|
Current research projects within the domain of the Earth observation group include:
Hyperspectral integrated modelling approach
Department of Science and Technology SumbandilaSat project
Cooperation Fund projects: CSIR with University of the Witwatersrand (Wits) and the University of Fort Hare
Potential of ground-based Lidar for system structural assessment project
Description of vegetation systems using hyperspectral data inputs project
Spectral structural interactions project
SEABOAT (South East Atlantic Biogeochemical Ocean Adjustments and Thresholds) Programme
Satellite remote sensing products development project
Frontier Programme for Mariculture Research & Development
Remote sensing of land degradation in the Inkomati catchment
Time-series analysis of hyper-temporal satellite data using high performance computing (HiTempo) project
Observations of land quality
Greenhouse gas inventory
Geo-information services
Integrated hyperspectral system modelling
Regional Centre for Spatial Analysis and Modelling (ReCSAM), Limpopo University, South Africa
Hyperspectral integrated modelling approach
This project investigates current remote sensing models used to gauge ecosystem state, and eventual integration of such models with existing empirical modelling strategies. This entails (i) a thorough review of existing models, (ii) evaluation of model input parameters to gauge which can be sensed using remote sensors and which parameters require in situ measurements, (iii) testing and validation of “remote-sensing-only” models, and (iv) robust extension of such models to ensure multi-temporal validity. Whereas a crop system model mainly involves growing season parameterization, an ecosystem approach would require a more extensive multi-temporal approach. Model evolution and parameterization are required throughout a yearly cycle, resulting in a dynamic approach to system state description. It therefore is crucial to develop strategies towards integrated modelling, such that researchers are able to incorporate hyperspectral-derived inputs along with additional parameters to model systems.
The proposal is closely linked to efforts of Prof. Pol Coppin (Katholieke Universiteit Leuven; M3-BIORES group) which involves integrated modelling of production systems. These studies also form part as application development efforts for the planned MSMISat satellite, a South African–Belgium collaborative spaceborne platform that will include a hyperspectral sensor (200 wavelength bands; 400-2500nm; 15m spatial resolution) as part of the payload and has a planned launch date of end-2009. It also is aligned to other research initiatives of various CSIR research groups:
- Ecophysiology group: Improved spatial and temporal prediction of water use efficiency of land surfaces in South Africa. This initiative aims to estimate water use efficiency using both in situ and remote sensing inputs, with a limited focus on hyperspectral component. The link to this effort lies in the facts that both projects will approach this from an “integrated sensing inputs” modelling strategy and that spatially explicit modelling is a goal of both initiatives. It is envisaged that the two projects will provide learning to each other in these aspects.
- A Young Researchers Fund project that links the hyperspectral parameters to in situ parameters for system state modelling, as well as providing opportunities for international exposure, e.g. conferences, workshops, and workgroups.
The hyperspectral integrated modelling approach project, on the other hand, looks at hyperspectral-specific approaches to forest system state (as defined by foliar chemistry description, e.g., chlorophyll a & b, Ca, Al, Mg, Mn, P, K, N, lignin, and cellulose) and state evolution modelling, with objectives of defining analytical approaches and modelling strategies with hyperspectral data in mind. Forestry was chosen due to its relatively uniform environment and growth stock (e.g., clonal material). It should be noted that this project constitutes only the first step towards integration of remote sensing and in situ parameters for system state modelling. The final goal will likely only be achieved after sustained research over a number of years.
For more information, contact Dr Renaud Mathieu
SumbandilaSat project
This project is initiated by the Department of Science and Technology, and in collaboration with the CSIR Defence Peace Safety and Security and Department of Water Affairs and Forestry. The project has as key objectives:
- evaluate the potential of the red edge wavelength band (690 - 730 nm) on SumbandilaSat for assessing vegetation vigour and state (chlorophyll content and water stress) with forestry as use case,
- evaluate the potential of the xanthophyll wavelength band (520 - 540 nm) on SumbandilaSat for assessing photosynthetic activity and vegetation state (e.g., described by growth) with forestry as use case,
- determine the degree with which leaf spectral responses to water stress in Acacia mearnsii from contrasting habitats (riparian and non-riparian areas) differ from each other using the red-edge and xanthophylls bands of the SumbandilaSat sensor. Focus potentially will expand to all bands, but special attention will be given to the two experimental bands, and
- determine the degree with which SumbandilaSat imagery can be used to assess the extent and fragmentation of coastal forests with specific reference to land use and fragmentation.
