http://www.eo.uni-jena.de Jena Interactions of savanna vegetation structure with C-band Synthetic Aperture Radar (SAR) at pixel level V. Odipo 1* , C. Berger, C. Schmullius Introduction 1 *,2,3 Friedrich-Schiller-University Jena Grietgasse 6, 07743 Jena * [email protected]; [email protected], [email protected] +49 3641 94 89 26/ +49 3641 94 88 80 / 94 88 82 Contact Savanna vegetation structure especially aboveground biomass (AGB) and canopy cover (CC) are important for assessment of the biome’s ability to provide ecosystem services under various disturbance scenarios. Studies aimed at tree-level structure mapping provides the best way to monitor intrinsic changes when performed at resolutions permitting detection of forest degradation at localized scales. ESA’s Sentinel-1 synthetic aperture radar (SAR) data is freely available at high temporal and spatial resolutions which can solve problems of atmospheric interference experienced with optical data in the tropics. We assess feature detectability (cover classes) and error propagation as a result of changing SAR resolution from high to low within a Lowveld Savanna using terrestrial laser scanner (TLS) derived CC and AGB as reference data over a 9 sq.Km area in KNP. The study found that whereas changes are visible at higher resolutions, the RMSE in savanna structure detectability reduced with downscaling SAR resolution, irrespective of polarization and season. Study Area Datasets used Proposed Methodology Preliminary Analysis Preliminary Results Conclusion & Outlook • High correlation on SAR backscatter and AGB in dry season due to open canopy caused by „leaf-off“ • High correlation on SAR backscatter and CC in the wet season due to closed canopy • RMSE reduced while R-squared increased with downgrading SAR resolution, and vise versa • High AGB values – Inability to reconstruct trees with H<5 m and DBH <10 cm (shrubs left our) References • Odipo, V.O.; Nickless, A.; Berger, C.; Baade, J.; Urbazaev, M.; Walther, C.; Schmullius, C. Assessment of aboveground woody biomass dynamics using terrestrial laser scanner and L-band ALOS PALSAR data in South African savanna. Forests 2016, 7, 294; doi:10.3390/f7120294 • Colgan, M.S.; Asner, G.P.; Levick, S.R.; Martin, R.E.; Chadwick, O.A. Topo-edaphic controls over woody plant biomass in South African savannas. Biogeosciences 2012, 9, 1809–1821. Doi: 10.5194/bg- 9-1809-2012 • Mathieu, R.; Naidoo, L.; Cho, M.A.; Leblon, B.; Main, R.; Wessels, K.; Asner, G.P.; Buckley, J.; van Aardt, J.; Erasmus, B.F.N.; Smit, I.P.J. Toward structural assessment of semi-arid African savannahs and woodlands: The potential of multitemporal polarimetric RADARSAT-2 fine beam images. Rem. Sens. Of Environ. 2013,138: 215-231. Doi: 10.1016/j.rse.2013.07.011 • TLS tree H & DBH used in AGB (t/ha) calculation AGB=0.109D (1.39+0.14In(D)) H 0.73 ρ 0.80 ….. Colgan et al 2013 ) • LAStools: TLS derived CHM for CC (%) • GAMMA: SAR processing: Rad. Calib.> Geocod.> Topo. Normalization and Linear to σ 0 (dB) conversion DBH (cm) Height (m) Tree biomass CC (%) tree plot tree plot kg/tree t/ha plot Min 10 10 5.3 5.3 274.4 1.8 0 Median 31 41 10.7 10.7 6669.9 46.5 11.3 Mean 39.6 40.2 10.7 10.7 7793.4 54.3 15.6 Max 75 75 16.1 16.1 35285.2 226 90.2 • Heterogeneous savanna in S. Africa • Mix of grass, shrubs and trees. • Disturbance & heterogeneity drivers: fire, herbivory, nutrient, climate • Dry winter: May- Sept, Wet summer: Oct- April • MAP = 650mm, MAT= 22°C Fig.1: LC map of Skukuza study area, KNP,S. Africa • Field data (Mar., 2015): 42 plots (30x30 m); H (>1m H); DBH at 1m H. Use: validation of TLS measured H & DBH • TLS data (Sept., 2015); Use: Tree H & DBH, DTM, DSM & CHM; LC classification in OBIA Fig.3: Graphical summary of methodology used in the study • Ground range detected (GRD) C-band Sentinel-1A; IWS; Ascending pass; • Period: Feb. 2015 - Mar. 2017 (VH & VV) Tab.1: Summary of dependent variables: AGB and CC for both tree- and plot-level Fig. 2: Datasets used in the study- tree DBH, H; TLS data; C-band data & soil moisture • AGB and CC vs. SAR variables (VH/VV; Wet/Dry) using RF („mlr“) • Trimble RealWorks and QuickTerrainReader used for tree H & DBH measurement from TLS • High resolution- ease of change monitoring and life form separation (visual)-pixel mixing at low resolutions • C-band SAR is critical in savanna monitoring: high temporal & spatial resolution; freely available • TLS data – identification and measurement of tree H & DBH • TLS data-lower vegetation not captured due to insufficient point to reconstruct the tree trunk • Error propagation-limited validation data (TLS area, few plots with corse resolution) • Possibility to assess effects of multi-temporal filters in a heterogenous savanna • Predicted AGB & CC at different resolutions: 10, 30, 70 and 75 m • Average σ 0 (dB) per LCC, and life-form separability • RMSE in AGB & CC mapping using C-band SAR Acknowledgement