Monitoring of forest extent and structure at national scale is essential task in the context of climate change for conserving biodiversity, developing national forestry inventories and projecting the future of terrestrial carbon sinks. Multi-temporal monitoring of woody canopy cover is needed to achieve the aforementioned task. The current study sought to demonstrate the potential of remote sensing sensors in generating multi-temporal woody canopy cover products for South Africa. The combination of freely available high spatial and temporal resolution from sentinel-2 (multispectral) and sentinel-1 (Synthetic Aperture Radar) and with canopy cover samples from the Global Ecosystems Dynamics Investigation (GEDI) enables multi-temporal assessment of woody canopy cover. The use of Random Forest (RF) machine learning algorithm to model woody structure over a period of five years (2019 – 2023) revealed changes in some woody biomes of South Africa, due to fragmentation, bush encroachment and degradation. These products provide national information on woody canopy cover on a consistent basis. Furthermore, the products provide a better understanding on the changes, distribution and extent of woody ecosystems in South Africa.

Forage nitrogen (N) and biomass measure rangeland quality and quantity. The N and biomass interactions (N*Biomass) are critical in understanding rangeland conditions, especially when deriving grazing or browse-carrying capacity indicators. Conventional methods have been used to measure N and biomass; their main shortcoming is labour intensiveness and time-consuming, especially when covering larger areas. Using remote sensing, limited studies focused on estimating the N*Biomass in the context of understanding rangeland conditions. This study, however, provides significant insights. It aims to understand the N*Biomass using vegetation indices and spectral bands in the Golden Gate National Park, South Africa. Stepwise Multiple Linear Regression (SMLR) and Random Forest modelling were performed to develop N*Biomass predictive models based on vegetation indices such as leaf area index (LAI), canopy chlorophyll content, and Sentinel-2’s spectral bands. The parsimonious SMLR model consisted of four variables, including LAI, spectral band 2 (Blue), 4 (Red) and 8a (Vegetation Red Edge), explaining 63% of N*biomass, whilst RF explained 41%. The study's findings indicate that the N*Biomass is explained by variables associated with the structure and the quality of the vegetation, providing valuable insights for future research and practical applications.

A comprehensive comparison and integration of retrieval methods is needed for accurately estimating vegetation biophysical variables such as leaf area index (LAI) over a multispecies grass canopy. This study tested the partial least squares regression (PLSR) and kernel ridge regression (KRR) for inversion of a radiative transfer model (RTM) to retrieve grass LAI in the Golden Gate Highlands National Park of South Africa during peak productivity. Furthermore, we constrained the inversion process using Active Learning techniques. Results show the most accurate LAI retrieval by KRR (over 34 sampled grass species) had a normalized root mean squared error of 19.28%. These findings have significant implications for the development of transferable rangeland monitoring systems in protected mountainous regions.

African rangelands are threatened by anthropogenic land-use activities, adverse climate phenomena such as droughts, and poor land management. These undermine their capacity to support various fauna and flora, provide ecosystem services, and sustain livestock agriculture, i.e., a key economic activity in Africa. Therefore, preserving the integrity of African rangelands is critical for addressing African food security challenges. Using multi-sensor Earth observation data and Random Forest classifier, this study characterized the spatial distribution of African rangelands, to support grazing and browsing capacity modelling, assessment of rangeland changes, and rangeland management policy development and decision making. The results show that rangelands could be characterised with good accuracies exceeding 70% in most AfriCultuReS pilot countries using the high-resolution land cover map and MCD12Q1 products as training and validation data. The spatial distribution maps can be used as masks that would aid accurate monitoring of rangeland health, productivity, phenology and changes.

The project “South African Land Degradation Monitor” (SALDi) contributes to the German-South African Programme “Science Partnerships for the Adaptation to Complex Earth System Processes in the Region of Southern Africa” (SPACES) by addressing the dynamics and functioning of multi-use landscapes with respect to land use, land cover change, water fluxes, and implications for habitats and ecosystem services. We are utilizing time series information from Sentinel-1 and Sentinel-2 of the European Space Agency (ESA) Copernicus program. The synergetic combination of both satellites hold large potential as both systems measure different properties of the earth's surface. This study analyses the feasibility of detecting breakpoints and land surface changes over time within the six SALDi study sites. The overarching aim is to link the findings of the time series analysis to different land degradation phenomena.

The overarching goal of SALDi (South African Land Degradation MonItor) is to implement novel, adaptive, and sustainable tools for assessing land degradation in multi-use landscapes in South Africa. This presentation demonstrates results from hyper-temporal Sentinel-1 and -2 timeseries concerning woody cover mapping in complex savanna systems, invasive slangbos bush encroachment in grassland areas and regional soil moisture retrievals. Validation has been performed by cross-comparisons, field trips and permanently installed soil moisture networks.

