talk

talk

The European Environment Agency (EEA) has produced ecosystem extent accounts for the European Union and its member countries for 2000 – 2018 on the basis of the European CORINE Land Cover (CLC) data set. This work is now being revised in the context of the proposal for a module on ecosystem accounting in the EU Environmental Accounts Regulation (691/2011) which integrates the UN SEEA EA standard into the EU statistical system. A technical guidance document in support of the implementation of ecosystem extent accounts suggests further detailed sub-divisions of ecosystem types based on the IUCN Global Ecosystem Typology (recommended within SEEA EA) and the European habitat classification system EUNIS. This increases the range of ecosystem types to 46 at level 2 and to 137 at level 3. Such detailed ecosystem sub-divisions support the use of ecosystem extent accounts for monitoring the achievement of EU policy targets but can only be achieved by combining new satellite data products with in situ reference data through the use of advanced techniques. This paper sets out current progress achieved in projects financed and managed by the EEA to develop and test methods for using satellite data as a key foundation for the mapping of ecosystem sub-types (habitats) in Europe. It describes what approaches have been used to combine Copernicus high resolution data sets with climate and soil variables as well as habitat in-situ data for validation and enhancement. It further sets out what understanding has been gained in the use of machine-learning methods for habitat mapping. The paper concludes with a discussion of the key challenges in the further development of detailed ecosystem maps based on advanced satellite data sets.

Annual maps showing the spatial distribution of ecosystems and their change over time are required for the development of ecosystem extent accounts. Many countries have been successfully using earth observation (EO) data to produce land cover/use maps but have yet to produce ecosystem extent maps in compliance with the System of Environmental Economic Accounting—Ecosystem Accounting. To address this issue, we developed a powerful and easy-to-implement method for deriving an ecosystem extent map suitable for ecosystem accounting at national and sub-national scales. The method comprises of three steps: 1) land cover mapping, 2) ecosystem modelling, and 3) aggregation of the results from the previous steps to produce a final ecosystem extent map. The method for land cover mapping uses a binary classification approach where each land cover class is mapped individually using Landsat imagery and various auxiliary geographical and topographical information. The ecosystem modelling uses Generalized Dissimilarity Modelling (GDM) with plant species data obtained from the Botanical Information and Ecology Network (BIEN) dataset and various environmental variables to produce a map of biotic plant dissimilarity. The GDM approach using plant species as one of the inputs is particularly effective for ecosystem classification since it can tease apart differences in forest ecosystem types that cannot be easily detected with EO data. In the final step, the two results are combined using a simple spatial overlay technique. We demonstrate the application of the method in Liberia and show how this approach can be used to generate an annual time series of ecosystem extent maps to advance ecosystem accounting in Liberia and beyond.

Background of our research is that the latest assessment by the EEA (The European environment – state and outlook 2020) shows that Europe’s biodiversity continues to decline at an alarming rate, with most protected species and habitats found not to have a good conservation status. Artificial Intelligence (AI) techniques, such as Machine Learning (ML) and Neural Networks or Deep Learning methods could enable an improved monitoring of biodiversity and ecosystems with satellite based high-resolution datasets such as Copernicus High Resolution Vegetation Phenology Product (HR-VPP) to better support European policy making. So understanding where habitats occur across Europe is a crucial element for understanding biodiversity conservation and taking specific actions. Since at the EU level information on habitats is currently limited at the 10km grid level (Annex I), more spatial detailed habitat maps are required using innovative methods such artificial intelligence/machine learning based on high resolution satellite and other ancillary data and integration of in-situ observations as training data from the European Vegetation Archive (EVA). Therefore we have been focussing lately one i) Production of high-resolution habitat suitability maps for Europe at 100 meter resolution for most EUNIS habitat types at level 3, ii) production of very high-resolution habitat suitability maps at 10 meter resolution for a.o Austria, Portugal and the Netherlands, and iii) explore deeplearning techniques for habitat mapping at regional scale. For all used methods the amount of training data from EVA is crucial for the classification/modelling precision, and differs a lot per region. In general much more effort should be made on the collection and the enhancement of the training data.