Spectral response for the defined bands will be simulated using resampled ASD spectroradiometer data and existing spaceborne sensors (e.g., Ikonos) and validated using actual SumbandilaSat imagery as the satellite comes online; critical research for instance pertains to the degree with which especially red edge migration can be assessed using such a multi-spectral instrument.
This research will be performed in the context of a larger consortium, namely the CSIR Natural Resources and the Environment, CSIR Forest and Forest Products Unit, and University of Kwazulu-Natal (red edge and xanthophylls; forest state assessment), CSIR Defence, Peace, Safety and Security (DPSS) (Acacia mearnsii water use assessment), and the Department of Water Affairs and Forestry (coastal forest assessment). It also forms part of a larger initiative whereby study sites will conform to those selected in the effort of the CSIR Meraka Institute(identification of priority areas and core validation sites) and eventual results will rely heavily on the calibration of SumbandilaSat imagery (DPSS proposal related to development of a calibration protocol for SumbandilaSat).
For more information, contact Dr Renaud Mathieu
Cooperation Fund projects: CSIR with University of the Witwatersrand (Wits) and the University of Fort Hare
University of the Witwatersrand (Wits) (collaborator: Dr. Barend Erasmus) The use of woody vegetation in rural savannas is well-documented. However, most of these studies rely on conventional field-ecology techniques to quantify standing woody biomass. Although this is sufficient for most purposes, it does not take into account the spatial variability of standing woody biomass. Given the increasing importance of sustainability due to the effect of global change on vulnerable rural communities, it is imperative that the access to such resources be quantified for all communities in a landscape. Such data will also show the historical use patterns, and inform future sustainability scenarios if use practices are changed or not.
The specific objectives of this study are to
- determine the temporal changes in woody biomass cover, using available and appropriate remotely sensed imagery,
- determine the current extent of and the amount of woody biomass in rural rangelands using high spatial resolution imagery,
- gauge the vegetation structure of the standing woody biomass, and
- develop and transfer skills in the analysis of remotely sensed imagery.
University of Fort Hare
(collaborator: Carryll Tyson): The stated outputs and benefits of this project are related to the short-term Honours-level projects, capacity building, and knowledge dissemination:
- development of a one-week course for Honours students-capacity building (Dr Renaud Mathieu; hyperspectral and lidar remote sensing),
- orientation survey data for developing an appropriate remote sensing monitoring system for Eastern Cape reserves,
- ground-based spectral signatures of Eastern Cape vegetation, including individual species, as well as mixed-species signatures,
- resampled spectral signatures and measures of calibration success to various satellite sensors, and
Development of cost-effective remote sensing techniques for vegetation state/stress monitoring.
For more information, contact Dr Renaud Mathieu
Potential of ground-based Lidar for system structural assessment project
(The project is executed in collaboration with CSIR Built Environment and the CSIR National Laser Centre Structural assessment of various man-made and natural environments is becoming increasingly important. Structural integrity, personal safety, and building progress can be considered key issues in a built environment, while ecosystem state, spectral-structural interactions, energy exchange interface, and vegetation inventory are crucial issues in the vegetation (natural) domain. One of the most exciting and novel developments in such structural measurement efforts is that of light detection and raging (lidar), a system that measures the return-trip of a laser beam in order to collect range information. The technical ability and human skills in South Africa is severely lacking in this field of technology, even though the CSIR National Laser Centre (NLC) has developed an integrated platform and system for atmospheric lidar measurements, based on slightly different principles. The goal of this pilot SRP project therefore is to (i) perform an in-depth literature review of such structural measurement platforms, (ii) define the needs and requirements of the Natural Resources and Environment, NLC, and CSIR Built Environment units in terms of structural system assessment, (iii) assess the potential of integrating such a structural system into the existing atmospheric system, (iv) identify key national and international collaborators, (v) host a workshop designed to address items (ii), (iii), and (iv) in an external context, and (vi) to hold an internal workshop and develop a draft SRP Full Project proposal on this topic for the 2008 financial year and onwards. This project has significant potential benefits to the CSIR and the country as a whole. Not only will technical capacity and skills be developed, but the potential for system and structural algorithm patents also exist. The project also links with existing initiatives within the CSIR, e.g., the NRE theme of linked natural systems, the recently developed research strategy of the Ecosystems Earth Observation group within NRE, and the international push towards integrated (in situ and remote sensing inputs) system modelling. Structural assessment will likely constitute an increasingly important component in such approaches.
