A global method to identify trees outside of closed-canopy forests with medium-resolution satellite imagery

Abstract

Scattered trees outside of dense, closed-canopy forests are very important for carbon sequestration, supporting livelihoods, maintaining ecosystem integrity, and climate change adaptation and mitigation. In contrast to trees inside of closed-canopy forests, not much is known about the spatial extent and distribution of scattered trees at a global scale. Due to the cost of high-resolution satellite imagery, global monitoring systems rely on medium-resolution satellites to monitor land use. Here we present a globally consistent method to identify trees with canopy diameters greater than three meters with medium-resolution optical and radar imagery. Biweekly cloud-free, pan-sharpened 10 meter Sentinel-2 optical imagery and Sentinel-1 radar imagery are used to train a fully convolutional network, consisting of a convolutional gated recurrent unit layer and a feature pyramid attention layer. Tested across more than 215,000 Sentinel-1 and Sentinel-2 pixels distributed from -60 to +60 latitude, the proposed model exceeds 75% user's and producer's accuracy identifying trees in hectares with a low to medium density (less than 40%) of tree cover, and 95% user's and producer's accuracy in hectares with dense (greater than 40%) tree cover. The proposed method increases the accuracy of monitoring tree presence in areas with sparse and scattered tree cover (less than 40%) by as much as 20%, and reduces commission and omission error in mountainous and very cloudy regions by nearly half. When applied across large, heterogeneous landscapes, the results demonstrate potential to map trees in high detail and accuracy over diverse landscapes across the globe. This information is important for understanding current land cover and can be used to detect changes in land cover such as agroforestry, buffer zones around biological hotspots, and expansion or encroachment of forests.

References in corpus (19)

• Convolutional LSTM Network: A Machine Learning Approach for Precipitation Nowcasting
• Road Extraction by Deep Residual U-Net
• MultiResUNet : Rethinking the U-Net Architecture for Multimodal Biomedical Image Segmentation
• ResUNet-a: a deep learning framework for semantic segmentation of remotely sensed data
• Focal Loss for Dense Object Detection
• Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification
• Towards Better Exploiting Convolutional Neural Networks for Remote Sensing Scene Classification
• Rethinking the Inception Architecture for Computer Vision
• Boundary loss for highly unbalanced segmentation
• DropBlock: A regularization method for convolutional networks
• Layer Normalization
• Land Use Classification in Remote Sensing Images by Convolutional Neural Networks
• Super-resolution of Sentinel-2 images: Learning a globally applicable deep neural network
• Batch Renormalization: Towards Reducing Minibatch Dependence in Batch-Normalized Models
• Pyramid Attention Network for Semantic Segmentation
• Adaptive Gradient Methods with Dynamic Bound of Learning Rate
• Zoneout: Regularizing RNNs by Randomly Preserving Hidden Activations
• U-Net: Convolutional Networks for Biomedical Image Segmentation
• The Lovász-Softmax loss: A tractable surrogate for the optimization of the intersection-over-union measure in neural networks