Development of a Forest Structure Essential Biodiversity Variable for ’s Biodiversity Observation Network Patrick Jantz1, Susana Rodríguez-Buriticá2, Maria Cecilia Londoño2, John David Armston3, Steven Hancock4, Hao Tang4, Laura Duncanson4, Ralph Dubayah4 and Scott J Goetz1, (1) Northern Arizona University, SICCS, Flagstaff, AZ, United States, (2) Instituto von Humboldt, Bogota, Colombia, (3) University of Queensland, St Lucia, Australia, (4) University of Maryland College Park, MD, United States

Structure Type 1 2 3 4 5 6 7 8 9 10 MOTIVATION APPROACH Caqueta moist forests 13 2 1 11 42 25 5 1 0 0 Catatumbo moist forests 24 10 2 40 18 2 2 1 0 0 Cauca Valley dry forests 86 5 4 3 1 0 0 0 0 0 • The vertical distribution of vegetation is an important component of habitat • We used LAI strata and derived metrics from ~50,000 GLAS footprints (Tang et al. Cauca Valley montane forests 31 10 7 33 16 1 1 0 0 0 suitability for many animals 2014) in Colombia to model vegetation structure as a function of climate, elevation, land cover, and the human footprint. We clustered footprints using LAI strata to Choco-Darian moist forests 43 4 2 14 24 5 4 3 1 0 • We do not currently have sufficiently detailed observations, sampled at high derive a set of characteristic canopy profiles and summarized these by ecoregion. Cordillera Oriental montane forests 58 7 5 14 15 0 0 0 0 0 enough density, to characterize the vertical distribution of vegetation in Earth’s Eastern Cordillera real montane forests 58 8 5 13 10 4 1 1 0 0 most diverse forests varzea 56 11 0 22 11 0 0 0 0 0 • This project leverages the Global Ecosystem Dynamics Investigation (GEDI) to Japurai-Solimoes-Negro moist forests 9 5 1 19 34 23 7 2 0 0 overcome limitations of existing data RESULTS Magdalena-Uraba moist forests 78 8 4 8 1 0 0 0 0 0 • Planned outputs are designed to address Aichi Targets and Sustainable Height Magdalena Valley dry forests 75 7 5 10 2 1 0 0 0 0 Development Goals related to forest structure and condition R2 – 0.79 Magdalena Valley montane forests 59 10 7 15 6 1 1 1 0 0 Napo moist forests 28 7 3 14 27 16 3 1 0 0 Aichi Targets Negro-Branco moist forests 19 3 1 17 41 15 2 0 0 0 5 - degradation and fragmentation of natural habitats is significantly reduced Northwestern Andean montane forests 16 8 5 35 27 3 4 3 0 0 11 - establish "ecologically representative and well connected systems of protected Patia Valley dry forests 25 0 25 25 25 0 0 0 0 0 areas" Purus varzea 8 1 0 7 46 27 8 2 0 0 14 - ecosystems that provide essential services are safeguarded, "taking into Rio Negro 77 5 0 8 9 1 0 0 0 0 account the needs of women, indigenous and local communities, and the poor and Santa Marta montane forests 0 0 0 100 0 0 0 0 0 0 the vulnerable" Cover 2 Sinu Valley dry forests 74 11 6 9 0 0 0 0 0 0 15 - "By 2020, ecosystem resilience and the contribution of biodiversity to carbon R – 0.73 Solimoies-Japura moist forests 5 1 0 6 32 36 13 6 1 0 stocks has been enhanced“ South American Pacific mangroves 33 13 2 14 29 4 3 2 0 0 Sustainable Development Goals 15.1 - "ensure the conservation, restoration and sustainable use" of forests are "in Western moist forests 38 8 5 40 7 0 0 2 0 0 line with obligations under international agreements" Percent of each forest structure type in each WWF ecoregion. Structure types are 15.2 - "promote the implementation of sustainable management of all types of ordered from shortest to tallest. forests, halt deforestation, restore degraded forests" 15.4 - "ensure the conservation of mountain ecosystems, including their Evenness biodiversity" R2 – 0.52 DISCUSSION • Vegetation height and cover were more predictable than vertical heterogeneity metrics with the set of explanatory variables used here. Evenness of LAI and OBJECTIVES standard deviation (not shown) were less predictable than height and cover although the models were still able to explain ~40-50% of variability. Land cover • Develop a consistent and scalable workflow that uses spaceborne lidar and elevation were the strongest predictors of vegetation structure. measurements to provide unbiased estimates of the extent and distribution of forest structure types • Vegetation structural types consisted of a handful of high cover, short stature Random forest variable importance plots for three vegetation structure metrics derived from GLAS types followed by progressively taller stature types with LAI shifted higher in the • data. Evenness is Shannon’s diversity normalized by maximum diversity expected for a given Work with the Humboldt Institute to incorporate forest structure data and canopy. workflows into Colombia’s biodiversity observation vegetation height. dsl=dry season length, ECO_NAME=WWF Ecoregion, map=mean annual precipitation, mint=minimum temperature, maxt=maximum temperature, hf=human footprint, srtm=elevation, lc=IGBP cover types from MODIS. • Moist forest and varzea ecoregions contained the largest fraction of tall stature types. The Choco ecoregion was heterogeneous in structure types, perhaps QUESTION showing a range of conditions related to recent disturbance.

• How does vegetation structure vary across environmental and land use • Additional information related to disturbance history could place structural types gradients in Colombia? in context, helping to distinguish between forests that are biophysically limited and those that are recovering from disturbance. Study Area • Additional remote sensing variables (reflectance, texture, backscatter) could • Colombia is a mega-diverse country with an active biodiversity monitoring improve vertical heterogeneity models, providing information on ecosystem program and active involvement in GEO BON structure above and beyond vegetation height and cover. NEXT STEPS • The Global Ecosystem Dynamics Investigation launched in December, 2018.

• The algorithms developed in our project will be applied to GEDI data to characterize forest structure across Colombia.

• Plot, aircraft lidar, and drone lidar data will be used to validate the vegetation structure EBV. Acknowledgements • NASA Group on Earth Observations Work Programme, Characteristic LAI profiles for 10 forest structure types in Colombia. Forest Grant #: 80NSSC18K0338 Study area showing Colombia’s major ecological regions (left panel) and screened structure types were defined via hierarchical clustering on scaled LAI estimates for Geoscience Laser Altimeter (GLAS) footprints (right panel). 5 m height strata. • GEDI Science Definition Team