For more information, contact Dr Renaud Mathieu
Description of vegetation systems using hyperspectral data inputs (CSIR-Mondi Business Paper joint project)
The main goal of this research was to investigate the species differentiation on the one hand, and specific site-age interactions, on the other hand, that exist in managed, homogeneous, even-aged Eucalyptus plantations in Kwazulu-Natal, South Africa using hyperspectral data. The study area is located approximately 45 km north of Pietermaritzburg and is characterised by undulating topography ranging from 800 m to 1200 m above mean sea level. Average daily temperatures range from 11 to 26 degrees Celsius with mean annual precipitation of less than 1200 mm. Eucalyptus stands used in this study were aged between four and ten years old. CASI hyperspectral (36 reflectance bands; 426-952 nm spectral range; 1 m spatial resolution) data were acquired during the spring (end-October) of 2005. The reflectance-processed hyperspectral data were used to (i) differentiate between various Eucalyptus species, a traditionally homogenous genus, (ii) discriminate between age classes within a species, and (iii) determine the degree with which different site quality classes can be defined based on spectral responses. The degree to which indicators derived from hyperspectral data, e.g., derivatives, integrals, and ratios, could be used to describe the nature of and differences among various species within the Eucalyptus genus, established a managerial base-line in the first instance. This addressed the question of distinct spectral differences between species in this largely homogenous group.
The second question pertained to the effect that age-class differences have on spectral response, and more specifically, the spectral variability found in homogenous even-aged, monoculture forest stands. The last topic aimed to evaluate age-site interactions and investigate to what extent various site classes can be determined using a hyperspectral approach. Specific implications relate to the use and interpretation of widely accepted forest state and management indicators, and how age-site interactions affect indicator outcomes. These issues are critical to our understanding of natural resource assessment using data sources with hyper-data characteristics, such as hyperspectral sensors.
For more information, contact Dr Renaud Mathieu
Spectral structural interactions project (jointly between Earth Observation and Forest Assessment research groups)
The main goal of this research is to investigate the structural-spectral interactions that exist in managed, homogeneous, even-aged Eucalyptus plantations in Kwazulu-Natal, South Africa. Eucalypts plantations used in this study range between four and ten years old. Small-footprint light detection and ranging (lidar; ALTM 3033 two-return laser system; 0.2 mrad footprint, 33 kHz pulse rate) and ASTER and Ikonos multi-spectral data were used in this study. Lidar and multi-spectral data were collected during the spring season of 2006. Segmentation (multi-resolution/hierarchical algorithm) of a lidar-derived canopy height model was performed in order to derive areas with homogeneous structural properties. Structural characterisation of segments were performed by comparing first and second return distributions, which mimic the integrated response found in waveform lidar sensors. The reflectance-processed multi-spectral data were used to (i) characterise the structural differences among different sets of homogeneous segments and (ii) determine the contribution of structural variability to spectral responses. The degree with which indicators derived from multi-spectral data, e.g., indices, could be used to describe the structural nature of and differences among various segment classes, was important in the first instance. This addresses the question of complimentarity between spectral and structural data sets in such forest environments. The second question pertains to the spectral variability found in structurally homogenous even-aged, mono-culture forest stands. Specific implications relate to the use and interpretation of widely accepted forest state indicators, and how structure affects indicator outcomes.
For more information, contact Dr Renaud Mathieu
SEABOAT (South East Atlantic Biogeochemical Ocean Adjustments and Thresholds) Programme
This project is the core of a four year programme to put in place a Climate Change Centred Oceanographic R&D capability in the CSIR that will be both globally recognised and make a significant contribution to the ACCESS research and education strategy.
The focus is primarily on ocean biogeochemistry: in the Benguela, as part of the South Eastern Atlantic basin responses to climate change, and in the southern ocean as a sub Antarctic CO2 sink zone. This project aims to examine the integrated response of the SE Atlantic to global warming by focussing on understanding the significant re-organization of its biogeochemistry since the last glacial period. Operationally, it integrates the three key activities required for earth system science: global scale models, remote sensing and in situ observations.
For more information, contact Dr Steward Bernard.
Satellite remote sensing products development project
This project aims to develop capacity in key areas that are fundamental for ultimately securing a multi-institutional marine Earth Observing centre. The core objective of the project is to provide the remote sensing component of a pre-operational marine observation system. The overall objectives of the project are: the development of ocean colour, sea surface temperature (SST) and other marine products for operational use in the southern African region, the provision of remote sensing and integrated products for the marine scientific community and resource managers, and the development of capacity in the region though student projects and training initiatives.
For more information, contact Dr Steward Bernard
Frontier programme for mariculture research & development: Development of a harmful algal bloom monitoring system in the southern benguela using multi-platform moorings and satellites.
