Click here to print this project

HyMedEcos-Gradients: Hyperspectral monitoring of Mediterranean ecosystems: gradients of land degradation

Start date: 07-04-2011 - End date: 07-04-2011

Aircraft:

Aircraft name: DO228 - NERC - ARSF

Airport: The data should be collected in the area between Castro Verde and Mértola, in the Alentejo region in southern Portugal (see uploaded file HyMedEcos_Grade_Flightdesign.pdf) for more details). The area is characterised by strong environmental gradients, including a sharp soil quality and land degradation gradient followed closely by a gradient of land abandonment and consequent scrub encroachment. Additionally, the study site includes the transition from a protected to a non-protected area with distinct land use regimes. The area close to Castro Verde is a Special Protection Area for birds (SPA), with a bespoke agri-environmental scheme aimed at maintaining the traditional cereal farming system to promote steppe bird populations, which are of international importance for some threatened species. However, this farming system also inadvertently promotes soil erosion, which is considered a major threat to the long-term sustainability of the cereal steppe of Castro Verde, which makes the area an interesting case study. At the same time, land abandonment can counteract some negative impacts on ecosystem services on the medium- to long-term time scale, e.g. based on natural succession in abandoned areas and related carbon sequestration. Additionally, the data collection in 2011, in a region partially overlapping with an existing dataset from 2006 (EUFAR project “STEPPEBIRD”), will also allow for a temporal analysis of e.g. vegetation cover patterns and steppe bird populations.

Project description

Project theme: Assessment of the use of hyperspectral remote sensing data for monitoring spatial and temporal transitions in Mediterranean ecosystems.

Science context: While remote sensing products are a reliable and often unique information source for the description and quantification of ecosystems and their services, existing data processing methods and algorithms still present many limitations when dealing with complex data such as hyperspectral imagery. In the proposed experiment, the area between Castro Verde and Mértola in southern Portugal, will be overflown for the acquisition of hyperspectral and laser scanning data at two spatial resolutions. The area is characterized by a strong natural gradient in terms of soil quality, vegetation and land use and offers multiple opportunities for ecosystem assessments with applied and methodological foci. Concurrent ground activities will be conducted in order to collect field spectroscopy data as well as several ecosystem parameters related with soil, vegetation and bird communities. A set of kernel-based approaches for combined qualitative-quantitative analyses will be developed for accurate ecosystem mapping and the quantification of environmental gradients. Methodological developments will be assessed with regard to the two spatial resolutions to help elucidate scale issues of future spaceborne hyperspectral analyses, such as for EnMAP imagery. Applied studies relating to soil erosion, biodiversity - vegetation and steppe bird - dynamics, estimates of forage crop quality and carbon storage potential will be carried out. Cross-disciplinary analyses are also foreseen. The analysis and outcomes of this study will be complemented by data and results from a 2006 campaign over the same region.