Les comptes d’occupation du sol de 2010 et 2015 du Sénégal ont été produits par les grâce à l’outil ARIES et à l’assistance technique de l’UNSD. Les données utilisées pour la compilation étaient les cartes d’occupation du sol issues de la cartographie réalisée par l’Agence nationale de l’Aménagement du Territoire (ANAT, images satellites Landstat Sentinel-2, 2010) et le Centre de suivi écologique (CSE, images satellites Landstat-8, 2015). Ces deux cartes comportent différentes classes d’occupation du sol. A cet effet, un premier travail d’harmonisation sémantique a permis d’établir une nomenclature nationale détaillée de 18 classes selon cinq thèmes : (i) surface artificialisée ; (ii) zone dénudée ; (iii) surface cultivée ; (iv) surface boisée et (v) région hydrique. Des corrections topologiques et géométriques ont aussi été apportées aux cartes pour corriger les écarts non justifiés par l’évolution spatiale entre les deux années et permettre une comparaison fiable. Les résultats montrent que le Sénégal est principalement constitué de surface boisée, 69,83% de la superficie totale en 2010. Sa superficie a toutefois connu une régression de 5,7% sur la période 2010- 2015. La deuxième classe la plus importante est celle des surfaces cultivées (24,54% en 2010) qui par contre, a vu sa superficie augmenter de 13,0%. Les régions hydriques ont une faible part dans la répartition de la superficie (3,74% en 2010) mais se sont étendues de 17,4%. Les zones dénudées et les surfaces artificialisées représentaient moins de 2% de la superficie totale en 2010 (1,34% et 0,55% respectivement). Elles ont connu une expansion de 7,4% tirée par celle des surfaces artificialisées (21,9%). La compilation des comptes de l’occupation du sol a permis aussi de calculer l’indicateur ODD 15.1.1.

This presentation includes a review of the SDG 6.6.1 data flow and the availability of open and free satellite data available for taking stock of and monitoring freshwater ecosystems globally. The presentation will also review the design and development of novel open-source satellite-based applications and services ready to be deployed on regional and national digital platforms. SDG Indicator 6.6.1 tracks the extent to which different types of water-related ecosystems are changing in extent over time. The satellite data used to monitor indicator 6.6.1 has been disaggregated into ecosystem types, thereby enabling ecosystem-level decisions to be taken. The satellite-based reporting methodology developed for SDG 6.6.1 indicator is one of the most advanced non-traditional data indicators. Yet, despite its promise a key challenge remains to decentralize these global data products and build ownership and action at national level. Global efforts to monitor SDG 6.6.1 with satellite data has been coordinated by UNEP, but in fact the UN Member States own the SDG monitoring and hence a need to look beyond a global monitoring system and look at how national EO based monitoring can be enabled. The implementation of a full end-to-end national SDG monitoring system needs to consider; [i] data sources and access; [iii] provision of data processing tools, [iii] ensure efficient mechanism for reporting and dissemination and [iv] and adopting an ecosystem typology (e.g., IUCN Global Ecosystem Typology) that not only ensure effective progress reporting for the SDGs but also can help guide the reporting and transformation of ecosystem policy and management for other important global frameworks e.g., SEEA Ecosystem Accounting, the Convention on Biological Diversity and the Ramsar Wetland Classification.

Land account is used to assist the monitoring and understanding of how land cover has evolved, which is crucial for both national and provincial spatial planning. To support this policy relevance, Statistics Indonesia has started to compile land account since 2016. The result is published annually in SISNERLING (Integrated System of Environmental-Economic Accounts of Indonesia). KLHK (Ministry of Environment and Forestry) and BIG (Geospatial Information Agency) serve as the primary data sources for land account compilation. Under one map policy (PKSP, read https://satupeta.go.id/), the two institutions are required to provide consistent extents that refer to one geospatial reference, one standard, one database, and one geoportal so it can be used as accurate and accountable guidance in the spatial-related policy making. However, differences in findings between KLHK and BIG's data processing results were discovered due to differences in the base maps used by the two institutions. Disregarding the differences, one key finding from land account in SISNERLING publication remains the high rate of change in classes related to forest cover of Indonesia. This information on forest cover area and forest cover loss are used as indicators that support SDG number 15 that aims to protect, restore and promote sustainable use of terrestrial ecosystems, sustainably manage forests, combat desertification, and halt and reverse land degradation and halt biodiversity loss.

We introduce the development status of the open-source demo software "WaterMaskAnalyzer" (WMA) for the determination of water body extents using satellite data and the Google Earth Engine (GEE). Freely available satellite images from the EU Copernicus program can record water surfaces precisely and at high temporal resolution. The application allows a simple to use on-demand monitoring of inland water dynamics by the Otsu-thresholding algorithm that automatically classifies water bodies. The WMA can be used to work out the water change zones over multi-year time series. The dynamics of the ecosystems in the riparian zones of water bodies can thus be described in detail. Also, the current status of a water body can be recorded with pixel accuracy using various tools. It will be easy in the future to send out ecological mappers to dynamic water bodies in a targeted and day-precise manner. In this way, the exact species (e.g. mud-dependent flora and fauna) that are being sought can be recorded. Mapping is thus no longer left to the chance of a predetermined mapping date. One advantage of the WMA is that, unlike comparable products, it not only offers predefined content, but also offers the user a wide range of setting parameters. Thus, consideration is given to the individual specialist question of the user. The WMA is available at the following link: https://bit.ly/3wEdnMI Note: A communication introducing the WMA is newly published: https://www.mdpi.com/2072-4292/14/18/4485

