The estimation of soil carbon stocks is an important component in environmental planning, policy and land management, particularly in the context of climate change mitigation. The following work presents national-scale soil property mapping at 10 m resolution, a level of detail not previously attempted in Scotland. The ultimate aim of this work is to facilitate carbon stock estimation from the soil properties to help inform planning and policy. Scottish organisations possess extensive field-based soil property records obtained through various initiatives, but coverage is inevitably constrained by cost and labour. A modelling-based approach informed by both field measurement, remote sensing imagery and other spatial covariates has the potential to fill the gaps in field-based accounts, providing contiguous estimates of soil carbon content, bulk density, profile depth, organic layer thickness and peat extent. Carbon content, bulk density and profile depth were predicted using machine learning techniques, yielding r2$$ {r}<^>2 $$ scores of 0.78, 0.65 and 0.7, respectively. Presence of peat was determined by the thickness of the organic layer for every grid cell profile (30% carbon content) that was greater or equal to 50 cm. National carbon stock was calculated by integrating the predicted soil properties down the full profile depth. This work demonstrates that soil properties can be mapped effectively using digital soil mapping techniques at high resolution, on a national scale, providing estimates of carbon stock and peat extent to aid policy makers in decision making.

Peat makes up roughly 28% of Scotland’s soil and is critical in many areas, including biodiversity and habitat support, water management, and carbon sequestration. The latter is only possible in healthy, undisturbed peatland habitats where the water table is sufficiently high, otherwise this potential carbon sink becomes a carbon source, that if left untreated will disappear forever. Drainage and erosion features are crucial indicators of peatland condition and are key for estimating national greenhouse gas emissions. Previous work on mapping peat depth and condition in Scotland has provided maps with reasonable accuracy at 100 metre resolution, allowing land managers and policymakers to both plan and manage these soils and to work towards identifying priority peat sites for restoration. However, the spatial variability of the surface condition is much finer than this scale, limiting the ability to inventory greenhouse gas emissions or develop site-specific restoration and management plans. This work involves an updated set of mapping using high-resolution (25 cm) aerial imagery which provides the ability to identify and segment individual drainage channels and erosion features. Combining this imagery with a classical deep learning-based segmentation model, enables high spatial resolution, national scale mapping to be carried out allowing for a deeper understanding of Scotland’s peatland resource and which will enable various future analyses using this data.

Peat makes up approximately a quarter of Scotland's soil by area. Healthy, undisturbed, peatland habitats are critical to providing resilient biodiversity and habitat support, water management, and carbon sequestration. A high and stable water table is a prerequisite to maintain carbon sink function; any drainage turns this major terrestrial carbon store into a source that feeds back further to global climate change. Drainage and erosion features are crucial indicators of peatland condition and are key for estimating national greenhouse gas emissions. Previous work on mapping peat depth and condition in Scotland has provided maps with reasonable accuracy at 100-m resolution, allowing land managers and policymakers to both plan and manage these soils and to work towards identifying priority peat sites for restoration. However, the spatial variability of the surface condition is much finer than this scale, limiting the ability to inventory greenhouse gas emissions or develop site-specific restoration and management plans. This work involves an updated set of mapping using high-resolution (25 cm) aerial imagery, which provides the ability to identify and segment individual drainage channels and erosion features. Combining this imagery with a classical deep learning-based segmentation model enables high spatial resolution, national scale mapping to be carried out allowing for a deeper understanding of Scotland's peatland resource and which will enable various future analyses using these data.

Abstract. Peatlands, which only cover 3 to 5 percent of the global land area, can store up to twice the amount of carbon as the world’s forests. Although recognised for their significant role in the global carbon cycle, discovering the global extent of peatlands and their carbon stock remains challenging. Referring to the UNEP's global peatland map, here we present PEATGRIDS, a data product containing global maps of peat thickness and carbon stock created created using the digital soil mapping approach. We compiled over 25,000 observations of peatland thickness, bulk density (BD) and carbon content (CC), globally. Using the Random Forest (RF) algorithm, we estimated peat thickness and peat BD and CC at ~1 km resolution at multiple depths (0–2 m) globally. The estimates were generated using 19 land surface covariates from digital maps and remote sensing images of land use, soil characteristics, topographical features, and climate parameters. The RF models for peat thickness were trained on 25,200 points grouped into six geographic regions. Validation of the peat thickness estimates showed a good performance, with the coefficient of determination (R2) ranging from 0.15 to 0.72. The prediction for peat BD and CC followed the same model architecture and were trained on 17,000 and 7,000 points, respectively. Overall, BD and CC models performed well and consistently across soil layers with average R2 values of 0.61 for BD and 0.48 for CC. Based on the estimated peat thickness, BD and CC, the carbon stock of global peatland was estimated to be 1,029 Pg C for peat dominated area of 6.57 million km2. PEATGRIDS is made available at https://doi.org/10.5281/zenodo.12559239 (Widyastuti et al., 2024) to support further analyses and modelling of peatlands across the globe.

