SORA 2.5-Guided BVLOS UAS for Wildlife Conservation in Kenya: Reducing Friction Between Safety and Field Operations

Highlights

What are the main findings?

• In conservation environments, ground risk is largely driven by transient populations (e.g., tourism and ranger activity), which are not reliably captured by standard population raster datasets.
• Formal safety requirements and field realities are often misaligned in wildlife operations, creating operational friction between prescribed procedures and practical execution.

What are the implications of the main findings?

• Ground-risk assessments in conservancies should prioritise bottom-up, activity-based population estimates, using raster products primarily for triangulation and sensitivity analysis.
• Reducing friction between safety requirements and field needs, through iterative crew optimisation and mission-specific procedure design, can improve both operational efficiency and procedural compliance.

Abstract

Safe Beyond Visual Line of Sight (BVLOS) operations are increasingly required for wildlife monitoring and conservation, yet existing regulatory frameworks are rarely tailored to protected areas characterised by low population density and limited infrastructure. This paper presents a field-based use case illustrating how the Specific Operations Risk Assessment (SORA) methodology can be applied to conservation-oriented BVLOS missions under Kenyan airspace conditions, including coordination within military-controlled airspace. We evaluate three population-density estimation approaches (qualitative, bottom-up, and top-down) against available ground truth, and compare tabulated and analytical SORA methods for deriving the Ground Risk Class. The work illustrates how SORA 2.5 structures ground and air risk reasoning in a conservation context, while retrospective review identifies limitations in containment, Operational Safety Objectives, and tactical mitigation performance requirements. Field trials involved five concurrent teams and 30 personnel conducting over 260 flights and more than 60 h of UAS activity across the Ol Pejeta Conservancy, providing insights into multi-team coordination under field conditions. Field implementation revealed areas of misalignment between prescribed safety requirements and operational realities, prompting iterative adaptation of workflows and procedures. Observed outcomes included reductions in team size (25–50%) and procedural steps (18%), derived from retrospective comparison of field procedures. A lightweight Uncrewed Traffic Management prototype was also trialled, revealing practical limitations in conservancy environments. Finally, we present a ten-step framework for developing field-ready safety procedures to support risk-informed decision-making in non-standard operational contexts. The findings provide empirically grounded guidance on applying SORA principles to conservation UAS missions, without proposing a new risk framework or generalised operational model.

1. Introduction

Uncrewed Aerial System (UAS) have become an important tool for wildlife conservation by enabling high-resolution data collection across large or inaccessible areas with minimal disturbance to animals [1,2,3]. Their flexibility supports increasingly demanding applications, including Beyond Visual Line of Sight (BVLOS) and multi-UAS operations [4,5]. As such operations scale in range, duration, and coordination complexity, the primary challenge shifts from platform capability to ensuring missions are conducted safely within regulatory and operational constraints. Despite a growing body of literature on conservation UAS applications, few peer-reviewed studies document how established risk assessment frameworks [6] translate into field-ready procedures in remote conservation environments. Empirical evidence remains limited on how Specific Operations Risk Assessment (SORA)-derived requirements interact with transient population exposure, limited infrastructure, and non-standard airspace governance [7,8]. As a result, friction arising from applying risk-based frameworks in conservation campaigns remains under-reported.

The SORA provides a structured framework for assessing and mitigating risks in UAS operations [9,10,11], but its application in conservation settings is challenged by remote environments, limited infrastructure, and the need to adapt flight plans to animal movement and changing local conditions. As a result, ensuring that safety requirements remain compatible with research objectives requires field testing and context-specific adaptation [12]. Within this context, this study reports on 2025 field campaigns conducted at the Ol Pejeta Conservancy in Kenya (360 km²), building on earlier operations in 2024 [13] and extending them to multi-team BVLOS deployments involving multiple UAS, pilots, and lightweight coordination tools as part of the WildDrone project [14].

1.1. Kenyan Regulatory Context

The Kenyan Civilian Aviation Authority (KCAA)’s UAS Regulations 2020 [15] classify UAS operations into Categories A–C according to operational risk and complexity. BVLOS flights fall under Category C and require integration with Air Traffic Control (ATC) and compliance with the KCAA’s Rules of the Air, summarised in Appendix A.1.

At the time of writing, the KCAA did not issue standard BVLOS authorisations pending the establishment of a national Uncrewed Traffic Management (UTM). The reported operations were therefore conducted under a specific exemption within prohibited military airspace covering Ol Pejeta, subject to prior approval and real-time coordination with Kenyan Air Force (KAF) ATC. In the absence of formal separation standards between crewed and uncrewed aircraft, airspace safety relied on procedural coordination and radio communication. The operations required both a temporary import permit for each UAS and an operational permit for the BVLOS missions [16], facilitated through the local partner Kenya Flying Labs (KFL). Further regulatory definitions, permit requirements, and document templates are provided in Appendix A.2, and the full submission set is available in an open-access repository (https://github.com/GuyMaalouf/wilddrone-hackathon-2025-bvlos-ops, accesssed on 29 January 2026).

1.2. Study Aim

This study examines how SORA 2.5 can be applied to the planning and execution of BVLOS operations in protected conservation areas characterised by low population density and limited infrastructure, within a specific Kenyan regulatory context that includes coordination with military-controlled airspace. Drawing on field campaigns conducted in Kenya, the analysis evaluates how SORA 2.5 can be adapted to conservation environments and identifies sources of friction between formal safety requirements and the practical constraints of fieldwork. The study makes two main contributions: (i) an empirical assessment of how ground risk can be characterised in conservation areas, including the treatment of transient populations and the practical use of different intrinsic Ground Risk Class (iGRC) assessment approaches, and (ii) an analysis of how risk assessment outcomes can be translated into scalable, field-ready procedures for multi-team BVLOS operations supported by lightweight coordination tools. The aim is to provide empirically grounded insights into the practical application of SORA in non-standard operational contexts, rather than to propose a new risk framework.

1.3. Research Questions

The analysis is structured around four research questions:

• RQ 1—How can SORA 2.5 guide the planning of BVLOS operations in conservation areas?
  • RQ 1.1: How can SORA 2.5 ground-risk approaches be used in conservation settings, and how do they compare in that context?
  • RQ 1.2: Which safety constraints affect conservation fieldwork, and how can they be addressed?
• RQ 2—How should safety procedures be designed to meet the needs of conservation fieldwork?
  • RQ 2.1: How much efficiency can be gained by tailoring procedures to specific missions?
  • RQ 2.2: Which procedural elements can be automated, and how does automation affect crew workload and reliability?
• RQ 3—How can BVLOS operations be scaled to multiple teams while preserving safety in the field?
  • RQ 3.1: How can crew configurations be optimised to improve scalability while maintaining safety, and what trade-offs arise?
• RQ 4—What needs to happen for field-supported BVLOS operations to become fully remote?

1.4. Reading Guide

Readers seeking the main insights can begin with Section 5 (Discussion), which answers the research questions and synthesises regulatory, procedural, and operational findings. Section 2, Section 3 and Section 4 provide the supporting analysis: Section 2 outlines the SORA-guided risk assessment, including the comparison of population-density methods and tabulated versus analytical computation of intrinsic Ground Risk Class; Section 3 describes procedure design, crew coordination, and multi-team operations; Section 4 presents field results and recurring operational challenges. Appendix A summarises the Kenyan regulatory framework and permit process, Appendix B provides supplementary SORA calculations and definitions, and Appendix C contains mission-level field trial cards.