Radar satellite Earth Observation time series data capture the changes of Earth’s surface features in the imaged area. These temporal changes contain information about land cover type (e.g., forests vs built-up areas vs waters). In this presentation, we will share results of joint research efforts of Statistics Canada and Environment and Climate Change Canada that demonstrate the potential of functional principal component analysis as a powerful feature engineering technique to extract dominant temporal trends from Synthetic Aperture Radar satellite time series data and its potential application to land cover classification, a key element of the Canadian Census of Environment program.

Mexico is considered one of the megadiverse countries together Brazil, Peru, Indonesia, Madagascar and Colombia. Ecosystems are subject to disturbances that affect its structure and functioning. In order to estimate the extension covered by ecosystems in Mexico and analyze the dynamics that they have experienced over time, we analyzed 4 Land Use and Vegetation series that have been developed and published by the National Institute of Statistics and Geography of Mexico. To calculate the extension, a methodological adaptation of IPCC-CONAFOR-N3 classification system was used. To analyze the dynamics of the ecosystems, the change in their extension was compared in the period between 2007 (Series III USUEV) and 2018 (Series VII). The results show that as of 2018, the natural ecosystems with the greatest extension are primary non-woody xeric scrub (330,778 km2), primary woody xeric scrub (176,719 km2) and the primary coniferous forest (127,006 km2). The most extensive anthropized ecosystems are annual agriculture (312,043 km2), human settlements (24,036 km2) and perennial agriculture (20,434 km2). Ecosystems have been dynamic in the analyzed period, one of the most important is the semi-deciduous forest, which presented the highest negative rate of change (-3.18% per year). The opposite occurs with ecosystems that had positive rates of change such as hydrophilic secondary woody vegetation (3.13%), aquaculture (3.49%), human settlements (3.77%) and cultivated forests (9.85%). An important case is the Mountain Cloud Forest, which has presented a loss of vegetation cover, presenting a negative annual rate of change in vegetation cover, having -0.60 % in secondary vegetation and the -0.04% in primary vegetation. Is important to implement the use of artificial intelligence tools to obtain information more frequently. These results are the basis for the analysis of the condition of the ecosystems, from which it could estimate its capacity to provide environmental services.

talk

The EU Biodiversity Strategy for 2030 anticipates the development of an EU-wide methodology to map, assess and achieve good condition of ecosystems. The Joint Research Centre (European Commission) has led the development of such methodology, making use of the System of Environmental Economic Accounting - Ecosystem Accounting (SEEA EA) of United Nations as reference framework. Specifically, the EU-wide methodology follows the rules of the SEEA-EA ecosystem condition accounts, presenting useful insights to operationalise this framework for all ecosystem types in the EU. The EU-methodology provides a comprehensive set of condition variables per ecosystem type as well as recommendations on methods for setting reference levels and thresholds to determine good ecosystem condition. In this presentation, we introduce the EU-wide methodology, its relationship with the SEEA-EA, and the challenges identified during the development of this work. Regarding challenges, the presentation will introduce those that could be mitigated through further advances in Earth Observation data. The case of urban ecosystems will be used to illustrate the challenges in a practical form, to facilitate their comprehension by a broad audience. To conclude, the presentation will open a discussion on the role of Earth Observation for the mapping and assessment of good ecosystem condition, and which should be the next steps.

The Federal Statistical Office of Germany is currently working on setting up comprehensive Ecosystem Condition Accounts. The work builds on the already published national Extent Accounts that distinguish up to 74 national ecosystem classes. Following the SEEA-EA framework, we draw up an Ecosystem Condition Typology (ECT) for each of these ecosystem classes including variables and indicators for abiotic, biotic and landscape characteristics. A multitude of data sources, mainly earth observation data and in-situ monitoring, is used to calculate variables locally and nationally. These data sources are harmonized and aggregated to accounting areas using a standardized and automated approach. Where possible, condition indicators, i.e. ecosystem variables set against reference values, are derived. In order to find appropriate and local reference values using various methods depending on the ecosystem type. Examples include the use of arguably pristine local references for forest condition, thresholds from the EU Marine Strategy Water Framework Directive and historical references for marine ecosystems and expert-based references for anthropogenic ecosystems. In this paper, we first present a comprehensive ecosystem condition typology and methods of finding reference values when setting up the German Ecosystem Condition Accounts. Different implementation challenges, such as different spatial and temporal resolution of input data, data validity, aggregation rules and the feasibility of referencing as well as potential solutions are discussed. In addition, awareness to data gaps that can be tackled by remote sensing products is raised.

