SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. References

Remotely-sensed proxies have been acknowledged as powerful tools for estimating species’ spatial distributions, whatever the taxonomic group being considered. Jiang et al. (2013) provide a robust example of seeing the unseen by remote sensing, predicting the distribution of epiphyllous liverworts from SPOT Vegetation remotely-sensed data.

Whatever the ecological property being investigated, from the distribution of a certain species or group of species to the amount of variability in plant species over space and time, field sampling inevitably presents a number of issues, such as the development of a robust sampling design, the definition of the statistical population being sampled, and the time and cost of performing sampling in the field (Chiarucci 2007).

This is particularly true when considering spatially complex ecosystems. It is practically impossible to gather exhaustive information about the geometry of environmental and species variation over space at a certain time (Palmer 2007). However, remote sensing is a powerful tool for obtaining reliable information about species distributions in space and time, since it guarantees whole spatial coverage in a short period of time (Feilhauer & Schmidtlein 2009; Rocchini & Neteler 2012).

The first satellite ever launched, on 4 October 1957 – Sputnik 1, was capable of providing only one rough type of information, i.e. a discontinuous beep sound (see http://www.youtube.com/watch?v=lPFKd5p_t0s). Since the launch of Landsat in 1972, remote sensing has represented one of the most powerful tools for detecting, directly or indirectly, biodiversity properties over the entire Earth surface within a reasonable time span.

In other words, remote sensing has valuable spatial and temporal information content. Famous examples of sensors with a high temporal resolution, i.e. with a continuous acquisition of data with short revisit intervals, are: (1) SPOT Vegetation, managed by VITO (Vision on Technology, Belgium), which since 1998 has provided data for vegetation scientists each day at a spatial resolution of 1 km, covering a spectrum from 430 nm to 1750 nm in four bands, and (2) MODIS sensors (Moderate Resolution Imaging Spectroradiometer) onboard the Terra and Aqua satellites, operated by NASA, which generate four global coverages per day at pixel resolutions ranging from 250 m to 1000 m in 36 spectral bands, covering an electromagnetic spectrum from 620 nm to 14 385 nm.

A number of higher spatial resolution sensors exist, e.g. QuickBird (reaching 0.61 m) or Ikonos (reaching 1.0 m), but in some cases a high temporal resolution is needed to detect plant phenological status.

This is particularly true when aiming to map or estimate species distributions over large regions, whatever the hierarchical taxonomic group being considered. Once remote sensing is used as a proxy of e.g. plant biomass or vegetation phenology, those properties that are ‘unseen’ in the field can also be revealed. Jiang et al. (2013) provide an example of estimating the probability of the spatial distribution of epiphyllous liverworts – bryophytes mainly related to evergreen broad-leaved forests – based on weighted regression models using as explanatory variable the normalized difference vegetation index (NDVI) achieved in different time periods, finally focusing on spring NDVI. A similar example with multi-date NDVI is provided in He et al. (2009). NDVI is based on the difference between the reflectance of objects in the near-infrared (NIR) and the red portion of the electromagnetic spectrum, divided by their sum. It is one of the most-used proxies for vegetation-related studies, including diversity prediction (Gillespie 2005), species composition change estimation (He et al. 2009) and net primary productivity measurement (Irisarri et al. 2012), since it increases at high biomass due to (i) the high reflectance of vegetation in the NIR, which is linked to scattering processes at the leaf scale, and (ii) the low reflectance in the red spectrum due to the absorption by chloroplasts for photosynthesis (e.g. Fig. 1).

image

Figure 1. A MODIS (Moderate Resolution Imaging Spectroradiometer) NDVI image for the United States. NDVI ranges from -1 (yellow in this rendering) to 1 (red in this rendering), basically indicating low to high biomass content. Intermediate values are indicated by green and light blue (low to medium values) and blue colour (medium to high values). MODIS images are freely available from glcf.umiacs.umd.edu/ and from https://lpdaac.usgs.gov/products/modis_products_table. Data processed by GRASS GIS (Neteler et al. 2012).

Download figure to PowerPoint

The paper of Jiang et al. (2013) has four major strengths: (1) it clearly demonstrates in a quantitative manner how ‘unseen’ organisms can be mapped using remote sensing; (2) it shows that images based on a single date acquisition cannot allow the detection of intra-annual fluctuations in biomass, failing in some cases to robustly estimate species distributions; (3) it makes explicit use of the probability (or, better, suitability, sensu Bradley et al. 2012) of presence of epiphyllous liverworts, maintaining the ‘vagueness’ needed when simulating species occurrence from remotely sensed data; and (4) the authors propose a robust ecological explanation of the use of some parts of the spectral signal instead of blindly relying on the whole of the spectral information.

The only weak point in Jiang et al. (2013) is the absence of an explicit map of the uncertainty of the distribution. Such information is crucial for allowing reliable forecasting of species (or groups of species) distributions, by explicitly accounting for bias and/or simply vagueness deriving from the uncertainty of (1) field-sampled (Schmidtlein et al. 2012) and remotely-sensed (Lechner et al. 2012) data or (2) the modelling technique being adopted in spatial analysis, which is not just related to the arity used in spatial functions (Kühn & Dormann 2012).

This theoretical problem could be addressed by Jiang and colleagues in the future: a further challenge in seeing the unseen using remote sensing.

References

  1. Top of page
  2. Abstract
  3. References