The proliferation of machine learning models, architectures, and datasets for Earth observation (EO) continues to rise dramatically. This pattern is expected to continue growing bringing with it an increase in the generation of remote sensing derived data products powered by geospatial artificial intelligence (GeoAI) techniques. Rigorous quality assessments and accuracy analysis needs to be undertaken for the science community to adopt many of these data products for scientific discovery of changes of the Earth’s land surface. While there is existing literature supporting and documenting best practices for the validation of GeoAI data products, the software to support large-scale collaborative validation efforts is limited. In this study we present the design and software implementation of a flexible commodity-based framework for large-scale global to regional validation of GeoAI data products. This framework’s main purpose is to enable, speed up, and optimize the acquisition of validation data for large-scale science projects with support across multiple sensors and spatial resolutions with little to no code. In addition, we present several use cases where this framework has enabled and streamlined the validation of global to regional data products at different spatial resolutions and within different computational platforms.

Human-induced land cover land use (LCLU) changes such as agricultural extensification and forest degradation and loss have extensive negative impacts including biodiversity loss, land degradation, and a disruption to ecological services. In Senegal, where people are heavily reliant on dryland agricultural production, climate change and land degradation pose particularly significant threats especially as rapid population growth continues to fuel frequent LCLU change. Considering these challenges, approaches that facilitate increased insight into the spatial and temporal dynamics of land use are needed to implement sustainable land management practices and mitigation strategies. However, difficulties associated with Senegal’s highly variable phenology, sparse woody cover and small, irregular fields necessitate the use of Very High Resolution (VHR; < 3 m spatial resolution) data and modern techniques for modeling land use at sufficient scales.We take advantage of VHR data’s spatial resolution and Sentinel-1’s high temporal resolution by implementing an object-based data fusion strategy to model land use. By generating high resolution vector objects from single-date WorldView imagery and using the corresponding Synthetic Aperture Radar (SAR) time series to train a One-Dimensional Convolutional Neural Network (1D CNN), we can effectively leverage deep learning techniques to extract land use signals from multi-resolution and multi-temporal data in a near-autonomous manner.

Land cover mapping has been a valuable tool in capturing changes in many developing regions in Africa. Senegal has been a hotspot of change where agricultural activity has rapidly increased. Agriculture in this region is often a complex mosaic of small fields which makes them difficult to classify using conventional land cover mapping methods and coarse-resolution satellite imagery. WorldView (WV) satellites provide very high-resolution imagery that is ideal for semantic segmentation using convolutional neural networks (CNN). In this study, we introduced training strategies that scale up the training data for the U-Net model using 2 m WV-2 and 3 imagery to overcome the challenges of regional mapping with a patchwork of hundreds of images. The proposed strategies increased the number of training data for the U-Net model in three main scenarios, (i) conventional training, (ii) model transfer, and (iii) transfer learning, and we evaluated model generalizability on test sets for two different regions in Senegal. The results showed that models rapidly reached a high level of performance with a limited increase in additional training in conventional and transfer learning strategies. In these two strategies, the U-Net consistently produced >87% average accuracy for trained images and >70% average accuracy for all test images at the final scale level. The research opens opportunities to produce regional land cover maps in West Africa without generating a prohibitively large amount of training data.

Validating land cover classification results from a machine learning model is a vital step in ensuring that further decisions are based on sound and robust results that can be trusted. Calculating pixel-wise validating statistics from a stack of land cover classification results, while computationally trivial for low-resolution imagery with a small spatial footprint, poses a significant challenge for very-high-resolution (VHR) imagery spanning a larger spatial footprint. Here we describe an open-source unified Python framework and workflow for the compositing of VHR imagery based on climatic and spatial information leveraging hardware acceleration. We additionally describe the implementation of per-pixel reduction algorithms which are used to reduce the stacked composite into a robust and accurate composite that is validated.

Timely and up-to-date bathymetry maps over large geographical areas have been difficult to create, due to the cost and difficulty of collecting in-situ calibration and validation data. Recently, combinations of spaceborne ICESat-2 lidar data and Landsat/Sentinel-2 data have reduced these obstacles. However, to date there have been no means of automatically extracting bathymetry photons from ICESat-2 tracks for model calibration/ validation and no well established open source workflows for generating regional scale bathymetric models. Here we provide an open source approach for generating bathymetry maps for the shallow water region around the island of Andros, Bahamas. We demonstrate an efficient means of processing 224 ICESat-2 tracks and 221 Landsat-8 scenes, using the C-SHELPh algorithm and Extra Trees Regression to provide 30 m pixel estimates of per-pixel depth and standard error. We map bathymetry with an RMSE of 0.32 m and RMSE% of 6.7 %. Our workflow and results demonstrate a means of achieving accurate regional–scale bathymetry maps from purely spaceborne data.

Forest aboveground biomass is a key environmental variable needed for constraining models of the global carbon cycle, monitoring stocks and fluxes of carbon in forests, and optimizing forest management toward climate mitigation. To date, limited satellite data have been available that are sensitive to Aboveground Biomass Density (AGBD), and the availability of new satellite lidar data streams from NASA's Global Ecosystem Dynamics Investigation (GEDI) [1] and Ice Cloud and Elevation Satellite (ICESat-2) [2] enable a new generation of AGBD estimates representative of 2018–2022 conditions. Here we explore the transferability of GEDI's AGBD estimation framework to ICESat-2. We compare distribution of Relative Height (RH) metrics from both products between 50 and 52° N, and find that ICESat-2's RH metrics are biased high compared to GEDI. We reprocess ICESat-2 RH metrics to make them more comparable to GEDI height metrics, and present a comparison of biomass estimates based on the original and new ICESat-2 RH metrics in boreal forests.

