Figure 1). Our rigorous exploration spanned a comprehensive three-year period, traversing three distinct ecological areas: the encompassing forest, the adjacent non-forest expanse, and the core confined field. This endeavor encompassed three essential activities, including intensive vegetation surveys, rigorous analyses of biodiversity indexes, and the harnessing of drone-enabled aerial observations to dissect the intricate shifts in topography.

Before initiating the environmental release assessment for LMOs, a comprehensive survey was conducted on the communities of wild plants surrounding the LMO confined field, and the spatiotemporal dynamics of vegetation were carefully analyzed.

Table 1. Changes in plant species inside and outside the confined field, 2020 to 2022.

Location	Year	Plant Growth Form (%)							Life Form (%)				Origin (%)
		Herb	Shrub /Tree	Tree	Shrub	Herb /Shrub	Graminoid	Fern	Annual	Annual /Biennial	Biennial	Perennial	Native Plant	Non -Native Plant
Confined field	2020	50.0	0	0	4.5	0	27.3	4.5	52.4	19.0	0	28.6	76.2	23.8
	2021	52.4	0	4.8	4.8	4.8	19.0	0	40.0	15.0	5.0	40.0	70.0	30.0
	2022	41.7	0	8.3	8.3	8.3	8.3	0	45.5	9.1	0	45.5	63.6	36.4
Forest 25 m	2020	35.9	1.3	6.4	11.5	1.3	10.3	7.7	23.4	3.9	1.3	71.4	85.7	14.3
	2021	22.1	0	8.8	14.7	0	7.4	1.5	7.6	12.1	0	80.3	97.0	3.0
	2022	37.7	0	3.8	13.2	0	11.3	7.5	30.8	7.7	0	61.5	88.5	11.5
Forest 50 m	2020	25.6	0	7.0	11.6	2.3	16.3	11.6	20.4	5.6	1.9	72.2	90.7	9.3
	2021	22.0	0	10.0	16.0	2.0	16.0	8.0	21.4	4.8	2.4	71.4	92.9	7.1
	2022	22.0	0	10.0	16.0	2.0	16.0	8.0	16.3	4.1	0	79.6	93.9	6.1
Forest 100 m	2020	36.1	0	3.3	9.8	3.3	18.0	8.2	28.8	5.1	1.7	64.4	84.7	15.3
	2021	36.5	0	3.2	14.3	3.2	12.7	3.2	33.9	3.2	3.2	59.7	88.7	11.3
	2022	46.9	0	3.1	10.9	1.6	14.1	3.1	30.2	4.8	4.8	60.3	84.1	15.9
Non-Forest 25 m	2020	63.4	0	0	0	0	24.4	2.4	47.5	7.5	7.5	37.5	62.5	37.5
	2021	67.7	0	0	0	3.2	12.9	3.2	33.3	3.3	13.3	50.0	73.3	26.7
	2022	63.4	0	2.4	2.4	2.4	19.5	2.4	50.0	2.5	7.5	40.0	62.5	37.5
Non-Forest 50 m	2020	62.5	0	0	3.1	0	12.5	3.1	38.7	3.2	6.5	51.6	61.3	38.7
	2021	66.7	0	0	4.8	0	14.3	4.8	45.0	5.0	5.0	45.0	70.0	30.0
	2022	63.0	0	3.7	0	3.7	18.5	3.7	34.6	3.8	7.7	53.8	61.5	38.5
Non-Forest 100 m	2020	47.4	0	2.6	0	0	31.6	0	45.9	2.7	8.1	43.2	62.2	37.8
	2021	62.5	0	0	0	0	20.8	0	39.1	0	4.3	56.5	65.2	34.8
	2022	51.5	0	3.0	0	0	21.2	3.0	37.5	3.1	3.1	56.3	65.6	34.4

presents a representative list depicting the spring flora in the herbaceous strata of the study area encircling the LMO confined field from 2020 to 2022 (Figure 2). Notably, the plant species within the surveyed confined field exhibited consistent classification ratios across the years. Nonetheless, divergences in plant growth forms were evident among distinct locations. Specifically, herbs constituted 41.7% to 67.7% of observed plants in the confined field and non-forest areas, whereas, within the forest area, this range extended from 22.0% to 46.9%. Further, the forest area encompassed a tree proportion ranging from 3.1% to 16.0%, while confined fields and non-forest domains had a lower range of 0.0% to 3.7%. Intriguingly, the forest located at a 50 m distance from the field exhibited the highest tree proportion among all locations. In contrast, the non-forest region at a 100 m distance boasted the most incredible abundance of graminoids. Noteworthy disparities were observed in the distribution of life forms, where annual plants predominated in confined fields and perennials held sway in forested expanses. The non-forest regions contained an annual plant proportion ranging from 33.3% to 50.0%, whereas, in the forest areas, this ratio ranged from 7.6% to 33.9%. Conversely, the forest domain displayed a perennial proportion range of 59.7% to 80.3%, while the non-forest area spanned 37.5% to 56.5%. The proportion of native plants spanned from 61.3% to 97.0%, whereas foreign plants accounted for 3.0% to 38.7%. These results underscore the disparities in plant species composition between forested and non-forest locales, attributable to variances in habitat features like light availability, soil nutrients, and moisture content [31]. The forested regions harbored species like tree seedlings, shade-tolerant herbs, and understory shrubs, unlike the non-forest sectors, which featured abundant light-demanding herbs and grasses [32].

