Indirect Estimation of Structural Parameters in South African Forests Using MISR-HR and LiDAR Remote Sensing Data

Parametric Models to Characterize the Phenology of the Lowveld Savanna at Skukuza, South Africa

Mathematical models, such as the logistic curve, have been extensively used to model the temporal evolution of biological processes, though other similarly shaped functions could be (and sometimes have been) used for this purpose. Most previous studies focused on agricultural regions in the Northern Hemisphere and were based on the Normalized Difference Vegetation Index (NDVI). This paper compares the capacity of four parametric double S-shaped models (Gaussian, Hyperbolic Tangent, Logistic, and Sine) to represent the seasonal phenology of an unmanaged, protected savanna biome in South Africa’s Lowveld, using the Fraction of Absorbed Photosynthetically Active Radiation (FAPAR) generated by the Multi-angle Imaging SpectroRadiometer-High Resolution (MISR-HR) processing system on the basis of data originally collected by National Aeronautics and Space Administration (NASA)’s Multi-angle Imaging SpectroRadiometer (MISR) instrument since 24 February 2000. FAPAR time series are automatically split into successive vegetative seasons, and the models are inverted against those irregularly spaced data to provide a description of the seasonal fluctuations despite the presence of noise and missing values. The performance of these models is assessed by quantifying their ability to account for the variability of remote sensing data and to evaluate the Gross Primary Productivity (GPP) of vegetation, as well as by evaluating their numerical efficiency. Simulated results retrieved from remote sensing are compared to GPP estimates derived from field measurements acquired at Skukuza’s flux tower in the Kruger National Park, which has also been operational since 2000. Preliminary results indicate that (1) all four models considered can be adjusted to fit an FAPAR time series when the temporal distribution of the data is sufficiently dense in both the growing and the senescence phases of the vegetative season, (2) the Gaussian and especially the Sine models are more sensitive than the Hyperbolic Tangent and Logistic to the temporal distribution of FAPAR values during the vegetative season, and, in particular, to the presence of long temporal gaps in the observational data, and (3) the performance of these models to simulate the phenology of plants is generally quite sensitive to the presence of unexpectedly low FAPAR values during the peak period of activity and to the presence of long gaps in the observational data. Consequently, efforts to screen out outliers and to minimize those gaps, especially during the rainy season (vegetation’s growth phase), would go a long way to improve the capacity of the models to adequately account for the evolution of the canopy cover and to better assess the relation between FAPAR and GPP.

Super-Resolution Restoration of MISR Images Using the UCL MAGiGAN System

High spatial resolution Earth observation imagery is considered desirable for many scientific and commercial applications. Given repeat multi-angle imagery, an imaging instrument with a specified spatial resolution, we can use image processing and deep learning techniques to enhance the spatial resolution. In this paper, we introduce the University College London (UCL) MAGiGAN super-resolution restoration (SRR) system based on multi-angle feature restoration and deep SRR networks. We explore the application of MAGiGAN SRR to a set of 9 MISR red band images (275 m) to produce up to a factor of 3.75 times resolution enhancement. We show SRR results over four different test sites containing different types of image content including urban and rural targets, sea ice and a cloud field. Different image metrics are introduced to assess the overall SRR performance, and these are employed to compare the SRR results with the original MISR input images and higher resolution Landsat images, where available. Significant resolution improvement over various types of image content is demonstrated and the potential of SRR for different scientific application is discussed.