Chimera: A deep-learning approach for fusing multi-sensor data for forest classification and structural estimation
Abstract
In recent years, the increased availability and amount of high-resolution (<30-m) imagery and advancements in machine learning algorithms have opened up a new opportunity to fuse multiple datasets of varying spatial, spectral, and temporal resolutions. In this talk we describe a new approach to simultaneously classify forest land cover type and estimate continuous forest structure metrics using a deep learning model ensemble. Our approach applies an ensemble of multi-task convolutional neural network (CNN) model we call Chimera. The Chimera ensemble integrates varying resolution, freely available aerial and satellite imagery (e.g., NAIP, Landsat), as well as relevant environmental factors (e.g., climate, terrain) to classify five forest cover types ('conifer', 'deciduous', 'mixed', 'dead', 'none' (non-forest)) and to estimate four continuous forest structure metrics (biomass, quadratic mean diameter, basal area, canopy cover). We demonstrate our approach by training the Chimera ensemble on georeferenced Forest Inventory and Analysis (FIA) field plots from the USDA Forest Service within California and Nevada, and highlight validation metrics. Our modeling approach can estimate forested conifer type locations and structural attributes with high accuracy in a repeatable and cost efficient manner. Applications include inputs to fire behavior modeling and conservation area monitoring. Future implementations of the Chimera ensemble on a distributed computing platform could provide annual estimates of forest structure (biomass) and measurements of land-use change in the Western US over time.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.B21A..04C
- Keywords:
-
- 0410 Biodiversity;
- BIOGEOSCIENCES;
- 0439 Ecosystems;
- structure and dynamics;
- BIOGEOSCIENCES;
- 0480 Remote sensing;
- BIOGEOSCIENCES;
- 1910 Data assimilation;
- integration and fusion;
- INFORMATICS