Abstract:
The efficient implementation of REDD+ programs and local sustainable forest management needs reliable data on species composition and distribution, forest biomass, and carbon storage, which are presently lacking in the majority of African vegetation formations. The study explored the use of unmanned aerial systems (UAS) imagery and associated processing tools in the management of the Miombo woodlands in Zambia. Four different studies were undertaken to meet the overall objective of this study. In order to have an overall understanding of the global application of UAS in forestry and the implications for its application to the Miombo woodlands, the first study was based on a review of the application of UAS in forest management and monitoring with a focus on challenges and opportunities for use in the Miombo woodlands. UAS technology, key attributes of the Miombo woodlands, and applications of UAS in forestry at the global and sub-Saharan African levels were reviewed, which enabled us to identify key prospects and challenges for UAS applications in the Miombo region.
As a demonstration of potential applications of UAS technology for managing the Miombo woodlands, the second study was focused on the use of multi-date and multi-spectral UAS imagery to classify dominant tree species in the wet Miombo woodlands in the Copperbelt Province of Zambia. Multi-date, multispectral UAS images taken at key phenological stages (leaf maturity, transition to senescence, and leaf flushing) and object-based image analysis (OBIA) with a random forest algorithm were utilized to classify the five dominant canopy species of the wet Miombo woodlands. The research found that combining multi-date raw band multi-spectral data with derived spectral indices produced better classification results (87.07% overall accuracy (OA), 0.83 Kappa) than using the best single-date multi-spectral data (80.12% OA, 0.68 kappa). The results from this study demonstrated the potential of using multispectral UAS imagery and phenology to map individual tree species in the Miombo woodlands of Zambia.
The third study was based on the application of UAS-Lidar to estimate forest structural attributes, which are critical to the management of the Miombo woodlands. UAS-Lidar data was used to estimate above-ground biomass, basal area, diameter at breast height, and volume, using multiple linear regression. The results indicate that the UAS-Lidar estimations provide the requisite degree of precision (relative root mean square error (RMSE): 3.40 - 20.89%) required for fulfilling international carbon reporting requirements and local forest management objectives. Furthermore, the use of unmanned aerial systems (UAS) equipped with Light Detection and Ranging (Lidar) technology offers a significant enhancement to the already utilized approaches for assessing Forest Structural Attributes (FSA) in the Miombo woodlands.
The fourth study was focused on bridging the spatial data gap that exists between detailed field inventory methods and satellite-based remote sensing methods that are required for wall- to-wall mapping of the Miombo woodlands. This study conducted a two-phase sampling design for wall-to-wall forest structural attributes estimation, where areas covered by a UAS-lidar were sampled by field plots and areas covered by wall-to-wall satellite imagery were sampled using a UAS-lidar. The results revealed that using UAS-lidar as reference data for predicting AGB using Sentinel-2 image metrics produced better results (Adj-R2 = 0.70 Mg/ha, RMSE = 27.97 Mg/ha) than using direct field estimated AGB and Sentinel-2 image metrics (R2 = 0.55 Mg/ha, RMSE = 38.10 Mg/ha). The results obtained demonstrated a practical solution to managing the Miombo woodlands using the available technology at multiple spatial scales.
The synthesis of these studies provides a holistic contribution for utilization of UAS technology and its accompanying processing tools in improving the acquisition of inventory data for the purpose of managing the Miombo woodlands in Zambia. This is a crucial necessity in effectively managing the diverse forested landscapes in the region.