This study trialled a convolutional neural net (CNN)-based approach to mapping peat ditches from aerial imagery. Peat ditches were dug in the last century to improve peat moorland for agriculture and forestry at the expense of habitat health and carbon sequestration. Both the quantitative assessment of drained areas and restoration efforts to re-wet peatlands through ditch blocking would benefit from an automated method of mapping, as current efforts involve time-consuming field and desk-based efforts. The availability of LiDAR is still limited in many parts of the UK and beyond; hence, there is a need for an optical data-based approach. We employed a U-net-based CNN to segment peat ditches from aerial imagery. An accuracy of 79% was achieved on a field-based validation dataset indicating ditches were correctly segmented most of the time. The algorithm, when applied to an 802 km(2 )area of the Flow Country, an area of national significance for carbon storage, mapped a total of 27,905 drainage ditch features. The CNN-based approach has the potential to be scaled up nationally with further training and could streamline the mapping aspects of restoration efforts considerably.

Land degradation directly affects around 25% of land globally, undermining progress on most of the UN Sustainable Development Goals (SDG), particularly target 15.3. To assess land degradation, SDG indicator 15.3.1 combines sub-indicators of productivity, soil carbon and land cover. Over 100 countries have set Land Degradation Neutrality (LDN) targets. Here, we demonstrate application of the indicator for a well-established agricultural landscape using the case study of Great Britain. We explore detection of degradation in such landscapes by: 1) transparently evaluating land cover transitions; 2) comparing assessments using global and national data; 3) identifying misleading trends; and 4) including extra sub-indicators for additional forms of degradation. Our results demonstrate significant impacts on the indicator both from the land cover transition evaluation and choice or availability of data. Critically, we identify a misleading improvement trend due to a trade-off between improvement detected by the productivity sub-indicator, and 30-year soil carbon loss trends in croplands (11% from 1978 to 2007). This carbon loss trend would not be identified without additional data from Countryside Survey (CS). Thus, without incorporating field survey data we risk overlooking the degradation of regulating and supporting ecosystem services (linked to soil carbon), in favour of signals from improving provisioning services (productivity sub-indicator). Relative importance of these services will vary between socioeconomic contexts. Including extra sub-indicators for erosion or critical load exceedance, as additional forms of degradation, produced a switch from net area improving (9%) to net area degraded (58%). CS data also identified additional degradation for soil health, including 44% arable soils exceeding bulk density thresholds and 35% of CS squares exceeding contamination thresholds for metals.

The commitment by over 100 governments covering over 90% of the world’s forests at the COP26 in Glasgow to end deforestation by 2030 requires more effective forest monitoring systems. The near real-time (NRT) change detection of forest cover loss enables forest landowners, government agencies and local communities to monitor natural and anthropogenic disturbances in a much timelier fashion than the thematic maps that are released every year. NRT deforestation alerts enable the establishment of more up-to-date forest inventories and rapid responses to unlicensed logging. The Copernicus Sentinel-2 satellites provide operational Earth observation (EO) data from multi-spectral optical/near-infrared wavelengths every five days at a global scale and at 10 m resolution. The amount of acquired data requires cloud computing or high-performance computing for ongoing monitoring systems and an automated system for processing, analyzing and delivering the information promptly. Here, we present a Sentinel-2-based NRT change detection system, assess its performance over two study sites, Manantlán in Mexico and Cartagena del Chairá in Colombia, and evaluate the forest changes that occurred in 2018. An independent validation with very high-resolution PlanetScope (~3 m) and RapidEye (~5 m) data suggests that the proposed NRT change detection system can accurately detect forest cover loss (> 87%), other vegetation loss (> 76%) and other vegetation gain (> 71%). Furthermore, the proposed NRT change detection system is designed to be attuned using in situ data. Therefore, it is scalable to larger regions, entire countries and even continents.