The primary focus of the project is the provision of near real-time harmful algal bloom (HAB) related data from multi-sensor moorings and multi-platform satellite sensors in the southern Benguela. The principal products will consist of algal biomass and assemblage descriptors, physical data relating to water column structure and movement, and operational risk indices developed through the course of the project. The project will make use of ocean colour data from the MERIS and MODIS sensors; sea surface temperature data from the AVHRR and MSG sensors; and multi-parameter mooring data from the current BCLME HAB monitoring buoy on the Namaqua shelf. The project is executed in collaboration with the Department of Science and Technology, and the Department of Environmental Affairs and Tourism.
For more information, contact Dr Steward Bernard.
Remote sensing of land degradation in the Inkomati catchment
This activity is a component of a larger project titled ‘The flow of benefits to people provided by ecosystems at multiple scales: a spatial and economic assessment’ (short title: Mapping the flow of benefits to people from ecosystems). The activity aims at producing a land degradation spatial layer for the Inkomati catchment. Additional aims are to:
- find and define a basis for identifying reasonably homogenous strata in the catchment, and
- find a way to scale relative productivity of those strata to enable:
- scaling of strata relative to each other
- assessment of degrees of degradation (loss of productivity) within strata.
Summertime Landsat TM and ETM+ images covering Inkomati catchment are being used. Four scenes are needed to cover the catchment – WRS 168-77, 168-78, 169-77, and 169-78. The catchment is subset from the images using a vector file. Following geometric correction, the images undergo atmospheric correction using ATCOR 2 software. Thereafter, computation of the Normalised Difference Vegetation Index (NDVI) and albedo values enable the identification of possible degradation areas, which are then verified in the field. Vegetation polygons from the SANBI vegetation map of the catchment are used as strata. Depending on location (and consequently the different perturbations like human harvesting, fire, grazing, moisture availability), different vegetation polygons have different vigour and degradation indications. For example, on a 14 December 2006 image, two vegetation polygons representing sweet arid basalt lowveld and eastern highveld grassland vegetation types, respectively, show the following NDVI profiles, indicating the former to be the healthier and more vigorous of the two:
For more information, contact Dr Christopher Munyati.
Time-series analysis of hyper-temporal satellite data using high performance computing (HiTempo) project
Sustainable development requires spatially-explicit information on environmental change and long-term satellite data archives (e.g. MODIS, AVHRR) are the only sources of such information at regional scales. Existing methods for detecting environmental change from satellite data sets are insufficient, since data volumes are prohibitively large (>10 Terabytes), conventional image data structures cause a massive Input/Output bottleneck which impedes effective time-series analysis and its computational intensity far exceeds the capability of single workstations. As a result a number of potentially very useful signal processing methods have not yet been applied to this problem.
The purpose of this project is to unlock the wealth of information in the large data sets of hyper-temporal (daily, 10-daily), coarse resolution satellite images using advanced time-series analysis in a High Performance Computing (HPC) environment.
Objectives: (i) establish the CSIR as a world leader in time series analysis of satellite data, (ii) develop and optimize an HPC time-series analysis system (the “HiTempo” system), including a novel per-pixel, time-sequential data structure, (iii) compare and refine various time-series analysis methods (e.g. Hidden Markov model, Singular Spectrum analysis and wavelet analysis) for characterising patterns, trends and changes in earth surface properties by assessing the accuracy of their outputs (change or trend maps) with independent data (e.g. Landsat images).
This project under the lead of the CSIR Meraka Institute (contact: Konrad Wessels) is conducted in cooperation with other CSIR research groups. The contribution of the NRE Earth observation group (contact: Melanie Vogel) is to build up data sets with information about expected and observed vegetation change patterns on the high resolution satellite imagery scale (Landsat TM, SPOT5). The results will be used as input and validation base for the processing conducted by the Meraka team on the hyper temporal-low spatial resolution MODIS scale.
This basic information will be derived from field explorations and workshops under the lead of the The Processes and dynamics group (contact: Sally Archibald).
The research will be supporting the South African Earth Observation System (SAEOS) and Global Earth Observation System of Systems (GEOSS). It is envisaged that the HiTempo system will deliver information (e.g. maps depicting environmental changes) to inform government decision making and policy development. This research will ultimately allow us to interpret the present in the context of the past, in near-real time.
 |
Figure 1. Overview of proposed project indicating activities and responsibilities (in colours). |
Observations of land quality
An integrated system for monitoring land cover and land condition: Land cover and condition are fundamental inputs to sustainable development, prioritization of development interventions, conservation planning and environmental monitoring.
The aim of this project is to assess land cover types and land cover condition in different vegetation and land use systems with greater accuracy and lower cost than current methods by combining remote sensing, in situ observations and models.
