Testing very high resolution Earth observation features as predictors of biodiversity surrogates at multiple scales for habitat quality change detection

Authors and Affiliations: 

Paola Mairotaa, Rocco Labadessaa, Francesco Lovergineb, Cristina Tarantinob Barbara Cafarellic,

aDepartment of Agro-Environmental and Territorial Sciences, University of Bari “Aldo Moro”, Via Orabona 4, 70125 Bari

bNational Research Council - Institute of Intelligent Systems for Automation (CNR-ISSIA), Via G. Amendola 122, 70126, Bari, Italy

bDepartment of Economy, University of Foggia, Largo Papa Giovanni Paolo II, 1 71100 Foggia

*Corresponding author. Phone: +39 080 5443021; Email: paola.mairota@uniba.uk


This research fits in the comparative habitat modelling effort of the BIO_SOS - Biodiversity Multisource Monitoring System: from Space TO Species project (FP7-SPA-2010-1-263435). Such an effort aims at exploring the potential of very high resolution (VHR) Earth observation (EO) features as proxies for habitat attributes in habitat quality change detection across spatial scales and biogeographical regions (Mairota et al., 2012). Here it is presented an investigation of the relations between VRH EO features and biodiversity surrogates (BS) representing local community structure of three taxonomic groups (plants, insects and birds) of a semi-natural grassland habitat of conservation concern in a Natura 2000 site in southern Italy. A hierarchical nested sampling strategy was adopted for field and EO data collection across three observation extent levels (landscape, patch, plot). Species abundance/dominance data were collected at the plot level (30 line 80m transects in March-September 2012). BS included overall species richness and diversity indices and analogous indices relevant to functional groups associated to habitat quality characteristics. The EO image used was closest to the conditions corresponding to high species richness for target organisms (DigitalGlobe™ Worldview-2, April 2011). Selected EO data included both spectral and statistical features (Normalized Difference Vegetation Index (NDVI), and three Grey Level Co-occurrence Matrixes (GLCM) metrics). The setting of the values of GLCM algorithms’ parameters (window size, quantization, shift and direction) likely to affect the scale even if the image spatial resolution is not altered, was guided by the perception limits of the target organisms and three window sizes were tested. Traditional regression techniques were used to evaluate the relations between BS and EO variables. Significant relations result between several BS, and particularly those relevant to functional groups, and EO variables which differ across extent levels. Moreover, within the same level, each BS can be associated to the same EO feature computed at different window sizes. It seems that landscape, patch and plot levels are respectively most appropriate when dealing with birds, plants and insects. This is further supported by the indication that the relations between BS relevant to plants and insects, might be affected by the window size used to compute EO proxies. While confirming the potential of VHR EO for habitat modelling aimed at habitat quality change detection, we verify that great care should be taken with regard to the scaling issues behind the assumptions related to the selection of the extent and the grain of the analysis in connection to the target taxonomic group. We also point out the value of functional groups related indices as BS of well preserved specific habitat types, as their quality is not necessarily represented by high values of overall species diversity.


Mairota P, Cafarelli B, Labadessa R, Leronni V, Blonda P, Lovergine F, Pasquariello G, Pradinho Honrado J, Lucas RM, Charnock R, Bailey M, Nagendra H, Niphadkar M, Kosmidou V (2012). Landscape pattern analysis. PART2. BIO SOS Biodiversity Multisource Monitoring System: from Space TO Species (BIO SOS) Deliverable D6.4, pp82 http://www.biosos.wur.nl/UK/Deliverables/