Measurements to be made by aircraft: HyMedEcos: The climatic regime of the Mediterranean region typically causes high water stress in vegetation during the long dry period, and also features intense rainstorms in periods when land cover is sparse and soils are prone to raindrop impact. These characteristics, together with the wide occurrence of shallow soils and a long and intense history of human use, make Mediterranean ecosystems highly susceptible to land degradation (Vogiatyakis et al., 2006). Land degradation is the result of complex interactions between many variables that show dynamic behaviour in space and time, and encompass processes such as soil erosion and deposition by water and wind, impoverishment of vegetation and desertification (De Jong & Epema, 2006). Hyperspectral remote sensing data, with its continuous spectral information, is an ideal data source for ecosystem and land degradation monitoring and assessment. HyMedEcos – Hyperspectral monitoring of Mediterranean Ecosystems - refers to a series of hyperspectral campaigns aimed at tackling specific research questions (e.g. degradation gradients, hydrological modelling, etc.). HyMedEcos-Gradients: The quantification of gradual processes, such as land degradation, land abandonment, or gradual changes in biophysical properties and in biodiversity patterns, are a fairly unexplored field, although they are part of intrinsically dynamic ecosystems. This experiment therefore aims to develop methods capable of using hyperspectral imagery, coupled with laser scanning data, for monitoring gradients in semi-natural areas, and ultimately understanding the causes and driving forces of such ecosystem changes. Also, potential problems and challenges of cross-scale transfer of algorithm developments to forthcoming satellite data (such as e.g. EnMAP), will be taken into account through the collection and analysis of data at two spatial scales. We propose to do this in a region in southern Portugal with a typical dry Mediterranean climate, along a gradient of land degradation and land abandoned. While abandoned agricultural lands are a typical consequence of land degradation, natural succession may enhance other ecosystem services such as carbon storage potential. Ground activities will be conducted to collect field spectroscopy data, as well as several ecosystem parameters related with soil, vegetation and bird communities. Additionally, a subsequent hyperspectral (HyMap) campaign will take place during the summer months which will allow analysis of seasonal / phenological gradients. Also, the outcomes will be complemented by data and results from a campaign over (partially) the same region, undertaken in 2006 (EUFAR “Steppebird” project), allowing for an assessment of temporal changes on vegetation and bird communities. Data analysis will make use of kernel-based approaches such as Support Vector Regression (Schölkopf et al., 2000) and Import Vector Machines (Zhu & Hastie, 2001) for fitting the various (field-collected) ecosystem parameters with the airborne data, and subsequent generation of quantitative maps (such as land cover fractions within mixed pixels, or maps including both vegetation types and biomass estimates), to be used for quantifying the environmental gradients. Cross-disciplinary analyses, e.g. the use of ecosystem biophysical parameters to model species habitat requirements and carbon storage potential, are also foreseen. The primary output of the project will be thematic digital maps of land cover, soil quality, biomass, steppe bird distribution and biodiversity and other geo-/biophysical variables that will serve as inputs for ecosystem and ecological models. Additionally, carbon storage potential in vegetation and soils will be estimated. The strategies and approaches that were applied to generate these maps, as well as the subsequent analyses and respective results will be published in international, peer-reviewed journals. References: De Jong, S. M. & Epema, G. F. 2006 Imaging spectrometry: basic principles and prospective applications, Kluwer Academic Publishers, pp.65-86. Schölkopf et al. 2000 Neural Computation, 12, 1207-1245; Vogiatyakis et al. 2006 Progress in Physical Geography, 30, 175-200; Zhu, J. & Hastie, T. 2001 Journal of Computational and Graphical Statistics, 1081-1088.

Season: Preferred dates: 01-04-2010 to 08-04-10 Acceptable dates: Anytime between late March and mid-April

Weather constraints: Preferably clear sky conditions, although up to 10% cloud cover is also acceptable.

Time constraints: Time of data acquisition would be preferably within 1.5 hours before local solar noon, though acceptable anytime 2.5 hours before or after local solar noon. In order to answer our research questions, the imagery should be collected during the first half of April, corresponding to the vegetative peak of both annual and perennial vegetation, as well as to the steppe birds’ breeding season. This timing is thus the most appropriate for quantifying all variables relating to vegetation and agricultural use, which are central to the processes being investigated.

Flights (number and patterns): We estimate a flight time of c. 2.5 hours duration. The study area should be covered by four (4) parallel flight lines at 4500m altitude, re-flown at 1500m (see uploaded file HyMedEcos_Gradients_Flightdesign.pdf for more details). Additionally, one (1) short flight line perpendicular to and crossing the previous lines should be flown at 4500m altitude, to allow for BRDF effects correction.

Instruments: Specim Eagle Hyperspectral Sensor Specim Hawk Hyperspectral Instrument Leica ALS50 (II) Part of the study area has already been flown (by NERC ARFS’s Dornier 228 in 2006 – EUFAR “Steppebird” project), and CASI-2 imagery and ALTM-3033 data has been collected.

Other constraints: None

Scientific contact

Name: Patrick HOSTERT

PI email: patrick.hostert@geo.hu-berlin.de

PI website: http://www.hu-berlin.de