Use of partial-coverage UAV data in sampling for large scale forest inventories

- Combination of probability sample of UAV data with field plots
- Timber volume estimation under hybrid inference
- Comparison against field design-based and complete cover ALS model-based estimation
- UAV hybrid estimation four times more efficient than field design-based
- Precision of UAV hybrid estimation similar to complete cover ALS model-based

The use of image-based three-dimensional data from unmanned aerial vehicles (UAV) has proven effective for forest inventories. However, limitations in the range of UAV operations hinder their use in large scale applications. Use of partial-coverage UAV data in combination with field plots may increase precision of field-based estimates of forest resource parameters and may offer a cost-effective alternative to wall-to-wall acquisition. In this study, data from UAV collected in systematically distributed blocks and combined with field plots in a two-phase design with hybrid inference (UAVHYB) were used to estimate mean volume and its precision (standard error) for a 7330 ha forest area in Norway. Because in UAVHYB the field data do not necessarily need to come from a probability sample, the approach offers great flexibility in field data collection. The estimate of precision using UAVHYB was compared to two alternative inventory approaches differing in terms of sampling design and mode of inference, i.e., (1) single-phase probabilistic sampling relying only on field samples for which the design-based approach to inference using simple random sampling estimators was adopted (FIELDDB) and (2) a model-based approach independent from design assumptions for which wall-to-wall airborne laser scanning data were applied (ALSMB). Relative efficiency (RE), calculated as the ratio between the estimated variances of two different inventory approaches, was used as measure for improvement in variance for one approach over the other. Comparison of UAVHYB against FIELDDB revealed that the use of the former was up to four times more efficient than the latter (RE = 4.4). This translates to a need for 4.4 times as many field plots under simple random sampling for a FIELDDB estimate to be equally precise as the UAVHYB estimate. For ALSMB the increase in efficiency compared to UAVHYB was limited (RE = 1.6). The study also demonstrated that the precision under UAVHYB can be improved when including additional field data from other inventories to enhance the model. Cost estimates for each inventory approach were compared, revealing that UAV may be a cost-effective tool for large scale forest resource assessments.