2. SORA 2.5-Based Risk Assessment in a Conservation Environment

SORA 2.5 was used as guidance to structure the risk assessment and safety planning for BVLOS operations under Kenyan airspace conditions. This section summarises how the relevant SORA steps informed the pre-operation classification of ground and air risk, and how the remaining elements of the framework were evaluated retrospectively following the January 2025 Ol Pejeta campaign, with detailed calculations, procedural artefacts, and extended analyses provided in the Appendix A, Appendix B and Appendix C to support, but not interrupt, the main narrative. Particular emphasis is placed on ground-risk assessment and the conservation-specific considerations affecting population-density estimation. Supplementary calculations and extended analyses, including post-operation evaluations of Operational Safety Objectives (OSOs), containment, and tactical-mitigation requirements, are provided in Appendix B. It is important to note that the SORA addresses only third-party ground and air risks; internal field-team safety considerations (such as wildlife encounters or heat exposure) fall outside its scope and were managed through separate operational protocols.

2.1. Step 1: Documentation of the Proposed Operation

What—Mission Objective: Conduct BVLOS flights to support wildlife conservation at Ol Pejeta, including ecological mapping, animal tracking, and environmental monitoring.

Where—Area of Operation: Flights took place within the 360 km² fenced Ol Pejeta Conservancy in central Kenya, a sparsely populated wildlife reserve containing several lodges and camps relevant to ground-risk assessment. The area lies under KCAA jurisdiction within military-controlled airspace, requiring coordination with both civilian and military ATC.

Which—Platforms Used: A total of 14 UASs were used, combining Commercial Off-The-Shelf (COTS) and research platforms. COTS multirotors comprised DJI Mavic 3 variants (Enterprise, Thermal, Pro), DJI Mini 3, Parrot Anafi, and CoDrone systems (Aluco, Noctu). Research platforms included the “Papa Smurf” quadcopter developed by the University of Bristol. Two fixed-wing systems were deployed: the Ebee X and a solar-powered University of Bristol glider. This mix provided complementary endurance, range, and payload capabilities for the different conservation tasks undertaken.

2.2. Step 2: iGRC

Step 2 estimates the iGRC by assessing the population-density within the operational footprint. We first defined a top-down containment geometry, then applied three population-estimation methods (qualitative field knowledge, tourism-based bottom-up, and raster-based top-down) and mapped the resulting densities to iGRC using both the tabulated SORA 2.5 method and the analytical approach of SORA Annex F.

2.2.1. SORA Semantic Map and Top-Down Volume Definition

We constructed a semantic map for Ol Pejeta, seen in Figure 1, based on the SORA semantic volumes, more details available in Appendix B.1.

A top-down approach was adopted: starting from the fenced conservancy boundary, we defined a bounded risk footprint that excludes surrounding villages and marks campgrounds as no-fly zones. This configuration aligns the ground-risk footprint with the conservancy fence, simplifies regulatory justification, and helps keep both routine and failure scenarios contained within the conservancy. Buffer-sizing considerations and a comparison with a bottom-up, path-based alternative are provided in Appendix B.2.

Note on units: This paper uses metric, imperial, and nautical units according to operational convention. Altitudes and regulatory limits are expressed in feet to match KCAA practice, horizontal distances and buffer sizes are given in metres or kilometres following UAS specifications, and airspace separation distances use nautical miles where appropriate.

2.2.2. Qualitative Estimation Using Field Knowledge

The initial iGRC assessment followed the SORA Annex F guidelines using a qualitative estimate of the average population density within the iGRC footprint (flight geography + contingency + ground-risk buffer). This estimate drew on prior BVLOS campaigns at Ol Pejeta and input from rangers and conservancy management. Human presence inside the fenced conservancy consists primarily of resident staff and their families, concentrated around staff compounds, operational centres, and a small number of lodges. Tourist activity is similarly localised around camps and attractions, with additional transient movement during game drives. Because the conservancy is fenced and not accessible to surrounding communities, there is no casual or unregulated foot traffic from outside the reserve.

Taken together, this restricted and spatially clustered pattern of human activity, combined with the absence of dense settlements or large buildings within the risk footprint, supported classifying the area as “lightly populated” (<50 people/km²). Classification criteria and supporting evidence are summarised in Appendix B.3.

2.2.3. Tourism-Based Local Estimation (Bottom-Up Approach)

A bottom-up estimate of population exposure was derived by projecting 2025 tourist visits to Ol Pejeta from official conservancy statistics (2016–2022) and regional post-COVID recovery rates from UN Tourism [17,18], yielding approximately 116,800 visitors (Figure 2).

A low-season adjustment for January resulted in an average of 352 tourists per day. Assuming an even distribution across 10 active camps and adding 10 staff per camp produced an estimated 46 people/km², corresponding to the “lightly populated” category. Intermediate values are summarised in

Table 1. Summary of the calculation steps of the bottom-up population-density estimate.

Parameter	Estimate
Projected annual visitors	≈116,800
Daily visitors in January (seasonally adjusted)	320 ± 10% → 352/day
Visitors per camp (10 sites)	36 tourists/camp
Added staff per camp	10 staff/camp
Total per camp	46 people/camp
Area considered	1 km2
Resulting population density	46 people/km2

, with derivations in Appendix B.4.

2.2.4. Population Raster Analysis (Top-Down Approach)

As a complementary top-down method, residential population within the conservancy was estimated using the 2020 WorldPop gridded dataset for Kenya [20]. Rasters were analysed at 100, 200, and 500 m resolutions to assess grid-size sensitivity (Figure 3).

A distribution derived from the 100 m grid (Figure 4) was used to characterise density variation across the operational area, yielding an average density of 12 people/km² and a maximum of 105 people/km². Aggregation to 200 m and 500 m reduced the maxima to 58 and 53 people/km², respectively, illustrating how coarser grids smooth local peaks.

Because SORA relies on the maximum population density, the choice of raster resolution has a direct effect on iGRC classification. At 200 m and 500 m, WorldPop maxima were close to, but slightly above, the qualitative and tourism-based estimates (<50 people/km²), whereas the 100 m grid produced an isolated extreme value more than twice as high. Figure 4 shows this as a tail event, highlighting how fine-resolution rasters can introduce single-cell outliers that disproportionately influence SORA outcomes. As WorldPop represents residential, typically night-time population, it does not capture transient tourist or staff presence within the fenced conservancy. Processing steps are summarised in Appendix B.5, and the full analysis code is available on GitHub (https://github.com/GuyMaalouf/opc-worldpop-analysis, accessed on 29 January 2026).

2.2.5. Ground-Truth Validation of Population Estimates

To validate the population estimates, daily visitor records from Ol Pejeta Conservancy were used as time-bounded, post-campaign ground-truth data derived from administrative records rather than real-time tracking. The observed mean of 309 tourists/day closely matched the bottom-up estimate of 352 tourists/day, while a transient spike exceeding 550 visitors occurred on 31 January (Figure 5).

Visitor distribution across camps showed strong spatial clustering (Figure 6), with most camps hosting small groups of 5–15 guests/day and a clear outlier at Sweetwaters Tented Camp, which averaged 112 visitors/day and exceeded 200 visitors on one day.

Approximately 600 staff were present daily across multiple facilities. These observations confirm that the conservancy remains lightly populated on average, while revealing short-lived surges and localised hotspots that are not captured by static or coarse-resolution population models and that must be addressed through conservative planning assumptions within the SORA process.

2.2.6. Determination of iGRC Based on the SORA Table

To complete Step 2, we identified the iGRC using the tabulated method in the SORA 2.5 framework [9]. The population-density value used for the table lookup was the <50 people/km² category, consistent with the qualitative and tourism-based methods described earlier. For the platform characteristics, we selected the University of Bristol solar-powered glider as the representative UAS because it had the largest wingspan and highest cruise speed among all platforms deployed. Its 3.8 m wingspan places it in the 3–8 m characteristic-dimension category, and its 20 m/s cruise speed falls within the <25 m/s category. Following the SORA requirement to use the higher-risk of the two, the 3–8 m column was selected.

