Leveraging multiple data sources to create a time series of plant functional traits during the California megadrought

Ting Zheng,  University of Wisconsin-Madison,  tzheng39@wisc.edu (Presenter)
Zhiwei Ye,  UW-Madison,  ye6@wisc.edu
Natalie Queally,  University of Wisconsin-Madison,  queally@wisc.edu
Fabian Schneider,  Jet Propulsion Laboratory,  fabian.schneider@jpl.nasa.gov
Ryan Pavlick,  Jet Propulsion Laboratory,  rpavlick@jpl.nasa.gov
Ethan Shafron,  Jet Propulsion Laboratory,  ethan.shafron@jpl.nasa.gov
Philip Townsend,  University of Wisconsin,  ptownsend@wisc.edu

Imaging spectroscopy has been widely used to map and study the spatial patterns of foliar functional traits. However, creating multi-year time series of trait maps has been challenging due to consistency issues with the airborne image data over time and the availability and representativeness of training data used to build and validate trait models. In addition, there are challenges associated with evaluating retrospective analyses.
Here, we addressed these challenges using imaging spectroscopy to retrieve foliar functional traits for the Central and Southern Sierra Nevada region in California from June of each year 2013-2018.
We first developed a processing pipeline to improve the spectral consistency and the quality of spatial registration for images retrieved by NASA’s Airborne Visible/Infrared Imaging Spectrometer-Classic (AVIRIS-C). We determined the most consistent wavelength ranges for the AVIRIS-C data by 1) resampling and comparing the hyperspectral data to concurrent Landsat bands; 2) comparing the reflectance from spectrally invariant targets across different years.
To create representative training data for trait modeling, we pooled data from numerous projects and created a ~800-plot training set covering multiple years and including samples from California and the US Northeast and Upper Midwest. Using this training set, we developed partial least squares regression models for the pre-determined wavelength range and mapped traits across all years.
We compared the resulting trait maps to existing products for a few traits (chlorophyll content, nitrogen concentration, and leaf mass per area) and observed overall good agreement across the resulting outputs.
Our study demonstrates a practical approach to retrieving a time series of plant functional traits at large scales. Such time series will help us better understand the response of ecosystems to a changing climate.

Poster Location ID: 2-52

Presentation Type: Poster

Session: Poster Session 2

Session Date: Wed (May 10) 5:15-7:15 PM

CCE Program: BDEC