What makes the difference between memory and face of a landscape? A machine learning approach applied to the federal state Brandenburg, Germany

The paper introduces two types of models: the “memory of a landscape” and the “face of a landscape”. The memory of a landscape refers to the development of a landscape as a result of many small and some major events. It can be described by a multitude of features that are difficult to change by humans, such as the initial geological substrate and the availability of nutrients linked to it. The implementation of the “memory model” leads to a scientific modelling approach that models the influence of the basic factors on forest distribution. The face of a landscape on the other hand implements a Big Data approach. The face can be changed more easily, e.g. by clearing forest areas and converting them into arable land. Both types of models are used to conclude from today’s perspective on the development of historical forests around 1880. A machine learning algorithm is used to implement both model types and evaluate the importance of features. Both models show differences in accuracy and simulation, which are discussed in detail. The inherent evaluation of the importance of the model inputs can be used to critically review some doctrines. The combination of machine learning with the knowledge of experts who help to select and prepare the data can be used in the future to depict the memory of a landscape more comprehensively in a model than is possible with previous approaches.

[1]  Bernhard Schölkopf,et al.  Training Invariant Support Vector Machines , 2002, Machine Learning.

[2]  E. Tuittila,et al.  Status and restoration of peatlands in northern Europe , 2003, Wetlands Ecology and Management.

[3]  Other,et al.  Remaking the landscape: the changing face of Britain , 2002 .

[4]  David R. Foster,et al.  Land-Use History (1730-1990) and Vegetation Dynamics in Central New England, USA , 1992 .

[5]  Cardona Alzate,et al.  Predicción y selección de variables con bosques aleatorios en presencia de variables correlacionadas , 2020 .

[6]  H. Ellenberg,et al.  Ecology of Central European Forests: Vegetation Ecology of Central Europe, Volume I , 2017 .

[7]  J. Deckers,et al.  World Reference Base for Soil Resources , 1998 .

[8]  Ralf Wieland,et al.  Automated feature selection for a machine learning approach toward modeling a mosquito distribution , 2017 .

[9]  Gerd Gigerenzer,et al.  Homo Heuristicus: Why Biased Minds Make Better Inferences , 2009, Top. Cogn. Sci..

[10]  S. Carpenter,et al.  Global Consequences of Land Use , 2005, Science.

[11]  Leo Breiman,et al.  Bagging Predictors , 1996, Machine Learning.

[12]  Lluís Corominas,et al.  Transforming data into knowledge for improved wastewater treatment operation: A critical review of techniques , 2017, Environ. Model. Softw..

[13]  G. Brierley.  Landscape memory: the imprint of the past on contemporary landscape forms and processes , 2010 .

[14]  Martin Hermy,et al.  Legacies of the past in the present-day forest biodiversity: a review of past land-use effects on forest plant species composition and diversity , 2007, Ecological Research.

[15]  Uta Steinhardt,et al.  Lehrbuch der Landschaftsökologie , 2012 .

[16]  E. Mückenhausen,et al.  Zum Alter des Plaggeneschs , 1968 .

[17]  Deutsche Ausgabe.  World Reference Base for Soil Resources 2006 , 2007 .

[18]  Z. Lipský,et al.  The changing face of the Czech rural landscape , 1995 .

[19]  D. Tilman,et al.  The Importance of Land-Use Legacies to Ecology and Conservation , 2003 .

[20]  Garry D. Peterson.  Contagious Disturbance, Ecological Memory, and the Emergence of Landscape Pattern , 2002, Ecosystems.

[21]  Zhiliang Zhu,et al.  US forest types and predicted percent forest cover from AVHRR data , 1994 .

[22]  Judit Padisák,et al.  Seasonal succession of phytoplankton in a large shallow lake (Balaton, Hungary) - a dynamic approach to ecological memory, its possible role and mechanisms , 1992 .

[23]  Vivian Scheinsohn,et al.  The memory of the landscape: Surface archaeological distributions in the Genoa Valley (Argentinean Patagonia) , 2016 .

[24]  J. Metzger,et al.  Relief influence on the spatial distribution of the Atlantic Forest cover on the Ibiúna Plateau, SP. , 2007, Brazilian journal of biology = Revista brasleira de biologia.

[25]  Ehrenfried Pfeiffer.  The Earth's Face: Landscape and Its Relation to the Health of the Soil , 2008 .

[26]  G. Glatzel,et al.  The impact of historic land use and modern forestry on nutrient relations of Central European forest ecosystems , 2005, Fertilizer research.

[27]  Shaowen Wang,et al.  A 3D convolutional neural network method for land cover classification using LiDAR and multi-temporal Landsat imagery , 2018, ISPRS Journal of Photogrammetry and Remote Sensing.

[28]  Yoshua Bengio,et al.  Random Search for Hyper-Parameter Optimization , 2012, J. Mach. Learn. Res..

[29]  Jarrett J. Barber,et al.  Quantifying ecological memory in plant and ecosystem processes. , 2015, Ecology letters.

[30]  Tianqi Chen,et al.  XGBoost: A Scalable Tree Boosting System , 2016, KDD.

[31]  Leo Breiman,et al.  Random Forests , 2001, Machine Learning.

[32]  Ralf Wieland,et al.  Multi-Scale Landscape Analysis (MSLA) - A method to identify correlation of relief with ecological point data , 2011, Ecol. Informatics.

[33]  Gaël Varoquaux,et al.  Scikit-learn: Machine Learning in Python , 2011, J. Mach. Learn. Res..

[34]  McKinney Wes,et al.  Python for Data Analysis , 2012 .

[35]  Monika Wulf,et al.  Land cover composition determinants in the Uckermark (NE Germany) over a 220-year period , 2016, Regional Environmental Change.