Kurtis J. Nelson

Towards Integration of GLAS into a National Fuel Mapping Program

Comprehensive canopy structure and fuel data are critical for understanding and modeling wildland fire. The LANDFIRE project produces such data nationwide based on a collection of field observations, Landsat imagery, and other geospatial data. Where field data are not available, alternate strategies are being investigated. In this study, vegetation structure data available from GLAS were used to fill this data gap for the Yukon Flats Ecoregion of interior Alaska. The GLAS-derived structure and fuel layers and the original LANDFIRE layers were subsequently used as inputs into a fire behavior model to determine what effect the revised inputs would have on the model outputs. The outputs showed that inclusion of the GLAS data enabled better landscape-level characterization of vegetation structure and therefore enabled a broader wildland fire modeling capability. The results of this work underscore how GLAS data can be incorporated into LANDFIRE canopy structure and fuel mapping.

An Optimal Sample Data Usage Strategy to Minimize Overfitting and Underfitting Effects in Regression Tree Models Based on Remotely-Sensed Data

Regression tree models have been widely used for remote sensing-based ecosystem mapping. Improper use of the sample data (model training and testing data) may cause overfitting and underfitting effects in the model. The goal of this study is to develop an optimal sampling data usage strategy for any dataset and identify an appropriate number of rules in the regression tree model that will improve its accuracy and robustness. Landsat 8 data and Moderate-Resolution Imaging Spectroradiometer-scaled Normalized Difference Vegetation Index (NDVI) were used to develop regression tree models. A Python procedure was designed to generate random replications of model parameter options across a range of model development data sizes and rule number constraints. The mean absolute difference (MAD) between the predicted and actual NDVI (scaled NDVI, value from 0–200) and its variability across the different randomized replications were calculated to assess the accuracy and stability of the models. In our case study, a six-rule regression tree model developed from 80% of the sample data had the lowest MAD (MADtraining = 2.5 and MADtesting = 2.4), which was suggested as the optimal model. This study demonstrates how the training data and rule number selections impact model accuracy and provides important guidance for future remote-sensing-based ecosystem modeling.

Automated integration of lidar into the LANDFIRE product suite

Accurate information about three-dimensional canopy structure and wildland fuel across the landscape is necessary for fire behaviour modelling system predictions. Remotely sensed data are invaluable for assessing these canopy characteristics over large areas; lidar data, in particular, are uniquely suited for quantifying three-dimensional canopy structure. Although lidar data are increasingly available, they have rarely been applied to wildland fuels mapping efforts, mostly due to two issues. First, the Landscape Fire and Resource Planning Tools (LANDFIRE) program, which has become the default source of large-scale fire behaviour modelling inputs for the US, does not currently incorporate lidar data into the vegetation and fuel mapping process because spatially continuous lidar data are not available at the national scale. Second, while lidar data are available for many land management units across the US, these data are underutilized for fire behaviour applications. This is partly due to a lack of local personnel trained to process and analyse lidar data. This investigation addresses these issues by developing the Creating Hybrid Structure from LANDFIRE/lidar Combinations (CHISLIC) tool. CHISLIC allows individuals to automatically generate a suite of vegetation structure and wildland fuel parameters from lidar data and infuse them into existing LANDFIRE data sets. CHISLIC will become available for wider distribution to the public through a partnership with the U.S. Forest Service’s Wildland Fire Assessment System (WFAS) and may be incorporated into the Wildland Fire Decision Support System (WFDSS) with additional design and testing. WFAS and WFDSS are the primary systems used to support tactical and strategic wildland fire management decisions.

Mapping forest height in Alaska using GLAS, Landsat composites, and airborne LiDAR

Vegetation structure, including forest canopy height, is an important input variable to fire behavior modeling systems for simulating wildfire behavior. As such, forest canopy height is one of a nationwide suite of products generated by the LANDFIRE program. In the past, LANDFIRE has relied on a combination of field observations and Landsat imagery to develop existing vegetation structure products. The paucity of field data in the remote Alaskan forests has led to a very simple forest canopy height classification for the original LANDFIRE forest height map. To better meet the needs of data users and refine the map legend, LANDFIRE incorporated ICESat Geoscience Laser Altimeter System (GLAS) data into the updating process when developing the LANDFIRE 2010 product. The high latitude of this region enabled dense coverage of discrete GLAS samples, from which forest height was calculated. Different methods for deriving height from the GLAS waveform data were applied, including an attempt to correct for slope. These methods were then evaluated and integrated into the final map according to predefined criteria. The resulting map of forest canopy height includes more height classes than the original map, thereby better depicting the heterogeneity of the landscape, and provides seamless data for fire behavior analysts and other users of LANDFIRE data.