The intersection of the <50 ppl/km² row with the 3–8 m column results in iGRC 5. The complete table used for this classification is reproduced in Appendix B.6.

2.2.7. Refined iGRC Estimation Using Annex F

We then applied the analytical method in SORA Annex F, which replaces the stepwise table with a continuous risk metric using the critical area $A_C$ and population density $D_{\mathrm{pop}}$. Using the same UAS parameters and density estimates, Annex F yielded iGRC 4, one class lower than the tabulated iGRC 5. Full calculations are provided in Appendix B.7.

This difference reflects the discrete structure of the tabulated method: the University of Bristol solar glider lies close to the boundary of the 3–8 m characteristic-dimension column, where small changes in wingspan or population estimates can produce relatively coarse shifts in the tabulated result. Annex F offers a finer-grained calculation in such edge cases, producing a slightly lower intrinsic ground-risk value.

2.3. Step 3: Final Ground Risk Class (GRC)

Starting from the refined iGRC 4 (SORA Annex F, Section 2.2.7), we applied exposure-reduction (M1) ground-risk mitigations from Annex B of SORA 2.5, implemented primarily through operational and procedural controls rather than additional onboard systems, to reduce exposure of uninvolved people (

Table 2. Ground-risk exposure-reduction mitigations (M1): Measures to reduce the likelihood of uninvolved people being present within the operational footprint under SORA Annex B.

M1a Sheltering	Low robustness. Most tourists remained inside safari vehicles or buildings when outdoors. Rangers operated mainly along roads and bush tracks, where vegetation, trees, and terrain features provided partial shielding.
M1b Operational restrictions	Medium robustness. Operations avoided overflying people and roads wherever possible. When road crossings were necessary, they were performed perpendicularly to minimise time above potential traffic. A 400 ft (120 m) horizontal buffer was applied around all campgrounds. Together, these restrictions reduced the proportion of the risk footprint where people may be present by more than 90%.
M1c Observation	Low robustness. Remote crews conducted pre-flight and in-flight visual checks of the area and maintained continuous radio contact with the Ol Pejeta control room to identify gatherings or transient activity and adjust operations accordingly.

). No energy-reduction devices (M2) were available on the platforms.

According to the Annex B reduction scheme, the combination of M1a (low), M1b (medium), and M1c (low) is valid and yields a three-step reduction, lowering the ground-risk classification from iGRC 4 to a final GRC 1. Full details are provided in Appendix B.8.

2.4. Step 4: Initial Air Risk Class (iARC)

The initial Air Risk Class (iARC) reflects the expected encounter rate with crewed aircraft in the operational volume before applying strategic mitigations. All flights took place inside the military-controlled restricted area HKR10 (Nanyuki), which overlaps a prohibited zone (HKP2) frequently used by KAF. This resulted in a classification of iARC–c.

Although the conservancy itself is rural, low-flying crewed aircraft, primarily tourist and charter traffic, were regularly observed, and one overflight during daytime operations required the UAS to descend below 30 m Above Ground Level (AGL) and temporarily pause the mission. Night flights were conservatively assigned the same iARC–c category.

2.5. Step 5: Strategic Air-Risk Mitigations

Strategic mitigations were applied before take-off to reduce the likelihood of encounters with crewed aircraft in the restricted and prohibited airspace over Ol Pejeta. These measures, combined with coordination from KAF, reduced the initial classification from iARC–c to a residual ARC–b. The applied mitigations are listed in

Table 3. Strategic air-risk mitigations: Measures to reduce the encounter rate with crewed aircraft.

Altitude & geography	Operations were capped at 400 ft (120 m) AGL west of the Ewaso River and 250 ft AGL east of it, the latter imposed by KAF due to proximity to Nanyuki Air Base. For BVLOS missions without observers, an internal ceiling of 250 ft (75 m)AGL was used to further reduce the risk of low-level encounters.
Chronology	Daytime flights were limited to a 2 km radius to ensure that visual observers could detect any approaching crewed aircraft with sufficient time to trigger contingency procedures. At night, when crewed aircraft are not permitted in this airspace, operations extended to 5 km.
Vertical separation	A minimum 500 ft (150 m) separation was maintained between the operational ceiling and the published floor of crewed aviation.
Aerodrome buffers	Exclusion zones of 9 km around Laikipia Airport and 5 km around Kamok Airstrip were removed from the operational volume (Figure 1).
External coordination	For each flight period, the safety manager obtained KAF ATC approval for morning, afternoon, and evening slots to ensure coordinated use of the restricted airspace.

.

The residual assessment relied on procedural deconfliction established by KAF ATC, including a vertical buffer between UAS operations and the lowest crewed flight levels and direct coordination before each operation. These measures ensured consistent separation within a controlled airspace and justified reducing the initial iARC–c to ARC–b.

2.6. Step 6: Tactical Mitigation Performance Requirements

For ARC–b operations, we applied the low-robustness Tactical Mitigation Performance Requirements (TMPR) defined in SORA Annex D. Tactical mitigations primarily relied on observers, ATC information, and radio-based situational awareness rather than automated detect-and-avoid systems. The measures implemented during the January 2025 campaign covered detection, decision-making, command, execution, and feedback using the available systems.

While these measures satisfy the qualitative requirements for low-robustness TMPRs, the absence of independent ground-truth air traffic data over the conservancy prevented empirical verification of detection probability, which was therefore treated as a known limitation in the retrospective assessment. A concise summary is provided in

Table 4. Low-robustness TMPRs applied for ARC–b operations: The table summarises the layered detection approach, predefined contingency actions, command-and-control performance, execution times, and feedback latencies used to satisfy the Annex D low-robustness criteria.

Detect	Detection relied on three layers: (1) web-based traffic feeds (Flightradar24); (2) local 1090 MHz ADS-B reception via software-defined radio; (3) continuous aeronautical VHF monitoring. Each pilot was supported by an airspace observer equipped with binoculars, and extended-range BVLOS missions [2 km (day)/5 km (night)] included additional remote observers positioned at the far end of the flight path. These measures satisfy the Annex D requirement for low-robustness detection for ARC–b operations, although the true detection probability cannot be empirically verified due to the absence of ground-truth traffic data over the conservancy.
Decide	When an intruder was detected, pilots executed predefined procedures: a rapid descent to below 100 ft (30 m), or if required, flight termination via the kill switch. Annex D specifies 60 ft (18 m) as the safe altitude; future campaigns will align with this value.
Command	Command-and-control links used DJI OcuSync (O2/O3) or 2.4 GHz RC systems (WiFi, FrSky, Herelink). Measured latencies across all platforms, presented in Appendix B.10, remained ≤150 ms, well within the Annex D requirement of ≤5 s.
Execute	The slowest platform descended at 3 m/s, yielding a descent from 400 ft (120 m) to 60 ft (18 m) in 34 s, meeting the Annex D limit of ≤60 s (see Appendix B.10 for calculations).
Feedback	Feedback updates were required within 5 s and end-to-end latency below 10 s. Local ADS-B tracks refreshed every 1 s, and Flightradar24 updated at 2–3 s intervals [21].

, with an adapted overview of the Annex D requirements in Appendix B.9.

2.7. Step 7: Specific Assurance and Integrity Level (SAIL)

The final ground risk (GRC 1) and residual air risk (ARC-b) combine to give a Specific Assurance and Integrity Level (SAIL) II operation according to the SORA Step 7 matrix. This level determines the robustness levels of containment requirements and Operational Safety Objectives (OSOs) that must be addressed in subsequent steps. The full SAIL matrix and mapping are provided in Appendix B.11.

2.8. Step 8: Containment Requirements

Containment requirements were based on adjacent-area population density and the characteristic dimension of the largest platform (3.8 m wingspan). Raster analysis of the 5 km adjacent area (Appendix B.12) indicated densities of 45–50 ppl/km², local maxima below 500 ppl/km², and no large assemblies within 1 km of the operational boundary.