3.2. Unveiling Dynamic Species Diversity: Exploring Variations across Locations, Years, and Seasons

Significant variations in species diversity indicators were detected across different locations, with year and season exerting varying influences (

Table 2. Results of the general linear model (GLM) applied to the three years of diversity index values from the interior and exterior of the living modified organisms (LMO) confined field.

Source	DF	Richness Index		Evenness Index		Shannon Diversity Index		Simpson Diversity Index
		F	p	F	p	F	p	F	p
Year	2	0.95	0.393	11.71	<0.001	0.77	0.466	0.20	0.818
Location	2	160.89	<0.001	330.15	<0.001	254.93	<0.001	415.33	<0.001
Season	2	1.59	0.213	5.72	0.006	0.66	0.523	1.17	0.318
Year × Location	4	1.31	0.279	4.33	0.004	2.12	0.091	2.69	0.041
Year × Season	4	0.93	0.456	6.01	<0.001	1.06	0.385	2.92	0.029
Location × Season	4	2.07	0.097	2.87	0.032	2.97	0.027	5.26	0.001
Year × Location × Season	8	0.73	0.666	5.86	<0.001	1.43	0.207	4.35	<0.001

, Figure 3). The richness index exhibited substantial divergence by location (p < 0.001), as did the Shannon diversity index (p < 0.001); an interaction between location and season also surfaced (p = 0.027). Additionally, the Simpson diversity index reflected significant differences based on location (p < 0.001) and across all interactions. Notably, the evenness index displayed noteworthy disparities across all factors and interactions. Findings from the vegetation survey delineated a richer diversity of plant species within forested domains compared to non-forest realms. This observation is consistent with a prior study that found that forests harbor higher plant diversity owing to their intricate habitat structure and microclimate [33]. The Shannon–Weaver, Simpson, and Margalef species richness indices demonstrated elevated values within the forested regions, underscoring a more diverse and evenly distributed plant community [10].

3.3. Discerning Variations in Plant Importance across Seasons, Distances, and Locations: Implications for LMO Confined Field Management

Notable alterations in the importance values of plants were noted across seasons and distances, with variations depending on specific locations. The confined field exhibited distinct attributes compared to other areas, and distinctions between forested and non-forest zones were also discernible (Figure 4). Forest regions were categorized based on their proximity to the LMO confined field, revealing variations in vegetation relative to distance. Conversely, non-forest regions exhibited pronounced seasonal shifts, particularly in spring 2020, autumn 2021, and spring 2022. These findings suggest consistent segregation in the importance values of vegetation-associated spatial distribution encompassing LMO confined fields over time, heightening the ability to detect unintentional LMO release promptly. In tandem, Ni et al. [34] conducted a study investigating the spatial extent and impact severity of successful invasive species by assessing vegetation criticality. Furthermore, this study illuminates that seasonal variability in non-forest regions (waterways, playgrounds, and buildings) is more pronounced than in forested regions, attributed mainly to human interventions and artificial management practices. This underscores the significance of distinguishing between human disturbances and natural habitats in LMO confined field monitoring.

3.4. Aerial Reconnaissance for Precise Data Acquisition: Capturing Imagery at 150 m Altitude over 31.65 Hectares Using Drones

During drone flights, images were captured from an altitude of 150 m, spanning a total area of 31.65 hectares, with horizontal and vertical overlaps configured at 40%, as shown in Figure 5a. Within a 24 min flight time, 42 aerial photographs were procured. Subsequently, these images underwent approximately one hour of editing. Although most images displayed overlap with at least three aerial counterparts, a few edges exhibited less coverage. Analysis of the aerial images yielded a pixel error of 0.344 cm, facilitating the generation of digital elevation models (DEMs) and ortho-images (Figure 5b). The root means square error (RMSE) for the drone image was calculated to be 0.36, 0.67, and 2.03 m in the X, Y, and Z directions, respectively. This error introduced minor positional discrepancies in the edited images, primarily noticeable in buildings and identifiable terrain. QGIS was employed to divide the topography into polygons using contour lines extracted from drone ortho-images (Figure 5c,d).

3.5. Elevating Vegetation Monitoring with Drones: Real-Time Insights and Species Dominance Mapping

Vegetation maps conventionally relied on satellite map services or aerial images furnished by the National Geographic Information Institute. However, these approaches could not capture real-time vegetation changes. In contrast, drones offer a holistic view of actual vegetation shifts through their real-time, high-resolution image capture. Drones unveiled the evolution of buildings around the LMO confined field and topographical transformations due to precipitation and seasonal vegetation variations. In conclusion, using drones to capture high-resolution aerial images represents a valuable technique to monitor vegetation changes [12,35]. In our study, drones provided real-time images that allowed for a comprehensive view of the actual vegetation dynamics, including the construction process of buildings, topographical changes due to rainfall, and seasonal variations in vegetation. The use of drones holds promise for various applications, including land development, forest management, and plant sociological spatial analysis [36]. Furthermore, Kedia et al. [37] reported that drones have the capability to discern invasive plant species in arid regions that are susceptible to wildfires and floods.

