Majchrowska A.: Abandonment of agricultural land in central Poland and its ecological role. Ekologia (Bratislava), Vol. 32, No. 3, p. 320-327, 2013. The intent of this paper is to give an overview of the ongoing study of farmland abandonment in central Poland. The goal of this project, initiated in 2011, is to provide field-verified information: 1. What is the spatial distribution of abandoned agricultural land in central Poland? How is it related to biophysical features of the study area? 2. What are botanical characteristics of spontaneous vegetation succession in the abandoned agricultural lands? What are their environmental determinants? 3. How do the abandoned lands correspond to existing protected areas? How could they be used to strengthen the ecological landscape structure of central Poland? The paper summarises the general study approach, discusses some methodological difficulties, presents preliminary findings and formulates the expected results.
The wide availability of multispectral satellite imagery through projects such as Landsat and Sentinel, combined with the introduction of deep learning in general and Convolutional Neural Networks (CNNs) in particular, has allowed for the rapid and effective analysis of multiple classes of problems pertaining to land coverage. Taking advantage of the two phenomena, we propose a machine learning model for the classification of land abandonment. We designed a Convolutional Neural Network architecture that outputs a classification probability for the presence of land abandonment in a given 15–25 ha grid element by using multispectral imaging data obtained through Sentinel Hub. For both the training and validation of the model, we used imagery of the Łódź Voivodeship in central Poland. The main source of truth was a 2009 orthophoto study available from the WMS (Web Map Service) of the Geoportal site. The model achieved 0.855 auc (area under curve), 0.47 loss, and 0.78 accuracy for the test dataset. Using the classification results and the Getis–Ord Gi* statistic, we prepared a map of cold- and hotspots with individual areas that exceed 50 km2. This thresholded heatmap allowed for an analysis of contributing factors for both low and intense land abandonment, demonstrating that common trends are identifiable through the interpretation of the classification results of the chosen model. We additionally performed a comparative field study on two selected cold- and hotspots. The study, along with the high-accuracy results of the model’s validation, confirms that CNN-type models are an effective tool for the automatic detection of land abandonment.
Generative adversarial networks (GANs) are a type of neural network that are characterized by their unique construction and training process. Utilizing the concept of the latent space and exploiting the results of a duel between different GAN components opens up interesting opportunities for computer vision (CV) activities, such as image inpainting, style transfer, or even generative art. GANs have great potential to support aerial and satellite image interpretation activities. Carefully crafting a GAN and applying it to a high-quality dataset can result in nontrivial feature enrichment. In this study, we have designed and tested an unsupervised procedure capable of engineering new features by shifting real orthophotos into the GAN’s underlying latent space. Latent vectors are a low-dimensional representation of the orthophoto patches that hold information about the strength, occurrence, and interaction between spatial features discovered during the network training. Latent vectors were combined with geographical coordinates to bind them to their original location in the orthophoto. In consequence, it was possible to describe the whole research area as a set of latent vectors and perform further spatial analysis not on RGB images but on their lower-dimensional representation. To accomplish this goal, a modified version of the big bidirectional generative adversarial network (BigBiGAN) has been trained on a fine-tailored orthophoto imagery dataset covering the area of the Pilica River region in Poland. Trained models, precisely the generator and encoder, have been utilized during the processes of model quality assurance and feature engineering, respectively. Quality assurance was performed by measuring model reconstruction capabilities and by manually verifying artificial images produced by the generator. The feature engineering use case, on the other hand, has been presented in a real research scenario that involved splitting the orthophoto into a set of patches, encoding the patch set into the GAN latent space, grouping similar patches latent codes by utilizing hierarchical clustering, and producing a segmentation map of the orthophoto.