Using the SORA 2.5 tabulated method (SORA main body Table 11 [9]; Appendix B.13), the operation corresponds to a Low containment requirement because the UAS exceeds the 3 m characteristic-dimension category. For comparison, the analytical SORA Annex F method (§5.3) was applied. The refined calculation in Appendix B.14, using $D_{pop}\approx 50$ people/km², yielded $iGR{C}_{\mathrm{adj}}=4$. As shown in Appendix B.15, no ground-risk mitigations applied in the adjacent area, and Appendix B.16 reports the corresponding Annex F Table 27 [9] outcome of no containment requirement for a SAIL II operation.

Both methods indicate minimal population exposure outside the conservancy, with the Annex F analytical approach yielding a less conservative yet compliant result. This reflects the intended SORA 2.5 trade-off: the tabulated method provides a faster, conservative option, whereas the analytical method requires additional effort but offers a more representative assessment in non-standard environments such as wildlife conservancies.

2.9. Step 9: Operational Safety Objectives (OSOs)

Due to timeline constraints, a clause-by-clause OSO assessment was not performed before fieldwork; instead, operational safety was ensured through pre-defined procedures, training, and checklists. Following the campaign, the operation was retrospectively reviewed against SORA 2.5 Annex E requirements for a SAIL II mission to assess compliance.

The review showed that several OSOs were met, others were partially met (typically due to incomplete documentation such as maintenance records or the absence of an Emergency Response Plan), and one OSO was not met, with the identified gaps primarily reflecting limitations in organisational assurance and documentation maturity rather than deficiencies observed during in-field operations, thereby providing a structured basis for prioritising improvements in future campaigns. A summary of OSO compliance is provided in

Table 5. Summary of compliance with the Operational Safety Objectives (OSOs) applicable to a SAIL II operation. Detailed integrity and assurance justifications are provided in Appendix B.17.

OSO	Title	Status	Primary Limitation
#01	Operator is competent and/or proven	Partially met	No training and role records
#02	UAS manufactured by competent and/or proven entity	N/A (SAIL III+)	–
#03	UAS maintained by competent/proven entity	Not met	No formal maintenance system
#04	UAS components essential to safe operations designed to ADS	N/A (SAIL IV+)	–
#05	UAS is designed considering system safety and reliability	N/A (SAIL III+)	–
#06	C3 link characteristics are appropriate for the operation	Met	–
#07	Conformity check of the UAS configuration	Partially met	No conformity check procedure
#08	Operational procedures defined, validated, adhered to	Partially met	No ERP; limited assurance records
#09	Remote crew trained and current	Partially met	Incomplete training coverage
#13	External services supporting operations are adequate	Met	–
#16	Multi-crew coordination	Met	–
#17	Remote crew is fit to operate	Met	–
#18	Automatic protection of the flight envelope from human errors	N/A (SAIL III+)	–
#19	Safe recovery from human error	N/A (SAIL III+)	–
#20	Human Factors evaluation/HMI appropriate for mission	Partially met	No formal HF evaluation
#23	Environmental conditions defined, measurable, adhered to	Met	–
#24	UAS designed/qualified for adverse environmental conditions	N/A (SAIL III+)	–

, with detailed integrity and assurance justifications in Appendix B.17.

2.10. Step 10: Comprehensive Safety Portfolio (CSP)

The final stage compiles all evidence generated across the previous SORA steps into a CSP. Although a formal one was not submitted to the KCAA, the materials produced for this project (operational description, risk assessment, and permit-application documents) effectively constituted the project’s version of a CSP. Together, they demonstrated how the operation addressed the safety claims and derived requirements for a SAIL II context. Future iterations would benefit from a structured compliance matrix to explicitly map each claim to its supporting evidence and improve traceability within the CSP.

3. Procedures Setup and Development

Building on the SORA-based risk framework, this section describes how safety requirements were translated into operational practice during the January 2025 field campaign at Ol Pejeta Conservancy, including organisational setup, coordination workflows, and standard operating procedures co-designed by safety specialists and field biologists.

Methodological scope of efficiency metrics: All efficiency metrics are derived from retrospective operational comparisons rather than controlled experiments. Baselines reflect conservative early-campaign configurations, while reduced configurations result from validated adaptations introduced with increased operational experience.

3.1. Crew Configurations and Roles Refinement

Operations initially employed a conservative crew structure with one pilot, one co-pilot, and dedicated airspace and ground observers per UAS (Figure 7a), ensuring redundancy and clear task separation during procedure validation and establishing a baseline for subsequent crew-scaling decisions. As operational confidence increased, observers were shared between teams operating from the same site (Figure 7b), with overlapping monitoring sectors allowing one observer pair to provide full coverage without exceeding workload limits.

This Multi-UAS setup was used when two teams co-located, typically during coordinated missions. It reduced staffing from 8 to 6 people, corresponding to an observed 25% reduction, without observed safety degradation, through defined sectors and coordinated timing.

Table 6. Crew-size optimisation across operation types. Summary of how crew refinements supported safe scaling as the number of UAS increased. Values indicate total personnel for each configuration and the reduction relative to the initial setup.

Operation Type	UAS	Crew (Before)	Crew (After)	Reduction
Initial setup	1	4	–	0%
Multi-UAS operations	2	8	6	25%
Swarm operations	3	12	6	50%
Extended field campaign	4	16	12	25%

summarises crew-size progression.

For swarm operations, 3 UAS flew synchronised missions using identical checklists, allowing the co-pilot role to be shared through centralised oversight (Figure 7c). This setup required 6 operators for 3 UAS, a 50% reduction relative to the initial configuration. By contrast, independent multi-UAS missions retained dedicated co-pilots because differing objectives required continuous availability for monitoring and emergency response.

During the extended campaign (February–June 2025), biologists operated in separate areas and required independent teams. Lower personnel availability and long-duration missions made larger crews impractical, so the co-pilot and airspace-observer roles were merged resulting in a three-person crew per UAS (Figure 7d). Across four teams, staffing decreased from 16 to 12 people (25% reduction). Because this combined role increased cognitive workload and reduced airspace-awareness robustness, maximum altitude was limited to 250 ft (75 m) AGL to reduce the likelihood of encounters with crewed traffic.

Overall, crew structures were simplified progressively as operational evidence accumulated, enabling reduced staffing while maintaining equivalent safety assurance.

3.2. Operations Planning and Coordination

Daily morning and afternoon briefings structured the overall coordination of up to 5 concurrent teams. These map-based sessions used the conservancy’s predefined patrol blocks to allocate scientific tasks, assign personnel and vehicles, and confirm platform availability. Patrol blocks were selected to avoid spatial overlap. When two teams required the same area, they merged and conducted the activity as a coordinated multi-UAS operation, with pilots positioned together to maintain direct verbal communication.

3.2.1. Third-Party Coordination

Team movements were reported to the Ol Pejeta radio room upon departure and return to maintain awareness of active UAS operations and to support rapid incident response. External communication with ATC was centralised through a single safety manager via mobile phone. ATC was contacted once before and once after each operational window, and the conservancy was treated as a single UAS operation for these exchanges, removing the need for team-level notifications.

3.2.2. Digital Coordination with WildOps

To support airspace coordination, a lightweight digital UTM prototype, WildOps (https://github.com/GuyMaalouf/WildOps-public, accessed on 29 January 2026), was trialled during the campaign. Pilots submitted planned missions by specifying the operation type (VLOS/BVLOS), time window, and take-off location, which were visualised as airspace volumes for approval by the safety manager.

