Disentangling disturbances with nested hierarchy classification of Mediterranean garrigue/maquis shrub community compositions through remote sensing and GIS

The paper presents an innovative nested hierarchical classification (NHC) method to disentangle landscape patterns between anthropogenic and natural drivers. This approach is motivated by the research need to differentiate between phytogeographic climate signals in vegetation communities based on these drivers. The method utilizes the non-conformity between spatial structures and granularity in remote sensing imagery with a twofold objective: (1) differentiate features within vegetation community compositions; and (2) aggregate these features with structural causal explanations. A sub-regional study in the southern Levant (Israel) is presented in which long-term ancient land use (LTA-LU) over 2300 years is juxtaposed with variable landscape patterns in modern semi-arid shrub communities. The modern land-cover is classified into three nestable classification levels using spaceborne, airborne, and unmanned aerial vehicle (UAV) imagery. NHC uses aggregation between the classification levels and the LTA-LU through a “bin sorting” technique based on the granularity and data type of the spatial structure. This technique enables the statistical feature extraction of vegetation community parts for association with their structural cause. The features identified indicate that areas of more intense LTA-LU are affected by plagioclimax attributed to industrial monoculture during the Hellenistic, Byzantine, and Ottoman periods. That form of land degradation, caused by long-term disturbance, results in substrates and vegetation compositions that cannot develop and recover to a climax state. [Display omitted] •Nested hierarchical land-cover classification shows plagioclimax in shrub landscape•Granularity is dominant factor to differentiate features in vegetation communities•Vertical land-use structure explains nested spatial structure through aggregation•Anthropogenically and naturally driven landscape patterns are disentangled