Field-based monitoring of aquatic ecosystems, such as lakes and wetlands, is limited by logistic constraints and costs that hamper timeliness and spatial coverage of data collection, more than what happens in most terrestrial biomes. Yet, anthropic impacts – direct and indirect – are critically endangering inland and transitional habitats even more seriously and rapidly than other ecosystem types (IPBES, 2019). With the technical developments of spaceborne platforms and increasing operational uptake – developed for Europe in the context of Copernicus program, and supporting the implementation of Water Framework and Habitats directives – ecologically significant applications of remote sensing have become a reality in the last decade. Earth Observation (EO) has fast become the ideal candidate tool to map aquatic ecosystem assets and conditions quantitatively and efficiently, constituting targets for their accounting. In particular, EO can provide frequent and synoptic data at multiple scales (from local to global) that cover aquatic ecosystem variables, dealing with physical, structural, functional and landscape features (UN SEEA EA, 2021), such as: water parameters, water extent and level, phytoplankton blooms, aquatic vegetation composition (including diversity), functioning of primary producers. We will present quasi-operational examples showing monitoring spatial and temporal dynamics of freshwater and wetland ecosystems in a multidimensional, integrated framework, developed over selected case studies in Italy. In particular, the potential of EO is demonstrated for: i) assessing temporal evolution of key ecosystem variables – water turbidity, submerged macrophyte cover and riparian vegetation phenology - in perialpine Lake Mezzola, in response to environmental stressors; ii) mapping intra-annual dynamics of primary producers - phytoplankton abundance and floating/emergent macrophyte density - in eutrophic Mantua lakes system; and iii) near real-time detecting of anomalous features of water quality (Chlorophyll-a concentration) and aquatic plant communities (cover and density) along the growing season over shallow, turbid Lake Varese.

Blue carbon ecosystems (i.e., mangroves, seagrasses, and tidal marshes) have a critical ecological, societal, and economic role due to their ability to store carbon, contribute to climate change mitigation, protect the shoreline, improve water quality, support livelihoods, and provide shelter, food and habitat for numerous species. Assessing the condition of these ecosystems is critical because it reflects their ability to provide ecosystem services, including carbon sequestration. Condition assessments can also help guide the development of management strategies and priorities, including ecosystem restoration and nature-based solutions. They help identify pressures, ecosystem resistance, and resilience, contributing to understanding the effect of natural and anthropogenic drivers of change, and informing monitoring and reporting against local and international policy objectives (such as the Aichi Biodiversity targets or the Sustainable Development Goals). In the context of blue carbon ecosystems, they might also guide financing decisions because those depend on risk analysis and the measurement of outcomes. Notably, one of the core accounts within the System of Environmental-Economic Accounting – Ecosystem Accounting (SEEA - EA) is the ecosystem condition accounts, which track the state of ecosystem assets using selected metrics or indicators. As the SEEA – EA is an inherently spatial framework, spatially explicit condition metrics are critical. Thus, remotely sensed approaches can be used to obtain such data. A literature review was performed to support the definition of metrics for developing blue carbon ecosystem condition accounts. The condition indicators obtained by earth observation were compiled, categorized, and assessed regarding their advantages and disadvantages. Notably, the most used metrics for ecosystem condition using remote sensing approaches are temperature, salinity, percentage vegetation cover, and productivity. Indicators related to biodiversity, which depend on species identification, are challenging and demand intense ground truthing efforts, thus, tend to be less successful.

With humanity facing biodiversity loss, climate change, and escalating pollution; now is the time to act, as stated by UN Secretary-General António Guterres. With the UN Decade on Ecosystem Restoration ending at the same time as the Sustainable Development Goals (SDGs), restoration interventions need to be assessed in a systematic and objective manner to maximize the global community's progress towards these goals. However, the high-quality data records of essential climate variables that are required for this are often lacking in both space and time. Satellite data products can fill this gap, as they can quantify impact by detecting and attributing environmental changes consistently over space and time. Within the Restore-IT project supported by ESA (Contract No. 4000136484/21/I-DT-lr), the aim is to provide an impact monitoring tool that enables asset managers and green investments funds to conduct ecosystem accounting and communicate transparently on effectiveness towards their stakeholders and the EU in the context of several SDGs (12: Responsible production and consumption, 13: Life on Land, and 15: Climate Action). One of the case studies within Restore-IT focuses on restoration of degraded lands in Kenya and Tanzania. By combining climate data records of planetary variables (e.g. soil water content, surface temperature and NDVI) with high-resolution optical imagery (e.g. PlanetScope), we find promising results that show the effectiveness of restoration activities. By comparing the records of the intervention areas statistically to the records of reference areas, we have computed the number of liters of water retained by the topsoil, the number of degrees of temperature the soil cooled down and the percentage of vegetation greenness increase as result of each restoration activity. Hence these near real-time, consistent satellite data streams have shown to be a useful source of information as a base for reliable ecosystem accounting.