In practice, this workflow was poorly aligned with field operations: although patrol blocks were assigned in advance, take-off locations were typically determined only after arriving on site and locating target animals, making pre-entry of coordinates impractical. On-site use required a laptop and mobile hotspot, increasing setup time, reducing usability in bright outdoor conditions, and adding workload during time-critical mission phases, while the lack of integrated positioning on laptops further complicated location entry.

Operationally, the safety manager required only confirmation that teams remained within the authorised flight geography and sufficiently separated from one another. This was more efficiently achieved by sharing a GPS pin via WhatsApp directly from the pilot’s mobile device, enabling rapid manual verification without detailed airspace visualisation or additional hardware.

Overall, the trial illustrates that desktop-based, pre-flight UTM workflows are poorly suited to dynamic conservation environments. Effective coordination in such contexts requires mobile-first tools with minimal user input, automatic position reporting, and interfaces designed for rapid, in-field decision-making rather than pre-planned submissions.

3.3. Operational Procedures, Iterations and Refinements

Field activities followed standard procedures spanning planning, pre-flight, in-flight, and post-flight phases, supported by contingency and emergency procedures. To manage these consistently across multiple operation types, the WildProcedures (https://github.com/GuyMaalouf/WildProcedures-checklist-generator, accessed on 29 January 2026) software was introduced. It maintained a central JSON (JavaScript Object Notation) database in which procedures were tagged by operation type (day/night, VLOS/BVLOS, with/without remote observer) and number of UAS (single, multiple, or swarm). A single command generated (i) a detailed procedures document and (ii) a compact field checklist. This automated approach replaced numerous static templates used during the June 2024 campaign and enabled rapid iteration, traceability, and customisation.

3.3.1. Adaptation Across Operation Types

Procedures were adapted to reflect differences in visibility, communication requirements, observer roles, and platform coordination, while maintaining a consistent core safety structure across all missions.

Table 7. Procedural adaptations across operation types during WildDrone 2025 field activities.

Type	Key Procedural Adaptations	Underlying Rationale/Risk Context
Day/Night	- Day: high-visibility vests, sun protection. - Night: portable lights, warm clothing, strobe activation, and a mandatory daylight terrain survey.	Day missions focused on crew visibility; night missions required enhanced lighting and pre-survey of obstacles.
VLOS/BVLOS	- VLOS: relied on line-of-sight communication.   - BVLOS: added communication checks and, for longer ranges, VHF radios.	BVLOS requires additional communication and surveillance measures to maintain situational awareness.
BVLOS Sub-types	- No remote observer: limited to 1 km from pilot.   - With remote observer: up to 2 km with VHF and RLOS. - Night BVLOS: up to 5 km with strobes and lighting.	Remote observer presence and time-of-day govern achievable range; longer distances require stronger mitigations.
Crew & Platform Variations	- Single-UAS: baseline procedures.   - Multiple-UAS: required pre-flight coordination, inter-pilot notification, and vertical separation.   - Swarm: applied the same deconfliction principles with synchronised control under one operator.	Multi-UAS activities introduce shared airspace management needs. Vertical deconfliction prevents mid-air conflicts, and standardised templates support consistent application across platforms.
Common Elements	Airspace and weather checks, IMSAFE assessment, failsafe configuration, and post-flight logging were consistent.	These steps ensured consistency and traceability across all operations.

summarises how steps and equipment were tailored to match each operation type.

3.3.2. Measured Efficiency Gains from Customisation

Table 8. Efficiency gain (%) across operation types, shown as the reduction in checklist items relative to the non-customised baseline. Gains were highest for simpler configurations such as VLOS, with smaller improvements for multi-UAS and extended-range BVLOS operations.

VLOS	BVLOS 1 km	BVLOS 1 km	BVLOS 1 km	BVLOS 2 km	Night BVLOS	Average
(single UAS)	(single UAS)	(multi-UAS)	(swarm)	(single UAS)	5 km (single UAS)
22.6%	20.2%	16.7%	19.0%	15.5%	13.1%	17.9%

reports the observed reduction in checklist items achieved through customised procedures (excluding contingency and emergency protocols), showing an average decrease of about 18%. This reduction serves as a practical proxy for time savings.

Figure 8 shows that all efficiency gains occur before take-off, where streamlining has the greatest operational impact. These stages are time-sensitive, as rapid preparation affects the ability to observe or track wildlife.

3.3.3. Evolution of Procedures over Time

Procedures were refined iteratively through co-development between safety specialists and field biologists to streamline workflows while maintaining safety.

Table 9. Progressive refinement of operational procedures between January and March 2025.

Evolution	Key Improvements and Outcomes
Procedures re-organisation	Checklist structure was redesigned to separate one-time first-flight tasks from a shorter recurring checklist for subsequent flights, reducing repetition and cognitive workload while retaining essential safety steps.
Communication and role definition	Emergency and contingency procedures were made more explicit, with standard callouts, clear responsibilities, and contact numbers placed on the first page to support rapid response.
Containment standardisation	Criteria for triggering Return-to-Home, land-on-site, or kill-switch actions were harmonised across operation types to improve crew familiarity and ensure consistent containment logic.
Threshold quantification	Ambiguous steps were replaced with measurable thresholds to increase repeatability, such as specifying landing-pad clearance distances and marking the site orientation.
Equipment requirements	Portable VHF/ADS-B units, originally carried universally, were later used only when required for higher-risk operations, with fixed equipment used otherwise.
Digital traceability	Daily checklists were generated through WildProcedures with embedded time stamps and version identifiers, ensuring teams used consistent and traceable procedure sets.

summarises the main areas of improvement and their outcomes.

4. Field Trials & Observations

The field trials presented in this section include both the WildDrone Hackathon operations conducted at Ol Pejeta Conservancy in January 2025 and the extended field campaigns that followed between February and June 2025. Together, these missions spanned diverse operational conditions and team configurations, providing a broad empirical basis for evaluating the practical application of the SORA-based safety framework introduced in Section 2. Rather than analysing individual missions in isolation, the trials are used to identify recurring operational constraints and mitigation patterns that inform the cross-cutting discussion in Section 5. They also support the operational insights discussed in Section 5, where five recurring constraints and their management strategies are examined:

(A) 

Large teams limit scalability but can be reduced safely through gradual optimisation;

(B) 

Repeated missions improve efficiency but require training to sustain vigilance;

(C) 

Range limits create time pressure but expansion depends on safe airspace integration;

(D) 

Environmental factors elevate operational risk and require crew-safety measures;

(E) 

Co-development ensures procedures remain practical and field-ready.

These observations illustrate how regulatory planning translated into field practice and provide the empirical grounding for the synthesis and generalisation presented in Section 5.

Summary of Field Trial Cards

Each operation was documented using a standardised Field Trial Card, ensuring consistent reporting of objectives, procedures, and observations.

Table 10. Summary of field operations conducted during and after the WildDrone Hackathon 2025.