We map mangrove extents in Pongara National Park, Gabon using the Freeman-Durden Decomposition and InSAR Coherence derived from ALOS-2 imagery. Specifically, we obtain a land cover map derived from both this polarimetric decomposition and a 14-day repeat-pass coherence. Our classification model and results are highly interpretable based on a depth 2 decision tree. We further illustrate the correlation between InSAR coherence and height obtaining rough mangrove height estimates from TanDEM-X data. From our results, we observe that repeat-pass interferometric coherence provides invaluable information about mangrove extents and coastal forests. The clear identification of mangrove extents presents a significant opportunity for NISAR, which will provide 12-day repeat pass images over coastal areas globally.

Smallholder farms in South East Asia are characterized by small irregular field patterns, dense cloud cover and haze which limits our ability to observe changes in agriculture land-use. Very-high resolution (VHR,30 m) in regions with dense persistent cloud cover and haze from biomass burning.

High-resolution global-scale estimates of aboveground biomass density will soon be available from a suite of spaceborne LiDAR and radar missions. The 2016-2017 AfriSAR campaign was specifically designed to evaluate a suite of sensors for estimating biomass in a range of tropical forest environments. Here, we compare the calibration and biomass estimates from 5 different active sensors - ALOS Global Digital Surface Model (DSM), Shuttle Radar Topography Mission (SRTM), Uninhabited Aerial Vehicle Synthetic Aperture Radar (UAVSAR), NASA Land Vegetation Ice Sensor (LVIS), and TanDEM-X (TDX) - in the tallest known mangrove forest on Earth - Pongara National Park, Gabon. We leverage this comparison to evaluate the implications for future satellite missions that are aimed at improving global estimates of forest carbon storage. Our findings are directly relevant for space-borne missions estimating terrestrial carbon storage - GEDI, ICESat-2, NISAR, BIOMASS, Tandem-X, and Tandem-L - highlighting the specific uncertainty and bias that can be expected in several global biomass products in mangrove ecosystems.

From parks to shopping areas, smart technologies are being used throughout our cities to inform, guide and even persuade us into certain experiences. In terms of the technologies (and their usage), the emphasis is now very much on the mobile device and mobile applications that provide us with the digital media time to interact and share. Moreover, what we are increasingly witnessing and experiencing is how this mobile experience can fully absorb and disconnect us from the environment around us. The authors of this paper want to re-focus the actual role of the environment in the design of the smart city experiences. Integrating site-specific artworks with smart technologies, the goal of this research is to put the emphasis back into the environment as a place where everyone can engage and enjoy regardless of ability and/ or disability. This paper reports on the early conceptual stages of the Cardiff Bay Barrage project. It will highlight how the work (thinking and feeling) of artists, computer scientists, writers and engineers in alignment with the needs of industrial partners Cardiff Council and Philips Lighting Ltd. can bring `inclusivity' to the experience of all/any visitors to Cardiff Bay Barrage. This paper presents the `pattern making' process involved in the preparation for gathering and validating of initial requirements to support the overall design for this inclusive experience.

An effective particle swarm - quasi-Newton hybrid for the optimisation of continuous functions is developed, which is shown to work well on a range of test problems. This method exploits the global exploration abilities of the PSO algorithm and the fast convergence of the quasi-Newton method. New switching heuristics between the quasi-Newton and PSO methods are introduced, with the update pairs being used to generate new particles. The new hybrid, called L-PSO, is shown to be effective in obtaining the global minimum on a range of test problems, and outperforms previous hybrids with which it is compared.

Geographic features change over time, this change being the result of some kind of event or occurrence. It has been a research challenge to represent this data in a manner that reflects human perception. Most database systems used in GIS are relational, and change is either captured by exhaustively storing all versions of data, or updates replace previous versions. This stems from the inherent difficulty of modelling geographic objects in relational tables. This difficulty is compounded when the necessary time dimension is introduced to model how those objects evolve. There is little doubt that the object-oriented (OO) paradigm holds significant advantages over the relational model when it comes to modelling real-world entities and spatial data, and we believe that this contention is particularly true when it comes to spatiotemporal data. In this paper, we describe a generic, object-oriented model for representing spatiotemporal geographic data, called the Feature Evolution Model (FEM), based on a 'state-event-state' approach. The model exploits the expressiveness of OO technology by representing both geographic entities and change as objects, and the potential complexities introduced by the temporal elements of change are minimised by subtyping. The conceptual model is represented using UML and has the advantage of being implementable by any OO programming language and database development environment. The generic model is applied to real-world geographic data, that of OS MasterMap Integrated Transport Network (ITN) data.

This paper explores the use of the stochastic optimization technique of simulated annealing for map generalization. An algorithm is presented that performs operations of displacement, size exaggeration, deletion and size reduction of multiple map objects in order to resolve graphic conflict resulting from map scale reduction. It adopts a trial position approach in which each of n discrete polygonal objects is assigned k candidate trial positions that represent the original, displaced, size exaggerated, deleted and size reduced states of the object. This gives rise to a possible kn distinct map configurations; the expectation is that some of these configurations will contain reduced levels of graphic conflict. Finding the configuration with least conflict by means of an exhaustive search is, however, not practical for realistic values of n and k. We show that evaluation of a subset of the configurations, using simulated annealing, can result in effective resolution of graphic conflict.