Herbaceous biomass is an important indicator of rangeland quantity. Grasslands cover vast areas in South Africa. It supports livestock production which is crucial for livelihoods and biodiversity conservation including ecotourism and conservation purposes. Grasslands, especially the mountainous ones, are threatened by the number of factors including global environmental changes. Latter changes include climate change, overgrazing, a proliferation of invasive species, unmanaged fires, and other land-use changes. There is a need to continuously monitor grasslands and remote sensing is an ideal, particularly in mountainous areas. The objective of the study was to test and compare the applicability of Sentinel-2 and Landsat 8 data acquired before and after fire occurrence in estimating biomass in the Golden Gate Highland National Park (GGHNP). Random forest model was used to predict biomass using Sentinel-2 and Landsat 8 reflectance data. Sentinel-2 and Landsat 8 reflectance data explained over 80% of biomass variation in the mountainous areas. The results indicate that both Sentinel 2 and Landsat 8 provide useful information for grass biomass estimation in mountainous environments.

The tree Acacia mearnsii is native to south-eastern Australia but has become an aggressive invader in many countries. In South Africa, it is a significant threat to the conservation of biomes. Detecting and mapping its early invasion is critical. The current ground-based methods to map A. mearnsii are accurate but are neither economical nor practical. Remote sensing (RS) provides accurate and repeatable spatial information on tree species. The potential of RS technology to map A. mearnsii distributions remains poorly understood, mainly due to a lack of knowledge on the spectral properties of A. mearnsii relative to co-occurring native plants. We investigated the spectral uniqueness of A. mearnsii compared to co-occurring native plant species within the South African landscape. We explored full-range (400-2500 nm), leaf and canopy hyperspectral reflectance of the species. The spectral reflectance was collected biweekly from December 23, 2016 and May 31, 2017. We conducted a time series analysis, to assess the effect of seasonality on species discrimination. For comparison, two classification models were employed: parametric interval extended canonical variate discriminant (iECVA-DA) and nonparametric random forest discriminant classifiers (RF-DA). The results of this paper suggest that phenology plays a crucial role in discriminating between A. mearnsii and sampled species. The RF classifier discriminated A. mearnsii with slightly higher accuracies (from 92% to 100%) when compared with the iECVA-DA (from 85% to 93%). The study showed the potential of RS to discriminate between A. mearnsii and co-occurring plant species.

Over-grazing, bush encroachment and alien species invasion are having negative impacts on livestock production in the rangelands of Sub-Saharan Africa, thus threatening rural livelihoods in the region. Leaf area index (LAI) can be used as an indicator or an early warning signal of changes in rangeland conditions. The aims of this study were to (i) assess the accuracy of the existing MODIS LAI product for key land cover classes and biomes in South Africa and (ii) assess the accuracy of LAI retrieved by the inversion of PROSAIL radiative transfer model. The MODIS LAI estimates were particularly poor for grassland and Karoo biomes for the period under investigation (November 2012 to February 2013). More accurate estimates of LAI values (RMSE = 0.9) across three major biomes were generated by inverting model outputs from PROSAIL radiative transfer model on MODIS imagery.

The performance of PROSPECT-5 radiative transfer model for predicting leaf chlorophyll from reflectance measurements made with the Analytical Spectral Device (ASD) spectrometer was investigated using numerical inversion techniques. The reflectance data of various spectral regions in the visible to shortwave infrared (SWIR) were assessed i.e. the full range (400-2500 nm), VNIR (400-1060 nm), visible (400-700 nm), red (600-700 nm), red-red edge (600-760 nm) and red-edge (670-760 nm). Among the spectral regions, the red-edge region yielded the lowest root mean square error of prediction (RMSEP =10.3 μg/cm2) for a variety of plants including crops and wild plants (n = 463). It is therefore recommended that inversion of radiative transfer models to retrieve leaf chlorophyll content be limited to the red-edge region.

The ability of various ecosystems to perform vital functions such as biodiversity production, and water, energy and nutrient cycling depends on the ecosystem state, i.e. health. Ecosystem state assessment has been a topic of intense research, but has reached a point at which accurate large scale (e.g. regional to global scale) modelling and monitoring are hindered by limitations in conventional assessment methods such as direct field sampling, modelling from environmental drivers such as temperature, precipitation and available nutrients, and modelling from remote sensing data. The Ecosystem-Earth Observation (Eco-EO) research group at the Council for Scientific and Industrial Research (CSIR), South Africa has highlighted the need in remote sensing research for an integrated sensing approach at the systems level. This perspective is based on the assumption that a modelling approach that exploits the strength of the various techniques (in situ environmental variables, direct field observation and remote sensing data) could potentially improve the assessment of ecosystem state at various geographic scales. In this light, the Eco-EO research group has embarked on an agro-system state assessment project since 2007 as a first step towards the implementation of the integrated modelling approach for various ecosystems. The agro-system consists of a monoculture forest plantation of Eucalyptus grandis situated in KwaZulu-Natal, South Africa. This paper presents preliminary results from the KwaZulu-Natal E. grandis experimental study.