Glaciers are important sentinels of a changing climate, crucial components of the global cryosphere and integral to their local landscapes. However, many of the commonly used methods for mapping glacier change are labor-intensive and limit the temporal and spatial scope of existing research. This study addresses some of the limitations of prior approaches by developing a novel deep-learning-based method called GlacierCoverNet. GlacierCoverNet is a deep neural network that relies on an extensive, purpose-built training dataset. Using this model, we created a record of over three decades long at a fine temporal cadence (every two years) for the state of Alaska. We conducted a robust error analysis of this dataset and then used the dataset to characterize changes in debris-free glaciers and supraglacial debris over the last ~35 years. We found that our deep learning model could produce maps comparable to existing approaches in the capture of areal extent, but without manual editing required. The model captured the area covered with glaciers that was ~97% of the Randolph Glacier Inventory 6.0 with ~6% and ~9% omission and commission rates in the southern portion of Alaska, respectively. The overall model area capture was lower and omission and commission rates were significantly higher in the northern Brooks Range. Overall, the glacier-covered area retreated by 8425 km² (−13%) between 1985 and 2020, and supraglacial debris expanded by 2799 km² (64%) during the same period across the state of Alaska.

National forests in the western United States are divided roughly in half between lands without roads managed for wilderness characteristics and lands with an extensive road system managed for multiple uses including resource extraction. We investigated the influence of these land use designations on fire ignitions, fire extent, and fire severity over the last three decades. Although roadless areas experienced fewer fire ignitions and are generally cooler, moister, and higher elevation landscapes less conducive to fire, wildfire extent was far greater in these areas than in roaded areas. An area equivalent to approximately one-third of roadless areas burned in the last three decades, while an area equivalent to less than one-fifth of roaded areas experienced fire. Most of the largest fires that have burned on national forest land in recent years began in roadless areas. Despite greater fire extent in roadless areas, there was no significant difference in fire severity between roadless areas and roaded areas after accounting for biophysical differences between these management regimes. Although fire patterns in roadless areas may pose challenges to land managers, the available evidence suggests that the greater extent of fire in roadless areas may confer resilience to these landscapes in the face of climate change.

Light detection and ranging (LiDAR) has become a commonly-used tool for generating remotely-sensed forest inventories. However, LiDAR-derived forest inventories have remained uncommon at a regional scale due to varying parameters among LiDAR data acquisitions and the availability of sufficient calibration data. Here, we present a model using a 3-D convolutional neural network (CNN), a form of deep learning capable of scanning a LiDAR point cloud, combined with coincident satellite data (spectral, phenology, and disturbance history). We compared this approach to traditional modeling used for making forest predictions from LiDAR data (height metrics and random forest) and found that the CNN had consistently lower uncertainty. We then applied the CNN to public data over six New England states in the USA, generating maps of 14 forest attributes at a 10 m resolution over 85% of the region. Aboveground biomass estimates produced a root mean square error of 36 Mg ha⁻¹ (44%) and were within the 97.5% confidence of independent county-level estimates for 33 of 38 or 86.8% of the counties examined. CNN predictions for stem density and percentage of conifer attributes were moderately successful, while predictions for detailed species groupings were less successful. The approach shows promise for improving the prediction of forest attributes from regional LiDAR data and for combining disparate LiDAR datasets into a common framework for large-scale estimation.

Satellite-based forest alert systems are an important tool for ecosystem monitoring, planning conservation, and increasing public awareness of forest cover change. Continuous monitoring in tropical regions, such as those experiencing pronounced monsoon seasons, can be complicated by spatially extensive and persistent cloud cover. One solution is to use Synthetic Aperture Radar (SAR) imagery acquired by the European Space Agency’s Sentinel-1A and B satellites. The Sentinel 1A and B satellites acquire C-band radar data that penetrates cloud cover and can be acquired during the day or night. One challenge associated with operational use of radar imagery is that the speckle associated with the backscatter values can complicate traditional pixel-based analysis approaches. A potential solution is to use deep learning semantic segmentation models that can capture predictive features that are more robust to pixel-level noise. In this analysis, we present a prototype SAR-based forest alert system that utilizes deep learning classifiers, deployed using the Google Earth Engine cloud computing platform, to identify forest cover change with near real-time classification over two Cambodian wildlife sanctuaries. By leveraging a pre-existing forest cover change dataset derived from multispectral Landsat imagery, we present a method for efficiently developing a SAR-based semantic segmentation dataset. In practice, the proposed framework achieved good performance comparable to an existing forest alert system while offering more flexibility and ease of development from an operational standpoint.