PhD Defence Mr Abel Ramoelo
Department of Natural Resources
Title of defence
Savanna grass quality: remote sensing estimation from local to regional scale
Information about the distribution of grass foliar nitrogen (N) and phosphorus (P) is important to understand rangeland vitality and to facilitate effective management of wildlife and livestock. Grass N and P concentrations are direct indicators of rangeland quality, and they vary over space. Foliar N concentration is known to relate to the protein content. Protein is major nutrient requirement for the herbivores. On the other hand, foliar P concentration is a crucial requirement for reproduction and lactating animals. Successful estimation of foliar N and P concentrations could facilitate the computation of the key indicator of nutrient limitation, known as N: P ratio. Understanding the nutrient limitation could equally help the ecologists, farmers and resource manager to understand the feeding patterns, distribution and densities of herbivores both in protected and communal areas. Landscape view of the nutrient distribution and limitation as an interest to planners and managers could be achieved through remote sensing measurement and analysis. This study was undertaken in the north-eastern part of South Africa, in the savanna ecosystem. The study area was purposively selected as it covers the rangelands in the communal areas, Sabi Sands private game reserve and Kruger National Park (KNP). This study area offers experimental sites for foliar biochemical estimation, because of a pronounced contrast in soil fertility induced by various geological types, i.e. basalt and gabbro associated with high soil fertility and granite associated with low soil fertility. The main aim of the study was to develop and improve estimation of grass quality using remote sensing measurements from local to regional scale. The objectives were (1) to test water removed spectra for foliar N and P estimation as compared to the existing spectral techniques, (2) To estimate foliar N: P ratio using field spectroscopy or in situ hyperspectral data, (3) to investigate the applicability of the non-linear partial least square regression in integrating in situ hyperspectral remote sensing and environmental variables to estimate foliar N and P concentrations, and (4) to investigate the utility of the red edge band in RapidEye data for estimating foliar and canopy N at regional scale.
The success in remote sensing estimation of foliar biochemical is faced with various challenges which are yet to be addressed. The noted success is mainly based on the use of hyperspectral remote sensing. One of the challenges is the water absorption effects in short-wave infrared which mask weak or subtle foliar biochemical concentrations. At laboratory level using in situ hyperspectral, there were several attempts to address this challenge by drying leaf samples and measure the reflectance. The main problem associated with this approach was upscaling from laboratory (leaf) to canopy level. In Chapter 2, we proposed a technique that could be applied to minimize water absorption effects when estimating foliar biochemical using hyperspectral remote sensing data. This study was based on the greenhouse experiment. The aim of this study was to test the utility of water removed (WR) spectra in combination with partial least square regression (PLSR) and stepwise multiple linear regression (SMLR) to estimate foliar N and P concentrations, compared to spectral transformation techniques such as first derivative, continuum removal and log transformed spectra (Log(1/R)). The savanna grass species (Digitaria eriantha) was sown in the greenhouse. Spectral measurements were made using a spectrometer. D. eriantha was cut, dried and chemically analyzed for foliar N and P concentrations. WR spectra were determined by calculating the residual from the modelled leaf water spectra using the non-linear spectral matching technique and observed leaf spectra. It was concluded that the water removal technique could be a promising technique to minimize the perturbing effect of foliar water content when estimating grass nutrient concentrations. The performance of WR spectra was also evident in Chapter 3 on estimation of foliar N: P using in situ hyperspectral remote sensing data. The objective of Chapter 3 study was to investigate the utility of in situ hyperspectral remote sensing to estimate foliar N: P, in combination with PLSR. The results showed that foliar N: P can be highly estimated by water removed and continuum removed spectra. This was undertaken at field level using ASD FieldSpec 3®, which shows a potential of this technique across various remote sensing measurement levels. The crucial level at which this technique could be tested is at airborne hyperspectral level. Generally, Chapter 3 demonstrated that foliar N: P ratio could be estimated using remote sensing.
The other challenge in estimating nutrients is the diverse and heterogeneous nature of the savanna ecosystems, in terms of soil and plant moisture, soil nutrients, fire regime, grazing pressure, species composition and anthropogenic activities. This makes remote sensing estimation of foliar biochemical a challenging venture. In Chapter 4, we investigated the use of remote sensing and environmental variables to estimate foliar N and P concentration at field level, using ASD FieldSpec 3 measurements. The objective of the study was to test the performance of non-linear PLSR for predicting grass foliar N and P concentrations through integrating in situ hyperspectral remote sensing and environmental variables (climatic, edaphic and topographic), named as integrated modeling approach. The data consisted of: (i) in situ-measured hyperspectral spectra, ii) environmental variables and measured grass N and P concentrations. The hyperspectral variables included published starch, N and protein spectral absorption features, red edge position, narrow-band indices such as simple ratio (SR) and normalized difference vegetation index (NDVI). The results of the non-linear PLSR were compared to those of conventional linear PLSR. Integrating in situ hyperspectral and environmental variables yielded highest foliar N and P estimation accuracy using non-linear PLSR as compared to using remote sensing variables only, and conventional PLSR. This study demonstrated the possibility to use integrated modeling approach and non-linear PLSR in estimating foliar N and P concentrations.