Monitoring forest cover change from Earth observation data streams in near-real-time presents a challenge for automated change detection by way of a continuously updated big dataset. Even though deforestation is a significant global problem, forest cover changes in pairs of subsequent images happen relatively infrequently. Detecting a change can require the download and processing of tens, hundreds or even thousands of images. In geoscientific applications of Earth observation, machine learning algorithms are increasingly used. Once trained, a machine learning model can be applied to new images automatically. This paper introduces the open-access Python 3 package Pyeo - "Python for Earth Observation". Pyeo provides a set of portable, extensible and modular Python functions for the automation of machine learning applications from Earth observation data streams, including automated search and download functionality, pre-processing and atmospheric correction, re-projection, creation of thematic base layers and machine learning classification or regression. Pyeo enables users to train their own machine learning models and then apply the models to newly downloaded imagery over their area of interest. This paper describes in detail how Pyeo works, its requirements, benefits, and a description of the libraries used. An application to the automated forest cover change detection in a region in Kenya is given. Pyeo can be used on cloud computing architectures such as Amazon Web Services, Microsoft Azure and Google Colab to provide scalable applications and processing solutions for the geosciences.

We present an image processing method for accurately segmenting crop plots from Unmanned Aerial System imagery (UAS). The use of UAS for agricultural monitoring has increased significantly, emerging as a potentially cost effective alternative to manned aerial surveys and field work for remotely assessing crop state. The accurate segmentation of small densely-packed crop plots from UAS imagery over extensive areas is an important component of this monitoring activity in order to assess the state of different varieties and treatment regimes in a timely and cost-effective manner. Despite its importance, a reliable crop plot segmentation approach eludes us, with best efforts being relying on significant manual parameterization. The segmentation method developed uses a combination of edge detection and Hough line detection to establish the boundaries of each plot with pixel/point based metrics calculated for each plot segment. We show that with limited parameterization, segmentation of crop plots consistently over 89% accuracy are possible on different crop types and conditions. This is comparable to results obtained from rice paddies where the plant material in plots is sharply contrasted with the water, and represents a considerable improvement over previous methods for typical dry land crops.

Despite advances in data collection and analytical software, mapping of glacial geomorphology is still largely achieved through manual techniques. Manual mapping is a relatively slow, inefficient and potentially subjective process, particularly given the extensive spatial coverage of remotely sensed datasets. Thus, glacial geomorphic mapping would benefit from semi-automated techniques able to map large areas quickly and efficiently. Here, we develop novel semi-automated techniques for mapping glacial geomorphology, based on lidar derived data collected from Breiðamerkurjökull, Iceland using Object Based Image Analysis (OBIA). The landforms classified were moraines, eskers, flutes/drumlins, and glacio-fluvial outwash. Several land surface parameters (LSPs) were calculated during image segmentation and analysis, representing the characteristics of the landforms of interest. The segmentation and classification routines were developed in Matlab, whereas most attempts so far have used specialised OBIA software. Our techniques utilised exploratory data analysis, where statistical groupings within the data define different landform types. Accuracy was assessed against a reference map derived using traditional manual mapping techniques. Landforms were classified to an overall accuracy of 77%. Our semi-automated mapping techniques should, with further algorithm development, be applicable using other datasets commonly used in palaeo-glaciology such as radar derived elevation data and swath bathymetry.

Mapped topographic features are important for understanding processes that sculpt the Earth's surface. This paper presents maps that are the primary product of an exercise that brought together 27 researchers with an interest in landform mapping wherein the efficacy and causes of variation in mapping were tested using novel synthetic DEMs containing drumlins. The variation between interpreters (e.g. mapping philosophy, experience) and across the study region (e.g. woodland prevalence) opens these factors up to assessment. A priori known answers in the synthetics increase the number and strength of conclusions that may be drawn with respect to a traditional comparative study. Initial results suggest that overall detection rates are relatively low (34–40%), but reliability of mapping is higher (72–86%). The maps form a reference dataset.