National-scale status assessments of ecosystems requires spatially explicit information about the distribution and change of a range of ecosystem types. However, there remains a lack off-the-shelf spatial data on ecosystem distribution and change for the vast majority of the world, hindering efforts to develop ecosystem accounts, investigate their changing status, and taking stock of natural environments. In this talk I will detail the earth observation analyses developed to assess the status of Myanmar’s ecosystems under the IUCN Red List of Ecosystems protocol. We developed (i) a national ecosystem typology to structure the assessment, (ii) a new, national-scale map of terrestrial ecosystems from spectral and radar earth observation data to estimate the extent of each ecosystem, and (iii) a range of analyses to assess the condition of all of Myanmar’s terrestrial ecosystems. The Myanmar ecosystem map utilised around 60,000 point-occurrences of ecosystem types and more than 70 publicly available biophysical (e.g. slope, elevation), passive and active sensor data (from the Landsat Archive and Sentinel) and climatic (precipitation, temperature) datasets to model the probability of occurrence of each of the >60 ecosystem types in the Myanmar ecosystem typology. Our condition assessment included a broad range of publicly available datasets and extensive outreach to the SE Asian ecosystem science community. Our results, which are free and in the public domain, indicate widespread data deficiency and highlight the ecosystem most at risk of loss in the near-term future. Our generalised methods of rapidly mapping ecosystems at national scales can be quickly adapted to other countries given sufficient training data, and represents a key step in scaling up the IUCN Red List of Ecosystems to the global domain.

Earth observation tools have provided an opportunity to facilitate studies of land degradation due to human and natural disturbances. These tools have saved field, and laboratory costs and provide us with an accurate estimate of lands condition on a large scale. In this study, terrestrial ecosystem health has been quantified in the form of the sustainable development goal 15.3.1 over a period of thirty years using Landsat images in rangeland and semi-arid forest in central Iran. We tested the transferability of quantitative remote sensing indices from the recent time (2016) to the past time (1987) and assessed the state of terrestrial ecosystem health. The study area was classified based on a standard rangeland health field protocol as healthy, at-risk, or unhealthy. We compared the patterns of terrestrial ecosystem health considering the challenges of monitoring landscapes using Landsat images. SDG 15.3.1 is calculated as the proportion of land that is degraded relative to the total land area in the spatial extent (km2). Compared to 2016, there were 23 healthy sites, 11 unhealthy sites, and six at-risk in 1987. When transferring our remote sensing approach to assess terrestrial ecosystem health from 2016 to satellite data from 1987, we found 32% of degradation. Freely available earth observation data helped us assess the current health status and obtain an estimate of the past health status based on old satellite images despite the lack of past field information. These outputs not only emphasize the identification of land conditions using remote sensing but also lay the groundwork for future research in terms of the calculation of economic loss and profit from land destructions or restorations.

Monitoring vegetation health is key to understand the impact of human activities on the environment through direct or indirect causes such as deforestation or climate change. By quantifying vegetation status with explicit spatial and temporal information, stakeholders are able to take efficient and targeted nature conservation measures. Earth observation technologies such as satellite imagery has proven to be a reliable approach to meet these needs. Part of the ESA Copernicus program, Sentinel-2 satellites provide multispectral medium resolution imagery with a high revisit time of five days, while Sentinel-3 SLSTR satellite delivers spatial resolution of 1 km² with a high revisit frequency of one day. Based on these satellites, indices can be calculated to highlight specific features. For instance, the Normalized Difference Vegetation Index (NDVI) is used to highlight live green vegetation whereas the Land Surface Temperature (LST) allows to measure temperature at the earth surface. We propose an enhanced approach of the Vegetation Health Index (VHI) to give localized and up-to-date information of 10*10 m² combined with an update frequency of 10 days. This method consists of comparing the current NDVI and LST values to the range of values observed in previous years. The VHI gives an idea where the observed value is situated between the extreme values (minimum and maximum). Lower and higher values are used as a proxy to indicate the critical and optimal vegetation state conditions, respectively. This approach gives fast, explicit, and localized information about vegetation health. We used the ERA5 climate reanalysis database (ECMWF) for validation purposes. We selected total precipitation, evaporation, and surface temperature variables, used as a proxy for vegetation health as vegetation is vulnerable to lack of water and high heat. By correlating this data with our VHI values, we were able to validate our approach.

