Geostatistical estimation of forest biomass in interior Alaska combining Landsat-derived tree cover, sampled airborne lidar and field observations

Abstract

Lidar provides critical information on the three-dimensional structure of forests. However, collecting wallto-wall laser altimetry data at regional and global scales is cost prohibitive. As a result, studies employing lidar for large area estimation typically collect data via strip sampling, leaving large swaths of the forest unmeasured by the instrument. The goal of this research was to develop and examine the performance of a coregionalization modeling approach for combining field measurements, strip samples of airborne lidar and Landsat-based remote sensing products to predict aboveground biomass (AGB) in interior Alaska's Tanana Valley. The proposed modeling strategy facilitates mapping of AGB density across the domain. Additionally, the coregionalization framework allows for estimation of total AGB for arbitrary areal units within the study area-a key advance to support diverse management objectives in interior Alaska. This research focuses on characterization of prediction uncertainty in the form of posterior predictive coverage intervals and standard deviations. Using the framework detailed here, it is possible to quantify estimation uncertainty for any spatial extent, ranging from point-level predictions of AGB density to estimates of AGB stocks for the full domain. The lidar-informed coregionalization models consistently outperformed their counterpart lidarfree models in terms of point-level predictive performance and total (mean) AGB precision. Additionally, including a Landsat-derived forest cover covariate further improved precision in regions with lower lidar sampling intensity. Findings also demonstrate that model-based approaches not explicitly accounting for residual spatial dependence can grossly underestimate uncertainty, resulting in falsely precise estimates of AGB. The inferential capabilities of AGB posterior predictive distribution (PPD) products extend beyond simply mapping AGB density. We show how PPD products can provide insight regarding drivers of AGB heterogeneity in boreal forests, including permafrost and fire, highlighting the range of potential applications for Bayesian geostatistical methods to integrate field, airborne and satellite data.

Addressing uncertainty and bias in land use, land use change, and forestry greenhouse gas inventories

“…Model-assisted estimators that utilize LiDAR and plot data have increased aboveground forest biomass precision by 2.5-6 times compared to plot-based simple random sample or post-stratified estimators (McRoberts et al 2013;McRoberts et al 2016;Gregoire et al 2016). Historically, the necessary LiDAR and radar data has been costly to collect and only intermittently available over space and time, but new and planned global LiDAR and radar missions, including GEDI, ICE-Sat2, and NISAR, have the potential to greatly improve LULUCF monitoring precision and to help align aboveground biomass monitoring methods across countries (Duncanson et al 2020;Babcock et al 2018). Ongoing availability of LiDAR or radar data will be critical to ensure countries can sustain new LULUCF monitoring methods.…”

Using matrix models to estimate aboveground forest biomass dynamics in the eastern USA through various combinations of LiDAR, Landsat, and forest inventory data

“…Antiquated approaches to predict AGB dynamics have been extensively used to provide robust estimates relying solely on forest inventory plots (Hall et al 2006, Ma et al 2018, but they are inefficient and difficult to implement over large spatial and temporal scales. Indeed the technologies of light detection and ranging (LiDAR) and timeseries Landsat surface reflectance high-level data products provide not only relatively cost-effective means, but also accuracy and spatial resolution to predict AGB dynamics over large domains of space and time (Powell et al 2010, Wulder et al 2012, Babcock et al 2018.…”

“…The approach of linking ground measurements with remotely sensed data has greatly improved efficiency and cost-effectiveness of AGB estimation in forests, and can directly provide AGB estimates based on regression models (Running et al 1999, Hu et al 2016. Such a process has become increasingly popular with the advent of LiDAR and synthetic aperture radar , Goetz and Dubayah 2011, Avitabile et al 2012, Neigh et al 2013, Tanase et al 2014, Margolis et al 2015, Babcock et al 2016, Fayad et al 2016, Hu et al 2016, Treuhaft et al 2017, the public-release of Landsat time-series data (Hall et al 2006, Labrecque et al 2006, Blackard et al 2008, Powell et al 2010, and the incorporation of both LiDAR and Landsat time-series imagery (Babcock et al 2018, Deo et al 2017b in the context of growing awareness of the role of terrestrial systems in the global carbon balance (e.g. Pan et al 2011).…”

“…However, it is not easy to implement more suitable AGB dynamic predictions to actual AGB dynamics for any model-assisted estimator (e.g. Forest Vegetation Simulator) that depends on field plots and remote sensing data being established as control variables (Babcock et al 2018).…”

Spatial Statistics on Amazon Rainforest Assessment: Spatially Stratified Inventory Processing

“…In this context, applications of spatial statistics in forest research have been explored in a plethora of studies over the last decade, which represents a real breakthrough on the field (GALEANA-PIZAÑA et al, 2014;PELISSARI et al, 2017;BABCOCK et al, 2018). Particularly, mapping of tree richness and biomass distributions have flourished, and results are encouraging to future research (YADAV and NANDY, 2015).…”