In Chapter 5, we focused on the regional estimation of foliar and canopy N using spaceborne remote sensing measurements, which is new for the savanna ecosystems. The objective of this study was to estimate and map foliar and canopy Nitrogen (N) at a regional scale using a recent high resolution spaceborne multispectral sensor (i.e. RapidEye). RapidEye sensor contains five spectral bands in the visible-to-near infrared (VNIR), including a red-edge band centred at 710 nm. The importance of the red-edge band for estimating foliar chlorophyll and N concentrations has been demonstrated in many previous studies, mostly using in situ hyperspectral remote sensing data. The utility of the red-edge band of the RapidEye sensor for estimating grass N was investigated in this study. A two-step approach was adopted involving (i) vegetation indices and (ii) integration of vegetation indices and environmental or ancillary variables using a SMLR and non-linear PLSR. To ensure that grass N estimation is not comprised by biomass variability, the field work was undertaken when the grass has reached maximum productivity. The model involving the simple ratio (SR) index (R805/R710) defined as SR54, altitude and the interaction between SR54 and altitude (SR54*altitude) provided the highest accuracy for canopy N estimation, while the non-linear PLSR yielded the highest foliar N concentration estimation accuracy through integration of remote sensing (SR54) and environmental variables. The spatial pattern of foliar N concentrations thus mapped, corroborated with the soil fertility gradient induced by the geological parent material.
In conclusion, this study demonstrated that water removed spectra can be used to improve estimation of foliar biochemical concentrations. The study also revealed that foliar N: P ratio can be estimated using field spectroscopy or in situ hyperspectral remote sensing. The integrated modeling approach using non-linear PLSR showed to improve the estimation of foliar biochemical concentrations. For the first time, the regional estimation and mapping of savanna grass N was successfully done using the red-edge band embedded in the RapidEye multispectral sensor. The estimation and mapping of grass nitrogen could be used for understanding feeding patterns and changes in densities of wild and livestock (herbivores) at a regional scale.
Abel Ramoelo was born in Muduluni village, South Africa on the 02 November 1980. He completed his primary level school in 1993 at Madaheni primary school, Makhitha village, South Africa. In 1998, he completed matric certificate from Luvhivhini secondary school, Maebani village. Given his passion on mathematics, physical sciences (Physics and chemistry), biology and geography, he pursued a Bachelor of Environmental Science degree at University of Venda (South Africa). He graduated top of the school of environmental sciences in 2003, scooping several awards. In 2003, he registered an honours degree in environmental science, at University of Venda. During his honours study, he developed a passion on using remote sensing and geographical information system tools to address and model environmental related issues. In May 2004 he started working at Human Science Research Council (HSRC) as GIS intern. After 3 months, August 2004, He joined the Council for Scientific and Industrial Research (CSIR) as junior researcher (remote sensing and GIS). After a year, he started a Master of Science in Geo-information Science and Earth Observation for Environmental Management and Modelling in four European universities (University of Southampton-UK, Lund University-Sweden, University for Warsaw-Poland, University of Twente-ITC, The Netherlands), funded by Erasmus Mundus Scholarship. The title of the thesis was “An innovative method to map land cover changes at country level utilizing hyper-temporal satellite images”. Since he is an aspiring scientist, six months after acquiring his MSc., he took-up a PhD position under the supervision of prof.dr. Andrew Skidmore, which resulted in this thesis. In 2011, he was awarded “Emerging Researcher Award” during the “Excellence Award” for the Natural Resource and Environment Unit of CSIR, South Africa.
Ramoelo, A., Skidmore, A.K. (Promotor), Schlerf, M. (assistant promotor) and Heitkönig, I.M.A. (assistant promotor) (2012) Savanna grass quality : remote sensing estimation from local to regional scale. PhD thesis University of Twente, Summaries in English and Dutch. ITC Dissertation 207, ISBN: 978-90-6164-331-9.
|Event starts:||Thursday 24 May 2012 at 14:30|
|Venue:||UT Waaier room 4|
|City where event takes place:||Enschede|
|Country where event takes place:||Netherlands|