Effective monitoring of agricultural lands requires accurate spatial information about the locations and boundaries of agricultural fields. Through satellite imagery, such information can be mapped on a large scale at a high temporal frequency. Various methods exist in the literature for segmenting agricultural fields from satellite images. Edge-based, region-based, or hybrid segmentation methods are traditional methods that have widely been used for segmenting agricultural fields. Lately, the use of deep neural networks (DNNs) for various tasks in remote sensing has been gaining traction. Therefore, to identify the optimal method for segmenting agricultural fields from satellite images, we evaluated three state-of-the-art DNNs, namely Mask R-CNN, U-Net, and FracTAL ResUNet against the multiresolution segmentation (MRS) algorithm, which is a region-based and a more traditional segmentation method. Given that the DNNs are supervised methods, we used an optimised version of the MRS algorithm based on supervised Bayesian optimisation. Monotemporal Sentinel-2 (S2) images acquired in Lower Saxony, Germany were used in this study. Based on the agricultural parcels declared by farmers within the European Common Agricultural Policy (CAP) framework, the segmentation results of each method were evaluated using the F-score and intersection over union (IoU) metrics. The respective average F-score and IoU obtained by each method are 0.682 and 0.524 for Mask R-CNN, 0.781 and 0.646 for U-Net, 0.808 and 0.683 for FracTAL ResUNet, and 0.805 and 0.678 for the optimised MRS approach. This study shows that DNNs, particularly FracTAL ResUNet, can be effectively used for large-scale segmentation of agricultural fields from satellite images.

The Group on Earth Observations Global Agricultural Monitoring Initiative (GEOGLAM) considers agricultural fields as one of the essential variables that can be derived from satellite data. We evaluated the accuracy at which agricultural fields can be delineated from Sentinel-1 (S1) and Sentinel-2 (S2) images in different agricultural landscapes throughout the growing season. We used supervised segmentation based on the multiresolution segmentation (MRS) algorithm to first identify the optimal feature set from S1 and S2 images for field delineation. Based on this optimal feature set, we analyzed the segmentation accuracy of the fields delineated with increasing data availability between March and October of 2018. From the S1 feature sets, the combination of the two polarizations and two radar indices attained the best segmentation results. For S2, the best results were achieved using a combination of all bands (coastal aerosol, water vapor, and cirrus bands were excluded) and six spectral indices. Combining the radar and spectral indices further improved the results. Compared to the single-period dataset in March, using the dataset covering the whole season led to a significant increase in the segmentation accuracy. For very small fields (< 0.5 ha), the segmentation accuracy obtained was 27.02%, for small fields (0.5 – 1.5 ha), the accuracy was 57.65%, for medium fields (1.5 ha – 15 ha), the accuracy was 75.71%, and for large fields (>15 ha), the accuracy stood at 68.31%. As a use case, the segmentation result was used to aggregate and improve a pixel-based crop type map in Lower Saxony, Germany.

Agricultural intensification has simplified landscape composition and configuration, which has led to biodiversity declines. Increasing landscape‐wide crop heterogeneity can promote farmland biodiversity. However, knowledge is still lacking on how the effects of configurational and compositional crop heterogeneity (i.e. field size and crop diversity) are modulated by the amount of semi‐natural habitats in the landscape, especially across large scales. We tested how mean field size and functional crop diversity affect farmland bird diversity and abundance over three consecutive years, and how these effects are modulated by the amount of small woody features (SWF) in the landscape. We related data from a national bird monitoring scheme to field‐level information from a novel, high‐resolution remote sensing‐based crop type map. Smaller field sizes and higher functional crop diversity were not generally associated with a higher diversity or abundance of farmland birds. Associations varied with species' breeding habitat preferences and were modulated by the amount of SWF. In landscapes with a low SWF amount, species diversity and the abundance of species breeding in field edges or shrubs were negatively associated with increasing field size. However, where the amount of SWF was high, larger field size was associated with higher species diversity and abundance of field and shrub breeders. Diversity increased with higher functional crop diversity, as did the abundance of non‐field breeders in landscapes with a medium to high SWF amount. Field size tended to have a stronger effect on bird diversity and abundance than functional crop diversity. Policy implications: National and EU agricultural policies should adopt a landscape perspective by considering the amount of semi‐natural habitats when designing biodiversity‐enhancing measures that target field size and functional crop diversity. In landscapes with low SWF amount, decreasing field sizes may be particularly effective to promote farmland bird diversity and the abundance of non‐field breeders. In landscapes with a medium to high SWF amount, increasing functional crop diversity is likely more effective than reducing field sizes. Field and shrub breeders may be promoted by maintaining landscapes with large fields, only if these offer a high SWF amount, low agronomic yield potential and low productivity.

Image segmentation is a cost-effective way to obtain information about the sizes and structural composition of agricultural parcels in an area. To accurately obtain such information, the parameters of the segmentation algorithm ought to be optimized using supervised or unsupervised methods. The difficulty in obtaining reference data makes unsupervised methods indispensable. In this study, we evaluated an existing unsupervised evaluation metric that minimizes a global score (GS), which is computed by summing up the intra-segment uniformity and inter-segment dissimilarity within a segmentation output. We modified this metric and proposed a new metric that uses absolute difference to compute the GS. We compared this proposed metric with the existing metric in two optimization approaches based on the Multiresolution Segmentation (MRS) algorithm to optimally delineate agricultural parcels from Sentinel-2 images in Lower Saxony, Germany. The first approach searches for optimal scale while keeping shape and compactness constant, while the second approach uses Bayesian optimization to optimize the three main parameters of the MRS algorithm. Based on a reference data of agricultural parcels, the optimal segmentation result of each optimization approach was evaluated by calculating the quality rate, over-segmentation, and under-segmentation. For both approaches, our proposed metric outperformed the existing metric in different agricultural landscapes. The proposed metric identified optimal segmentations that were less under-segmented compared to the existing metric. A comparison of the optimal segmentation results obtained in this study to existing benchmark results generated via supervised optimization showed that the unsupervised Bayesian optimization approach based on our proposed metric can potentially be used as an alternative to supervised optimization, particularly in geographic regions where reference data is unavailable or an automated evaluation system is sought.