Skip to main content
All CollectionsHOW-TO GUIDES
Best Practices for Drone-Based Vegetation and Tree Imaging
Best Practices for Drone-Based Vegetation and Tree Imaging
Updated today

Introduction

One of the most commonly restored habitats that Restor encounters is woodland and forestry (including agroforestry) and Monitoring, Reporting and Verification (MRV) of these projects is fundamental to their success. There are many aspects to the restoration process that can benefit from MRV including: verification of tree planting pledges; measurement of tree mortality rates; profiling of vegetation species diversity; canopy cover and many others. Many of these MRV requirements can be met using drone imagery which is characterised by its high spatial resolution (although it does not have the same wide area coverage that is available from satellite imagery).

For many of these use-cases, the first step in the analytical process is in identifying the individual tree crowns. This is something that Restor has invested in, especially in open-canopy forests where the individual crowns don’t overlap (e.g. in early restoration forests where the trees are small, or in forests where the canopy cover percentage is low, such as Miombo woodland). Restor have trained an AI model to identify tree crowns from over 400,000 individual labelled trees in a range of forest types - achieving an accuracy of >75%. This model is used alongside a closed-canopy AI model (where the majority of tree crowns overlap) developed at the University of Cambridge in the UK. Using either of these models, individual tree crowns can be identified and then used as an input for more detailed analyses.

Capturing high-quality drone imagery of vegetation and trees requires careful planning to ensure optimal lighting, seasonal conditions, and image consistency. The goal is to produce data that can be reliably used for MRV and other ecological or forestry-related evaluations. This guide outlines the best practices for timing, environmental conditions, and flight planning to achieve the most appropriate imagery for your specific MRV use case.

Temporal/Environmental considerations

Season

  • Growing Season - Ideal for capturing peak foliage conditions, ensuring that the canopy is fully developed. This is beneficial for assessing canopy density, leaf condition and species differentiation. In most temperate climates, this occurs in late spring to early summer.

  • Dormant Season - When trees are leafless, it is easier to observe structural details like branch architecture and trunk distribution. This is useful for certain ecological studies, forest inventory, or biomass assessments. Generally late autumn, winter, or early spring (although there is no dormant season in evergreen forests), depending on the local climate. Identification of trees without leaves on is possible with the Restor model, but the accuracy is much less as there were not as many labelled trees without leaves on used in the training data.

Local Phenology

  • Identify the phenological stage of vegetation relevant to your requirements.

  • If assessing health (e.g., via multispectral or thermal imagery), capturing imagery during peak leaf maturity can maximise the differentiation in reflectance patterns.

Local Climate

  • Consider drought conditions or unseasonably warm/cold spells that might affect vegetation appearance.

  • In tropical regions, the dry season may be preferable to avoid heavy cloud cover (and therefore less illumination) and precipitation.

  • In arid regions, the cooler months may provide better visibility due to reduced dust and haze.

Time of Day Considerations

  • Late morning to early afternoon - (approximately 10:00–14:00 local time) the sun is typically high enough above the horizon to minimise shadows and provide even illumination across the canopy. Consistent lighting reduces image contrast issues and makes orthomosaic stitching more accurate. Neutral midday lighting is especially important for multispectral or hyperspectral sensors that rely on consistent illumination.

  • Early morning and late afternoon - the low sun angle can create long shadows, increasing contrast and potentially obscuring lower parts of the canopy or vegetation layers. This complicates image interpretation and image processing. However, if your goal includes capturing the three-dimensional structure or emphasising texture, consider slightly off-peak times for added shadow contrast (e.g., mid-morning or mid-afternoon, rather than high noon). In some specific use cases (such as estimating the height of trees from their shadows in low density forests) this time of day is ideal.

Weather and Environmental Conditions

  • Sunny or lightly overcast days provide stable, predictable lighting.

  • Avoid heavy cloud cover, as rapid changes in brightness can result in inconsistent exposure across the flight (see Figure 1).

  • Calm winds ensure stable flight paths and sharper images.

  • Minimal breeze reduces risk of motion blur and improves mosaic stitching accuracy.

  • Low humidity can improve image clarity and sensor performance.

  • Avoid smoke, haze, or dust, which can reduce image quality and spectral sensor accuracy.

Figure 1 Example of a series of drone flights with different illumination due to sun and clouds

Sensor and Camera Considerations

  • Use a fixed (manual) exposure setting if lighting conditions are stable.

  • Lock white balance to maintain colour consistency across images.

  • For vegetation health (e.g. NDVI), ensure the camera is equipped with the appropriate multispectral or hyperspectral sensors.

Flight Planning

  • Altitude - maintain a sufficient altitude and plan for image overlap (typically 70–80% forward and 60–70% side overlap) to ensure a seamless orthomosaic and 3D reconstruction. For tree crown delineation, aim to achieve a target resolution of 10cm or better - the AI models were trained on 10cm imagery.

  • Schedule flights to coincide with the chosen optimal lighting window.

  • Break large areas into manageable flight blocks to ensure consistent lighting conditions throughout each block.

  • Plan flight routes perpendicular to the sun’s position when possible to minimise variations in reflectance.

  • For monitoring changes in vegetation health or phenology, schedule repeated flights at the same time of day and under similar weather conditions. Ideally at the same time of year as well. Consistency in timing and conditions improves the reliability of comparative analyses.

Post-Processing Considerations

  • Apply radiometric calibration to ensure accurate reflectance measurements.

  • Perform image orthorectification and mosaic creation using consistent ground control points (GCPs) to ensure spatial accuracy.

  • Conduct quality checks and remove images captured under non-ideal conditions (e.g., sudden cloud shadows).

Figure 2 Example of image artefacts caused by orthomosaic stitching issues

Summary of ideal conditions for tree detection

  • Season: Peak growing season for foliage analysis; leaf-off season for structural analysis.

  • Time of year: Choose times when vegetation is at desired phenological stages relevant to the study.

  • Time of Day: Late morning to early afternoon for stable, even illumination and minimal shadows.

  • Weather: Clear or lightly overcast skies, minimal wind, stable atmospheric conditions.

Did this answer your question?