	Mission Objective	Operational Challenges & Lessons
#1	Night tracking of lions (Panthera leo) to study movement and hunting behaviour.	• Predator activity increases ground risk. • Vehicle-based flying requires adaptations. • Night conditions require audio logging. • Fatigue management at night is essential. • Fixed home point limits efficiency.
#2	Evaluating a UAS swarm framework for coordinated, multi-perspective zebra monitoring [22,23,24].	• Large teams reduce scalability. • Short-range limits observation time. • Vertical spacing limits fleet size. • Procedural refinement boosts efficiency.
#3	Assessing wildlife responses to controlled UAS approaches to study vigilance of Plains zebras.	• Pre-flight setup minimises disturbance. • Procedures training improves efficiency. • Fixed home point limits flexibility.
#4	Aerial photogrammetry of rhinos (Diceros bicornis, Ceratotherium simum) for individual recognition and monitoring.	• Compact crew enhances flexibility. • Low altitude disturbs wildlife. • Range limits operational potential. • Exclusion zones hinder monitoring. • Co-development improves efficiency.• Crew training ensures readiness.
#5	3D reconstruction and semantic segmentation of savanna habitats for ecological modelling.	• Dense vegetation increases collision risk. • Bushland heightens wildlife risk. • Field experience improves workflow.
#6	Testing collaborative multi-UAS mapping methods for dynamic savanna environments [25].	• Safety protocols limit crew scalability. • Crew continuity improves familiarity.
#7	Field-testing edge-AI systems (WildWing, WildLive) for autonomous wildlife tracking [26,27,28].	• Custom platforms require adaptation. • Standardisation enables fast onboarding.
#8	Evaluating WildLive for real-time multi-species detection and tracking [27,28].	• Higher altitude reduces tracking accuracy.
#9	Assessing behavioural responses of large herbivores to UAS noise for ethical-flight guidance.	• Large crews reduce scalability [29,30]. • Dual-UAS coordination increases workload. • Safety procedures slow mission turnover. • Range limits operational coverage. • Crew training improves efficiency.
#10	Evaluating geolocation and size-estimation accuracy using COTS UAS and monoplotting algorithms [31,32].	• Crew optimisation improves efficiency. • Routine operations risk complacency. • Emergency training enhances readiness.
#11	Testing a stereo-vision payload for non-invasive measurement of animal body condition.	• Few pilots limit flexibility. • Large platforms reduce agility. • Short-range limits observation capacity. • Procedural refinement improves efficiency.

summarises the missions at a high level, while Appendix C provides the full operational details. Together, they demonstrate how recurring patterns observed across heterogeneous missions underpin the higher-level discussion of safety, scalability, and operational trade-offs in Section 5.

5. Discussion

This section synthesises the main findings and addresses the research questions (Section 1.3), drawing on quantitative results and qualitative observations from Section 2, Section 3 and Section 4. It interprets how SORA 2.5 can guide BVLOS operations in protected areas characterised by low population density, limited infrastructure, and dynamic field conditions, and extracts transferable insights for other non-standard operational environments with comparable ground and air risk profiles, while recognising that quantitative efficiency gains are context-dependent.

• RQ1. How Can SORA 2.5 Guide the Planning of BVLOS Operations in Conservation Areas?

The findings indicate that SORA 2.5 can guide BVLOS planning in conservation areas by clarifying dominant ground- and air-risk drivers and supporting proportionate mitigation strategies. As illustrated in Section 2, the framework helped distinguish risks requiring strategic spatial mitigations from those manageable through field procedures, informing decisions on flight exclusion areas, separation from aerodromes and roads, and conditions under which night operations remained acceptable. While specific mitigations remain context-dependent, this SORA-guided reasoning is applicable to other low-density, time-sensitive deployments where air risk is managed procedurally rather than through dense supporting infrastructure, such as remote environmental monitoring, disaster response, and search-and-rescue operations [33,34,35,36].

The assessment resulted in a sail II classification, after which operational safety was delivered primarily through structured field procedures. A retrospective review of relevant OSOs showed broad consistency with sail II expectations, indicating that well-designed field practices can achieve outcomes compatible with the framework. Future campaigns should formalise this alignment through more comprehensive OSO documentation.

A continuing limitation is the absence of reliable real-time air-traffic information, which reduces confidence in air-risk assumptions [37,38]. Progress in such environments will depend on improved airspace situational awareness, particularly through simple, affordable, and privacy-preserving Electronic Conspicuity (EC) solutions.

• RQ1.1. How Can SORA 2.5 Ground-Risk Approaches Be Used in Conservation Settings, and How Do They Compare in That Context?

• A—Population estimation relies more on local knowledge than modelling

Three population-density estimation approaches were compared against post-campaign daily administrative records used as ground truth, collected after more than 260 flights. The qualitative method provided reasonable first estimates but remained subjective where prior field knowledge was limited, while the bottom-up approach yielded similar values but required access to camp records that may not always be available. Raster-based estimates aligned only incidentally with these methods, as nearby village populations spilled into the conservancy and transient groups such as tourists and rangers, who constituted most people present during operations, were undercounted. Ground-truth observations confirmed this mismatch, revealing concentrated activity near Sweetwaters Camp that none of the models captured. Overall, reliable population inputs for SORA applications in conservancies require direct engagement with site managers and, where possible, on-site reconnaissance; raster products should therefore be treated as a secondary reference when field data cannot be obtained [39].

• B—Raster-based population estimates have limitations

The raster analysis highlighted three factors influencing intrinsic ground-risk classification. First, grid resolution affected apparent density: at Ol Pejeta, WorldPop estimates ranged from 105 ppl/km² at 100 m to 53 ppl/km² at 500 m, with 200 m providing the most representative scale in line with SORA Annex F guidance [40]. Second, maximum-cell values overstated exposure, as isolated 100 m cells exceeded 100 ppl/km² while 99% of the area remained below 37 ppl/km², suggesting percentile-based thresholds may provide a more stable representation. Third, global rasters omitted transient activity, which dominated human presence in the conservancy. Together, these factors indicate that raster-based estimates require careful interpretation and supplementation with local knowledge.

• C—Annex F analytical estimation provides a more representative assessment

Using a representative density, the intrinsic ground-risk class was computed with both the tabulated and analytical SORA Annex F methods. The lookup table placed the 3.8 m UAS just above a category threshold, resulting in iGRC 5, whereas the continuous analytical method yielded iGRC 4. This one-class difference reflects the conservative nature of tabulated approaches near boundaries; while it did not affect the resulting SAIL in this case, it could ease downstream requirements in other contexts. Analytical estimation is therefore preferable near boundary conditions, as it provides a more proportional assessment while remaining feasible with standard computational tools.

• RQ1.2. Which Safety Constraints Affect Conservation Fieldwork, and How Can They Be Addressed?

This section outlines the main safety constraints observed during field operations and how they can be addressed within a SORA-guided framework.

• A—Large teams limit scalability but can be reduced safely through gradual optimisation

Early missions relied on large teams, increasing coordination overhead and limiting scalability. Roles were progressively combined (e.g., shared observers and merged co-pilot/airspace-observer), enabling reductions without overloading individuals [41]. Crew optimisation should proceed incrementally, with clear role definitions and checks that reductions do not degrade airspace awareness, supported by mitigations where needed.

• B—Repeated missions improve efficiency but require training to sustain vigilance

Daily repetition improved coordination and reduced setup times as teams became familiar with checklists and terrain. However, routine operations and increasing automation risked reduced vigilance, particularly in multi-UAS contexts [42]. In-flight checklists therefore prompted active cross-checks of telemetry, airspace status, and observer reports. Sustaining efficiency without loss of alertness requires structured and recurrent training.

• C—Range limits create time pressure but expansion depends on safe airspace integration

Limited range created time pressure when animals moved quickly, highlighting the need for efficient coordination and repositioning. These constraints were driven primarily by airspace integration: many crewed aircraft over conservancies are non-cooperative and electronically invisible, requiring human observers to maintain separation. Extended-range missions (≤2 km by day, ≤5 km at night) therefore relied on remote airspace observers to maintain visual coverage, improving safety but increasing logistical complexity. Future BVLOS expansion depends on improved airspace-situational-awareness systems capable of detecting cooperative and non-cooperative aircraft; until then, hybrid approaches combining digital tools and observers remain the most practical solution.

• D—Environmental factors elevate operational risk and require crew-safety measures

Vegetation, wind, wildlife proximity, and night operations affected visibility, positioning, and line-of-sight continuity [43], requiring adjustments to launch sites and observer placement. Environmental limits were encoded in checklists, while crew-safety risks were managed through field protocols. Procedures should explicitly define the environmental conditions under which they remain valid, integrating operational and crew-safety considerations.