Core research areas are the Kruger National Park (KNP) and adjacent areas in Mpumalanga and Northern Cape. The choice of these sites allows for analysis at savanna and grassland lowveld vegetation as well as observations at the savanna-Nama karoo ecotone.
|
The biomes of South Africa showing the two sites for testing the use of FAPAR and phenology indices. The eastern lowveld including savannas and grassland covers a range of altitudes, rainfalls, soils, and land management. The grassland-karoo boundary will give an indication of how well the algorithms work in arid, variable environments.
Source: Vegetation Map of South Africa, SANBI |
This project is a co-operation between the following research CSIR groups:
The Processes and dynamics group (contact: Dr Graham von Maltitz) contributes with extensive field data, such as field inventory vegetation data, locally measured biophysical parameters (Fraction of Photosynthetic Active Radiation FPAR, Leaf Area Index LAI) to assess land cover productivity and state of degradation in natural environments. A variety of contextual information and modelling approaches are also integrated in the project to provide accurate classification and attribute information.
The Meraka institute (Contact: Konrad Wessels) focuses on the analysis of hypertemporal (1 to 10 days repetition) low spatial resolution (250m to 1km) remote sensing data (MODIS) to distinguish land cover types by phenological descriptors on a large spatial scale.
The contribution of the Earth observation group (contact: Dr Melanie Vogel) is the assessment of savanna vegetation and vegetation state. As indicator for savanna condition the tree coverage will be assessed on different spatial and temporal scales. Very high resolution satellite imagery (IKONOS) will be used for accurate local tree cover assessment. The results will be used to train high resolution (SPOT5, Landsat TM) classificators, which shall be used to validate the results of the MODIS derived classifications from the Meraka team. Additionally the Landsat data will be used to explore new methods for land cover classification and change analysis.
The novel techniques involved are: the derivation of sensor-independent vegetation functional information from ‘hypertemporal’ (i.e. daily to 10 daily) remote sensing; the extraction of features from high-spatial resolution datasets using textural analysis; and the fusion of data sources with dissimilar spatial, spectral and temporal properties.
The outcome will be substantially improved procedures for land cover and land condition mapping for southern Africa.
For more information, contact Dr Melanie Vogel).
Greenhouse gas inventory
The CSIR is the specialist consultant for the Greenhouse Gas Inventory for South Africa project that is executed by the Department of Environmental Affairs and Tourism (DEAT) and the University of Cape Town. The Earth observation research group is specifically working in the agriculture, forestry and other land uses (AFOLU) sector and will determine whether there are any changes in the amount of greenhouse gas emissions for this sector. The intergovernmental panel on climate change (IPCC) coordinates the reporting structure and methodologies from all countries, and guidelines are available from their website: http://www.ipcc.ch/. The project will run over a year for the inventory, followed by another year of training to DEAT officials.
For more information, contact Heidi van Deventer
Geo-information services
The CSIR is in the process of setting up a geo-spatial data infrastructure (GSDI) that will enhance the accessibility of spatial data to all personnel of the CSIR. Options of open source map servers and inter-operatibility software to propriety geographical information systems (GIS) and remote sensing (RS) packages are being investigated.
The vision is to centralise storage of spatial data being used and curated by the CSIR, and enable access and use of the spatial data sets by most of the CSIR personnel that can benefit from it. A business plan, business requirement specifications (BRS) and user requirement specifications (URS) would document Phase II of the development of the CSIR Geoportal, in line with CoGIS. CoGIS aims to establish cooperation between Pretoria-based institutions to centralise spatial data and sharing access through portals.
More information about the CSIR Geoportal is available from http://gsdi.csir.co.za.
For more information on the service offered by the CSIR, contact Heidi van Deventer.
Current projects include:
Integrated system for monitoring land cover and land condition in urban and rural areas.
- Tree cover mapping using high resolution multispectral IKONOS imagery
- Co-operative project with Wits (Barend Erasmus/Patrick Dwyer) and Meraka institute (Dr. Konrad Wessels) looking at mapping/monitoring woody and herbaceous biomass in and around the KNP, using remote sensing and field surveys.
- SumbandilaSat project: Red-edge and xanthophyll - with focuses on the evaluation of vegetation specific applications for the SumbandilaSat red-edge and xanthophyll wavelength ranges.
For more information, contact Russell Main.
Integrated hyperspectral system modelling
This project investigates the usefulness of using hyperspectral data as inputs into integrated forest system modelling. It is a co-operative project with Mondi - focussing on using forestry foliar chemistry traits as an indicator of system state.
For more information, contact Russell Main.
|