Since the civil conflict and war that ended in 2003, Liberia’s remaining forests have been under increasing pressure from extraction activities and development such as logging, wood harvesting, mining, forest clearing for agriculture, and housing and urban development. Earth observations have become the principal tool for monitoring gains and losses in forest area over time. However, not all standing forest is equal and other measures that quantify changes in forest health and condition are particularly useful for designing effective policies and conservation actions on the ground. In this study, we quantify forest fragmentation in Liberia between 2000-2018 using two metrics: 1) number of forest fragments, and 2) average size of fragments. We use the Theil-Sen non-parametric regression to evaluate changes in annual fragmentation for each metric across the 19-year time series. The results show an exponential increase in the number of forest fragments across Liberia starting in 2015 and a corresponding exponential decrease in average fragment size as larger forest patches become increasingly divided. The findings highlight the need to understand and address the yet unknown drivers of fragmentation that dramatically increased in 2015. We discuss pros and cons of these metrics for ecosystem condition accounts.

talk

Ecosystem Services (ES) accounts, as framed and assessed in INCA (Integrated system for Natural Capital Accounting), are able to provide information on the interaction between ES Potential and ES demand. This interaction can generate a match, that represents the ES actual flow from ecosystems to economic units. This interaction can also generate mis-matches, that can be caused by: (i) the overuse of ES that eventually lead to ecosystem degradation, (ii) the absence of the needed ES that leave the ES semand unsatisfied, and finally (iii) the ES missed flow. Mismatches can be used to build ES vulnerability accounts that in turn enable to map risk. Because ES are allocated to economic units, ES vulnerability accounts can be directly “allocated” to the economic units that will eventually suffer from this increased ecological risk. This presentation is meant to provide examples for each of the mismatch categories that are collected by ES vulnerability accounts to explore how this information can be structured in map risk.

Within the INCA platform (Integrated system for Natural Capital Accounting), accounts have been developed for 9 ecosystem services for the EU following the SEEA EA standards, adopted by the UN. Based on INCA models, a QGIS plug-in tool has been developed to support practitioners and national statistical offices to account for ecosystem services. The tool will be demonstrated, highlighting the Earth Observation (EO) inputs required, with an example for the model of on-site soil retention. To achieve consistency in ecosystem accounts for decision-making, it is necessary to produce comparable time series, which allows to identify trends in time and to assess (for each ecosystem service) outputs for different countries or regions. However, not all required EO inputs are available for the requested accounting period and the continuation of these inputs is not always guaranteed, this leads to artificial trend breaks due to data gaps. Alternative datasets could be considered to keep the integrity of the time series of ecosystem accounts; however, sometimes it is necessary to combine different datasets towards a consistent series. Since the models developed under INCA rely heavily on EO data, a methodology will be presented to deal with this data variability. Extrapolation to fill data gaps could lead to errors in the estimation of the models and eventually decrease the confidence for comparability, but is sometimes unavoidable as will be highlighted with the ecosystem extent. The results of this methodology can be useful for policy makers, as well as the EO industry to understand the critical products that can aid in the calculation of future ecosystem accounts. INCA is a project funded by the European Commission and developed with the support of the Joint Research Centre.

Mapping of ecosystem service is useful for decision making. The ARIES (Artificial Intelligence for Environmental Sustainability) project focuses on integrated modeling of ecosystem services and is led by the Basque Centre for Climate Change, Spain. Under the ARIES project, k.LAB platform is developed as an integrated modelling platform which includes several ecosystem services related models on carbon stock, flood regulation, sediment, outdoor recreation, etc. In k.Explorer and ARIES for SEEA Explorer, two user interfaces of the k.LAB platform, several global scale ecosystem service models have been already included and used. Then, in collaboration with BC3, the member of k.LAB Japan have been developing several ecosystem service models which were customized for Japanese specific situations including crop and carbon stock models. Regarding the crop model, the FAO’s AquaCrop model was simplified and modified by using remote sensing data. For agriculture land, we firstly studied the relation between vegetation canopy cover and Normalized Difference Vegetation Index (NDVI) by Unmanned aerial vehicle (UAV). By using this relation, a test version of the simplified crop production model was developed. This model can also estimate yield and residues on a field. Also, carbon stock model has been developing by using field survey and remote sensing data sets. These models were imported into k.Modeler under the k.LAB platform as separate models and used from k.Explorer. In this presentation, the current development of the models under k.LAB platform will be presented.

