Assessing the utility of remotely sensed data and integrated topographic characteristics for determining tree stand structural complexity in a re-forested urban landscape.

Master of Science in Environmental Science. University of KwaZulu-Natal 2017.

Transformation of natural landscapes into impervious built-up surfaces through urbanisation is known to significantly interfere with urban ecological integrity and its ability to provide environmental goods and services as well as accelerate climate change and associated impacts. Urban reforestation is widely promulgated as an ideal mitigation practice against impacts associated with urbanisation, however reforestation often has to compete with multiple and more “lucrative” urban land uses. This necessitates the optimisation of ecological benefits derived from reforestation within the limited available land. Such optimisation demands spatially explicit monitoring and evaluation (M&E). The recent proliferation of tree stand structural complexity (SSC) – a multidimensional index of the ecological performance of tree stands - offers great potential as an alternative indicator of ecological performance, instead of the one-dimensional traditional indicators such as Leaf Area Index, stem diameter and tree height. Furthermore, the recent advancements in remote sensing (RS) technology offers an improved potential of determining ecological performance across an urban reforested landscape. However, remotely sensed data costs and reliability often hinder their operational adoption. Consequently, the recent advancements in the freely available Sentinel 2 (S-2) data offer great potential for a cost effective operational M&E of SSC. The aim of this study was to i) Examine the utility of the freely available S-2 multispectral instrument imagery to determine SSC using the Partial Least Squares (PLS) regression technique within a re-forested urban landscape ii) Explore the potential of integrating topographic datasets with the S-2 data to determine SSC and iii) To rank the value of these variables in determining SSC. Tree structural data from a re-forested urban area was collected and a SSC index used to determine the area’s ecological performance. Multiple vegetation indices (VIs) were derived from the S-2 imagery while topographic variables (i.e. Topographic Wetness Index (TWI), slope, Area Solar Radiation (ASR), and elevation) were derived from a Digital Elevation Model (DEM). Results showed that the PLS model (n = 90) using the most important S-2 VIs (S2 REP, REIP, IRECI, GNDVI) produced a moderate predictive accuracy (0.215 NRMSECV) while topographybased model produced a high prediction accuracy (0.147 NRMSECV). Integrating the S-2 data with topographic information produced the highest prediction accuracy (0.13 NRMSECV). Furthermore, results indicate that SSC significantly varied across all topographic variables, with TWI and slope as the most important determinants of SSC. These results provide valuable spatially explicit information about the ecological performance of the reforested urban areas. Additionally, the study demonstrates the value of topographic data as an alternative predictor of SSC as well as the value of integrating the S-2 data with topographic characteristics in determining the performance of reforested areas.