• E—Co-development ensures procedures remain practical and field-ready

Procedures developed during early missions were conservative and poorly adapted to field realities. During the January 2025 campaign, collaborative refinement by safety specialists and field teams removed redundancies, reordered steps, and strengthened contingency responses, including additional checks following an incoming crewed aircraft event. Iterative co-development maintains alignment with SORA expectations while ensuring procedures remain usable in daily conservation operations.

• RQ2. How Should Safety Procedures Be Designed to Meet the Needs of Conservation Fieldwork?

Developing safety procedures for conservation operations requires balancing regulatory requirements with field practicality. This section presents a compact methodology for designing field-ready procedures that remain clear, usable, and safe in dynamic field conditions, and that is applicable to other low-density, time-sensitive deployments.

1. 

Define scope and inputs: Specify operational scope (normal, contingency, emergency) and context (terrain, airspace, wildlife rules, UAS type, crew roles), consistent with SORA Annex E OSO #08(a) [9], ISO 21384-3 [44], and ICAO Doc 10019 [45].

2. 

Design the checklist family: Create checklists for planning, pre-/post-operation, all flight phases, emergency, and contingency. Keep emergency and contingency lists readily accessible, and maintain both detailed procedures and concise field checklists [44,45].

3. 

Apply clear placement rules and initiation anchors: Shift tasks that can be completed off-site to pre-operation; avoid repeated power cycles within a phase; use full checks only for the first flight and a reduced set for subsequent flights. Define initiation anchors for each checklist (e.g., pre-operation at the vehicle, first flight on arrival, subsequent flights before take-off, post-flight at landing) [46].

4. 

Set the right level of detail: Keep lists short and readable. Combine low-workload actions that occur together, separate safety-critical steps, use clear action verbs, avoid negations, and do not cluster demanding actions at the same moment [46,47].

5. 

Express environmental limits as field cues: Translate wind, visibility, terrain, and vegetation constraints into simple observable cues [9,45] so crews can apply limits without relying on instruments alone.

6. 

Minimise the “subsequent flight” checklist: After configuration and calibration are confirmed [9,46], retain only essential recurring steps: logging, power-up, battery and SD-card changes, error checks, GNSS and link confirmation, brief weather scan, team brief, go/no-go decision, take-off call, and immediate post-lift performance check.

7. 

Manage crew fatigue and workload: Apply fatigue-management principles (rest, hydration, role rotation, recovery days) rather than fixed-hour limits. Include the IMSAFE check in pre-operation and launch phases to confirm fitness to operate [9,45].

8. 

Separate emergency, contingency, and Emergency Response Plan (ERP): Define emergency procedures for immediate abnormal events, contingency procedures for predictable degradations, and the ERP for post-accident actions (e.g., crash, fire, injury, airspace incident), including relevant contact details [9,44,45].

9. 

Ensure formatting and legibility: Apply human-factors design principles: clear typography, left alignment, sufficient spacing, consistent structure, strong contrast, and visual emphasis on high-importance actions without implying optionality [46,47,48].

10. 

Validate and refine iteratively: Test procedures via table-top reviews and limited flights, refine them through daily debriefs, and maintain version control and change logs. Where higher assurance is required, follow SORA Annex §E.3 [9] and accepted means of compliance such as ASTM F3178-16 [49].

• RQ2.1. How Much Efficiency Can Be Gained by Tailoring Procedures to Specific Missions?

Tailoring procedures to mission type reduced the number of checks per operation by an average of about 18%, with most savings occurring before take-off, particularly during pre-flight and first-flight phases. Although this reduction is only a proxy for time saved and was not derived from a formal time–motion study, it indicates that modest adjustments to preparation steps can yield meaningful efficiency gains in daily fieldwork [47].

These gains resulted from aligning procedures with operational context. Differences in range, time of day, and number of UAS produced distinct risk profiles that warranted targeted adaptations: night missions required different equipment and omitted day-specific checks, while multi-UAS operations introduced coordination and vertical-separation checks to prevent conflicts during take-off and landing. Such tailoring removed irrelevant steps without compromising safety, and similar differentiation could extend to UAS-specific procedures as platform data become more standardised.

Using a centralised database in WildProcedures (https://github.com/GuyMaalouf/WildProcedures-checklist-generator, accessed on 29 January 2026) ensured that customisation did not erode standardisation. Automatically generating tailored checklists reduced checklist habituation [46,47] by removing items with no operational relevance, improving both efficiency and operator vigilance.

Overall, procedural tailoring offers the greatest benefit in operations with variable configuration or risk, such as changing range, complexity, or number of UAS, which is typical of field-deployed, time-sensitive missions. For repetitive single-UAS operations under uniform conditions, efficiency gains are more limited.

• RQ2.2. Which Procedural Elements Can Be Automated, and How Does Automation Affect Crew Workload and Reliability?

Automation can reduce workload by handling routine tasks, improving situational awareness, and reducing manual errors. Its greatest value lies in planning and pre-flight phases, where information gathering and verification dominate. In-flight tools can support vigilance through periodic prompts, while packing, inspection, and post-flight tasks remain largely physical and offer limited scope for automation. Similar task boundaries have been reported in other time-sensitive field operations, including emergency response and remote environmental monitoring, where preparation benefits most from automation while in-field decision-making remains human-centred.

Fully automatable elements include retrieving airspace and weather data, checking crew qualifications, and generating mission-specific checklists. Partial automation is appropriate near operational limits, where systems can flag uncertainty but require human confirmation. Safety briefings, final go/no-go decisions, and emergency coordination with ATC must remain human-led.

Automation also introduces human-factors risks, including reduced attention (automation complacency), over-trust (automation bias), and skill degradation (out-of-the-loop effects) [50,51,52,53]. Mitigation requires guardrails such as confirmation prompts, challenge–response calls, and periodic checks that maintain operator engagement.

Overall, automation can reduce workload and omissions while supporting situational awareness, but it should complement rather than replace human judgement, particularly during contingencies. Future work should examine how specific automation features affect workload, vigilance, and reliability in real operations.

• RQ3. How Can BVLOS Operations Be Scaled to Multiple Teams While Preserving Safety in the Field?

Scaling multi-team BVLOS operations requires clear spatial organisation, predictable workflows, and shared situational awareness [54]. The practices below summarise methods applied during five-team field operations (over 260 flights and 60 h) and are relevant to other low-density, time-sensitive deployments where multiple crews operate concurrently under procedural airspace separation. These insights are based on repeated operational use and qualitative safety outcomes rather than controlled comparative experiments.

1. Defined flight zones support effective deconfliction: A shared operational map defining authorised flight areas and contingency zones provides a common spatial reference and reduces overlap [55]. Dividing the area into clear zones simplifies coordination without relying on precise GPS positioning.

2. Operational windows reduce coordination load: Fixed activity windows (e.g., morning, afternoon, night) streamline logistics, improve resource sharing, and reduce the likelihood of overlapping operations [55].

3. Centralised supervision strengthens operational safety: A central safety manager maintains oversight across teams [54], confirms compliance with authorised geographies, and acts as the single point of contact with ATC and external authorities, preventing inconsistent reporting.

4. Robust coordination is essential for safe separation in shared airspace: When multiple UAS share airspace, co-located pilots can coordinate take-off and landing verbally, supported by standard vertical-separation rules. In multi-UAS missions this maintains independent trajectories, while swarm operations rely on synchronised checklists and a shared mission objective.

5. Reliable communication ensures safe coordination: Communication channels should be verified before each flight, with multiple technologies used to provide redundancy and maintain connectivity between teams.

6. Continuous feedback sustains safety at scale: Debriefs after each operational window identify coordination issues and inefficiencies. Feeding this information back into procedures and checklists helps prevent gradual erosion of safety margins as operational tempo increases.

• RQ3.1. How Can Crew Configurations Be Optimised to Improve Scalability While Maintaining Safety, and What Trade-Offs Arise?