In May 2022, vegetation shifts were recorded via drone images, culminating in creating of a vegetation map (Figure 6). The map elucidates variations in vegetation coverage by pinpointing dominant species within observed areas. It leverages the relative dominance of vegetation cover to discern primary species in the monitored regions. Notably, Conyza canadensis reigned supreme as the dominant species within the LMO confined field, while graminoid plants such as Oplismenus undulatifolius dominated the 25 m forest area. Shrubbery predominated the 50- and 100 m zones, with Akebia quinata claiming dominance. Non-forest domains saw a shift in dominant species, with Taraxacum officinale taking the lead in the 25 m region, Zoysia japonica encircling buildings, and Miscanthus sacchariflorus lining waterways at 50 m. In the 100 m expanse, Zoysia japonica, meticulously managed due to its proximity to structures, emerged as the predominant species, whereas Vicia sativa dominated the unattended areas. The vegetation importance value was harnessed to construct the vegetation map, spotlighting Conyza canadensis as the most significant species within the LMO confined field. Lamium amplexicaule held sway in the 25 m forest region, while Equisetum arvense commanded dominance at 50 and 100 m. Taraxacum officinale claimed the highest importance value in non-forest zones at 25 m, while Zoysia japonica secured the spotlight at 50 and 100 m. The outcomes unveiled that alterations in vegetation were steered by distance and human activity. Nevertheless, variations in dominant species were intricately linked to topography in the unmanaged sectors. These revelations underscore the essentiality of meticulous observation concerning shifts in unattended forest areas, especially following the release of LMOs into the ecosystem. Detecting LMOs can be formidable, fraught with diverse ecological risks [3]. In contrast, to comprehend the impact of LMO leakage on the ecosystem surrounding the confined field, it is imperative to accumulate spatial change data on a seasonal and yearly basis. In addition to the data obtained through temporary drone surveys, it is essential to conduct continuous data accumulation over time to analyze the surrounding changes, analogous to vegetation surveys [38]. Our findings specifically demonstrated the potential of using a Geographic Information System (GIS) database to establish, analyze, and manage comprehensive spatial and vegetation information derived from drone imagery, thereby enabling effective safety management of LMO confined fields. Nonetheless, it is important to acknowledge the limitations of this technology. Unlike general satellite imaging analysis of vegetation changes, the operation of drones in adverse weather conditions such as strong winds and heavy rains presents challenges [39,40]. Consequently, suitable weather conditions for drone flights are required, and with further advancements, restrictions imposed by climate conditions must be minimized. Future research endeavors should focus on discussing the management of spatial and vegetation information for the establishment of a GIS database to facilitate systematic data management. To achieve precise and extensive management, investigations should involve a wide range of drone photography and images encompassing various spectra.

4. Conclusions

Establishing a robust safety management system for LMO confined fields is paramount to avert the contamination of neighboring ecosystems. Particularly noteworthy is the integration of drones, which has revolutionized the production of vegetation maps, enabling nuanced monitoring of vegetation dynamics around LMO confined fields, surpassing the capabilities of conventional surveys. This study utilized a drone to capture high-resolution images and detailed information compared to satellite imagery. Another advantage of drone imagery is the segmented shooting intervals, enabling the prompt reflection of terrain changes around the LMO confined field and enhancing the mapping precision of the vegetation community. Furthermore, the aerial imagery process using drones helps to obtain comprehensive data, including areas previously inaccessible on target vegetation with enlarged survey coverage, swift investigation speed, and reduced workforce and expenses. This spatial data acquisition lays the groundwork for constructing a comprehensive database, a cornerstone for forging a sustainable, long-term approach to managing LMO leakage. This study’s focal point centered on monitoring shifts within the safety management system, pinpointing the cultivation of LMOs within established confined fields, and encompassing pertinent species and corresponding flora. By orchestrating vegetation surveys in peripheral regions and harnessing drones for instantaneous topographical and spatial data collection, the potential for the newly suggested system was demonstrated. The developed monitoring system could be utilized safely and conveniently for the currently set up LMO confined fields in South Korea. Furthermore, it would evolve into a standard safety management system of the MOE’s confined fields to be built in the future. In the forthcoming period, we intend to enhance the safety management framework for cultivating LMOs by employing drone technology to investigate plant clusters and leveraging artificial-intelligence-based deep learning techniques for the automated monitoring of vegetation alterations via image processing. In addition, we expect that the rapid and cumulative data acquisition could draw a preemptive strategy against the unintended release of LMOs into nature on vegetation monitoring, terrain change, etc. This will entail comparing drone-derived observations with actual vegetation survey results, fostering a deeper understanding of the system’s practical implementation.