As one of NASA's open and free data systems, NASA's Land, Atmosphere Near real-time Capability for EOS (LANCE) provides a wide range of near real-time, low latency and expedited data products from NASA and other earth science satellite missions to support users in their understanding of the value and distribution of ecosystem services. LANCE data can support the System of Environmental Economic Accounting-Ecosystem Accounting (SEEA-EA) framework by both facilitating the creation of ecosystem spatial extent and condition accounts, as well as the quantification and mapping of ecosystem services, much more quickly than routine data processing allows. During the past 13 years, LANCE near real-time satellite data products (e.g., surface reflectance, albedo, vegetation height, thermal anomalies, soil moisture, snow cover etc.) have been used to produce ecosystem-related indicators such as vegetation indexes, biomass, land cover maps, fire, and flood products. These ecosystem-related indicators can be integrated into ecosystem services models, tools, and platforms to monitor land cover and land use changes in ecosystem composition in support of land and natural resource management. When using ecosystem services accounts in decision-making, low latency is essential because having up-to-date information can be critical to support decisions being made. With adequate spatial resolution, long time series length, and low latency data products, LANCE would promote the sustainable use of land and natural resources.

This study describes an empirical approach for determining coastal protection value of mangrove ecosystems against coastal erosion in Sundarbans, India. Similar studies usually focus on flooding impacts and rarely assess erosion. Mangroves can substantially reduce vulnerability and risk by providing natural protection from flooding and erosion, which is important to millions of people in the Sundarbans. Accurate assessment of coastal protection benefits can influence conservation and restoration of mangroves, while informing coastal management decisions.Satellite data going back several decades is used to map coastal erosion and determine areas of mangrove ecosystem loss. A causal relationship between these phenomena are established by evaluating environmental factors and triggers and by comparing with nearby locations. Loss of mangrove ecosystems is seen to be the leading cause for coastal erosion, particularly where competing anthropogenic land uses, such as aquaculture and agriculture have likely resulted in the destruction of mangroves. Coastal protection value is determined based on coastal erosion loss prevention over a 20 year period, which would be enabled by the restoration and conservation of mangrove ecosystems. An expected damage function is used to determine protection value based on the land, buildings and infrastructure exposed to coastal erosion hazards. Values of coastal protection can vary widely, which is related to the value of coastal assets and land use behind mangrove ecosystems. These assets and the land use are mapped using high resolution satellite data to provide accurate and hyper-local coastal protection values. This study provides the key decision pathways and the critical information needed to move from planning to action. It facilitates the inclusion of natural coastal defences into national development and resilience strategies and shows mangrove restoration and conservation to be a cost effective solution for coastal protection.

Global agricultural systems are amongst the biggest contributors to ecosystem degradation. Much needed transformations towards more sustainable agriculture will require building capacities for accurate monitoring and accounting of multiple ecosystem services across spatial scales. Rapid developments and creative innovation in earth observation and process-based modelling approaches can contribute to filling this gap. The DAKIS (or Digital Agricultural Knowledge and Information System) project integrates digital technologies, including earth observation products, to support improved agricultural ecosystem service provision through i) ecosystem service accounting; ii) evaluation of trade-offs; and iii) generation of land management recommendations. To test this approach, we evaluated multiple spatial datasets in two 5 x 5 km landscape windows in Brandenburg, selected to represent variation of biophysical and anthropogenic factors in the study region (e.g., soil type, climate, field size, land cover). Based on crop-specific land-use maps, we integrate process-based models, remote sensing imagery and spatial analysis to derive maps quantifying agroecosystem service potentials using specific indicators (e.g., soil erosion risk, biodiversity and pollination potentials, yield productivity). We then apply a trade-off analysis to evaluate ecosystem service provision and to formulate future-oriented land-management recommendations.