Optimising crew configurations requires reducing team size without undermining workload balance or situational awareness. Field experience showed that once procedures stabilise and task demands become predictable, roles can be merged progressively in a structured manner [56]. Similar crew-scaling dynamics have been reported in other field-based BVLOS contexts relying on procedural airspace separation rather than infrastructure-supported detect-and-avoid, indicating applicability beyond conservation settings. The principles below summarise how teams can scale while managing safety trade-offs.

1. 

Begin with conservative crew configurations and iterate toward efficiency: Early missions should use larger teams to keep workload low while procedures mature. As processes stabilise, compatible roles can be merged through iterative review without loss of situational awareness [57].

2. 

Balance operational needs with safety trade-offs: Smaller teams may be necessary under field constraints but increase workload and reduce redundancy. When reductions are unavoidable, compensatory measures such as limiting altitude, range, or mission complexity should be applied, and reductions implemented only once their effects are understood.

3. 

Leverage shared roles in multi-UAS and swarm operations: In multi-UAS missions, observers can be shared if coverage areas are clearly defined. Swarm operations allow further role sharing (e.g., co-pilots) when UAS follow the same checklist and timing, provided responsibilities remain explicit.

4. 

Merge roles that share a single operational focus: Role consolidation is most effective when tasks contribute to a single operational focus. For example, co-pilot and airspace-observer roles may be combined due to their shared aviation-safety objective, whereas wildlife-observer roles require independence. Clear responsibility boundaries reduce cognitive switching and simplify qualification.

5. 

Maintain vigilance through cross-checks: Routine operations can reduce alertness; short challenge–response calls help maintain focus and prevent complacency during low-workload phases.

• RQ4. What Needs to Happen for Field-Supported BVLOS Operations to Become Fully Remote?

This discussion reflects the authors’ perspective on potential pathways toward remote BVLOS operations and is intended as an interpretative outlook rather than a statement of regulatory fact.

Transitioning from field-supported to fully remote BVLOS operations requires progressively replacing on-site roles with reliable airspace integration, communication, and autonomy. Drawing on evidence from this study, prior work on remote UAS integration [58,59], and current regulatory trends, this section outlines a plausible and transferable pathway rather than quantitative performance claims. The progression is incremental: extending range with minimal personnel, reducing local roles as assurance grows, and ultimately enabling remote or autonomous concepts such as UAS-in-a-box. Although derived from conservation deployments, these constraints and pathways apply to other low-density, time-sensitive operations, including rural disaster response, search and rescue, and remote environmental monitoring.

A. 

Airspace integration as the primary enabler

The main barrier to remote BVLOS operations is airspace integration, particularly in regions with mixed cooperative and non-cooperative traffic and limited surveillance infrastructure. Under the SORA TMPR framework [9], operators must detect a significant proportion (50∼90%) of intruding crewed aircraft, yet poor characterisation of traffic behaviour in regions such as sub-Saharan Africa prevents meaningful verification of detection performance. As a result, operational range remains constrained by visual observer coverage.

Emerging regulatory approaches emphasise EC for cooperative airspace management. The U.S. Part 108 draft rule links right-of-way to broadcast status [60], Europe’s U-space framework similarly relies on EC [61], and ADS-L provides a lightweight transmission option [62,63]. Trials such as the UK i-Conspicuity programme demonstrate detection ranges beyond 10 km, though cost, privacy, and interoperability remain barriers [64,65,66]. Universal EC adoption offers the most scalable path, while non-cooperative sensing suits specialised contexts [67].

B. 

Evolution of air-risk methodology

While SORA 2.5 strengthened ground-risk and containment structures, air-risk logic (ARC and TMPR) remains conservative. Field experience shows reliance on procedural mitigations and human-centred responses. As EC adoption increases, detect-and-avoid systems will require staged validation, beginning with conservative separation and relaxing criteria as evidence accumulates, mirroring earlier scaling of field-supported multi-UAS operations.

C. 

Connectivity and command resilience

Beyond airspace integration, Command and Control (C2) resilience becomes the next constraint. Current operations rely on RF line-of-sight links over 2–5 km, whereas remote missions require redundant architectures combining RF, cellular, or satellite links. Increasing range will favour higher-level intent-based control over continuous low-level inputs. Studies such as GENIUS indicate that 5G can meet safety-critical requirements, though rural coverage remains uneven [68]. Autonomous functions must tolerate short link losses and execute predefined contingencies, consistent with OSO #06 and OSO #13.

D. 

Incremental pathway toward remote operations

Progress toward remote BVLOS operations should follow a staged approach: start conservatively, validate performance, and expand complexity gradually. Interoperable EC and shared airspace situational awareness form the immediate foundation, enabling advances in connectivity and autonomy through evidence-based testing. Remote operations thus represent a continuation of learning cycles toward safer, more scalable UAS integration.

6. Conclusions

This study examined how SORA 2.5 can guide multi-team BVLOS operations in protected, low-density conservation areas, using field campaigns conducted in Kenya as an empirical basis. Rather than proposing a new risk framework, the work illustrates how SORA outputs can be interpreted and operationalised in environments characterised by limited infrastructure, transient populations, and procedural airspace management.

The analysis combined ground- and air-risk assessment with operational evidence from over 260 flights and more than 60 h of activity. Qualitative, bottom-up, and raster-based population estimates all indicated lightly populated conditions, but ground-truth records showed that global rasters can under-represent transient exposure associated with tourism and ranger activity. This reinforces the need for local data when assessing ground risk in conservation settings. Near threshold cases, the Annex F analytical method produced a less conservative intrinsic Ground Risk Class than the tabulated approach, while a retrospective review against Annex E clarified which Operational Safety Objectives were satisfied in practice and where further evidence is required. Although derived from a single conservancy, these risk-assessment challenges and trade-offs are representative of other low-density, time-sensitive field deployments.

Operationally, the campaigns demonstrated that structured spatial planning, shared procedures, and central oversight can support concurrent BVLOS missions with multiple teams. Progressive crew restructuring reduced team size by approximately 25–50% while maintaining observed safety margins, and mission-specific procedures generated using WildProcedures reduced checklist length by an average of 18%, with most gains occurring before take-off. These figures are indicative rather than statistically validated, but they show how procedural tailoring can improve efficiency without eroding safety. A lightweight coordination tool (WildOps) supported situational awareness but proved ill-suited to highly mobile crews and intermittent connectivity, highlighting that digital tools must be designed around field constraints rather than control-room assumptions. The resulting ten-step procedure-design framework provides a reusable, field-ready structure for translating SORA objectives and human-factors guidance into practical checklists for conservation and similar low-infrastructure operations.

Taken together, the findings show how SORA-informed risk assessment, locally grounded population estimation, compact crews, and tailored procedures can support scalable BVLOS operations while preserving safety transparency. For regulators and practitioners, the study illustrates how SORA 2.5 can be adapted to fenced reserves and procedurally managed airspace, how analytical ground-risk methods can avoid unnecessary conservatism, and how lightweight procedure-generation tools can support assurance arguments in the field.

Limitations and Future Work

This work is limited to a single conservancy, a specific regulatory context involving military-controlled airspace, and a restricted set of UAS platforms. Efficiency and safety claims are based on checklist structure, crew configuration, and observed operations rather than formal time–motion studies or incident statistics. Future work should strengthen evidence for partially satisfied Operational Safety Objectives, particularly those related to tactical mitigation performance, through systematic measurement of workload, timing, and safety events. The procedure-design approach and WildProcedures should be extended to additional platforms and environments, while coordination tools such as WildOps require mobile-first, connectivity-aware redesign to support wider EC adoption, improved detect-and-avoid performance, and resilient communications beyond short-range radio links, enabling a gradual transition to fully remote BVLOS operations.