Mapping Desert Blooms

Project title: Mapping the Spectral Colors of Blooming Deserts with EMIT

Despite the harsh conditions, the floristic composition of arid and semi-arid deserts is rich and diverse in endemic species, as evidenced by spectacular seasonal wildflower blooms. Resistant seeds, bulbs, and rhizomes can remain dormant under the dry soil for years, but after sufficient precipitation accumulates, an unusual and explosive development of ephemeral flowering plants occurs: a super bloom. Super blooms have been reported primarily in desert transitions towards Mediterranean ecoregions and are often triggered by climate-ocean conditions associated with El Niño. However, a detailed spatio temporal assessment of the blooming desert" phenomenon in arid and semi-arid ecological regions worldwide (e.g., California/Arizona, South Africa, Australia, Chile, Peru) is still needed. Ephemeral flowering triggers relevant changes in the ecosystem, such as temporary fractional cover expansions, outbreaks of pollinators and predators, and microclimate changes: higher evapotranspiration and lower surface albedo. Monitoring such phenology dynamics across large scales can be performed by integrating remote sensing with field-based ecology.

The unique optical properties of flowers allow imaging spectroscopy to be used as a non destructive tool to study flowers through space and time. The high spectral resolution and fidelity of imaging spectrometers like EMIT allow the retrieval of spectral signatures that comprise the weighted contribution of materials at smaller scales than the ground sampling distances (e.g., 60 m for EMIT). However, disentangling the mixed contributions of green leaves, flowers, non-photosynthetic canopy components, and background covers (e.g., soil, shadows) to spectroscopic reflectance measurements at different spatial resolutions is challenging. Moreover, most physical and hybrid modeling

and mapping methods at canopy scales ignore flowers, which have different spectral signatures and phenological patterns.

We aim to study wildflower blooms in arid and semi-arid desert ecoregions overlapping the EMIT data coverage. First, we will develop and evaluate spectral unmixing methods for quantifying fractional flowering coverage relying on the unique optical properties of flowers (especially in the 380-1000 nm spectral range). We will build an open-access

floral spectral library for endemic species in Mediterranean, arid, and semi-arid ecoregions, which, together with other available spectral libraries for soils, will be used as input for setting the unmixing framework to map the flowering abundance and their uncertainties based on the EMIT imagery. Informed by this spectral library, we will develop spectral unmixing techniques for mapping flowering areas and apply these techniques to EMIT scenes in these regions to identify and characterize blooming events. We will also evaluate whether blooms can be detected from other remote sensing measurements, including multispectral, thermal, and synthetic-aperture radar. Our current understanding of flower blooming in diverse ecosystems hinges predominantly on phenological data collected by field experts and volunteer networks, which is geographically limited and primarily concentrated in developed nations. Consequently, the majority of blooming forecasting indices are established based on ground-based meteorological data, calibrated using phenological models of a select few flowering species. This lack of comprehensive, globally-inclusive data constitutes a significant gap in our understanding of desert flower blooms. After using the EMIT hyperspectral data to identify desert flower blooms, we will investigate the underlying atmospheric and surface conditions inducing these blooming events. For each blooming event identified using EMIT, we will analyze antecedent remote sensing/modeled data for variables such as rainfall and temperature to understand the drivers of blooming events and develop predictive models for anticipating future blooming events.

Project Team

Yoseline Angel

University of Maryland, College Park