Although fisheries represent about 10% of Liberia’s economy annually, they have not been systematically measured using a consistent approach or on an annual basis. Using the System of Environmental Economic Accounting Ecosystem Accounting (SEEA EA) framework we quantified – in physical and monetary terms – the contribution of marine/coastal ecosystems to the provision of fish biomass and their input to the economy. We developed a marine fisheries supply table for the years 2016-2021 based on fishing activities in the Exclusive Economic Zone (EEZ), where commercial fishing is permitted. In recent years, due to tremendous advances in geospatial bigdata availability in the public domains, it is now possible to get a more accurate picture of industrial fishing activities. Global Fishing Watch (GFW), a cloud data platform, provides near real-time satellite monitoring data of industrial vessel locations and their movements. With the help of this data source, along with ground-based measurements such as catch per unit effort (CPUE) and market price, we were able to estimate industrial catch for each vessel operating in Liberian waters. We found that in addition to their national economic contribution, fisheries contribute to livelihoods of local communities through food security and employment generation. Besides measuring and valuing fisheries supply, we also analyzed spatial and temporal variation of industrial fish catch in relation to seafloor geomorphological classes, as a proxy for marine ecosystems, that can be used for policy decisions such as preventing overfishing in certain areas through marine spatial planning. In addition, the methodological approach developed in this analysis will help the government measure and monitor fishing activities using a standardized ecosystem accounting approach.

Reliable quantification of ecosystem services provision in agricultural landscapes depends on accurate mapping of the spatial configuration of land use and land cover (LULC). This talk will present experimental results on the benefits of higher spatial image resolution and better thematic resolution in land cover mapping for the estimation of ecosystem services. The choice of suitable land cover data is crucial for the production of relevant knowledge to inform natural capital-based land-use policies. A LULC map of a sub-catchment of the River Welland in England was produced from Copernicus Sentinel-2A and 2B satellite images at 10 m resolution. The satellite images were processed using Google Earth Engine. The 10 m map was compared to LULC datasets at 20 m, 25 m and 100 m resolution. Ecosystem services provisions of crop yield, carbon storage and pollinator abundance were estimated from each map using the 'Integrated Valuation of Ecosystem Services and Tradeoffs' (InVEST) model. Spatial resolution had a significant effect on the abundance and spatial configuration of land cover types. For example, detected woodland cover in the finest resolution dataset was twice that in the coarsest data. Finer spatial resolution also allowed small, fragmented patches of woodland and grassland to be identified. The finest resolution map resulted in 21% lower predicted wheat production (due to lower estimates of cultivated land cover), 7% higher predicted carbon stocks and 43% higher predicted wild bee abundance compared to the coarsest resolution map. The estimated monetary value of ES provision increased by 23.2% between the 10 and 100 m dataset. We recommend that a LULC resolution of 10 m or finer should be employed in agricultural landscapes to accurately preduct ecosystem services provision for the development of future natural capital policies.

Cultural ecosystem services (CES) are some of the most significant benefits nature provides to humans, but often suffer from poor understanding and quantification, due to inconsistent conceptualizations and methodological challenges. In this research, we first examined and proposed a CES theoretical framework, and then relying on this, we developed a method employing location data that can be extended and generalized to quantify CES. This method considers both visitor arrivals and visitor satisfaction, which is different from other CES evaluation methods, like social media photos used to evaluate the concentration of visitors,or questionnaire mainly focusing on visitors’ attitudes.Then we used Nagoya City, Japan, as a case study and found that Nagoya Castle, Atsuta Jingu, and Tsuruma Park accounted for 66% of CES in Nagoya City. Among the various CES subtypes, cultural heritage, aesthetic, and mental health were the three most perceived. With the proposed methodology, CES can be evaluated in a more objective and comprehensive fashion; compared across space, time, and cultural diversities; and further integrated into ecosystem services assessments.

Sitting at the interface between land and sea, nearshore water bodies are a key part of the UK’s natural capital. They are the location of multiple marine industries, the primary feature for coastal tourism and recreation, and a critical home for diverse and productive biodiversity. Healthy nearshore waterbodies support productive fisheries, coastal economic activity and wellbeing and moving towards Net Zero. However, nearshore water bodies are heavily influenced by land-based activities and require an integrated approach between land-based management and marine conservation. Nutrient enrichment, turbidity, sedimentation, and pesticides all affect the resilience of UK coastal and marine systems, which can have negative impacts on the marine Natural Capital assets and delivery of ecosystem services at both local and regional scales. The combination of a changing baseline from climatic shifts in weather patterns, alongside the chronic stresses of reduced water quality, coupled with other cumulative pressures may reduce the ability of these systems to support and provide marine ecosystem services. We have developed a supervised classification of true-colour satellite imagery to map the extent of estuarine and coastal plumes and the associated areas of freshwater influence around the UK coast, using the Forel Ule index. Monthly composite Forel Ule values have been derived between 2017 and 2021 and compared with in situ water quality data (inorganic nutrients, suspended particulate matter). Using the relationship between Forel Ule class and water quality, we can map the water quality with greater spatial resolution than conventional in situ monitoring. A cost-distance based function has been applied to produce spatially distributed pollutant loads across the plume and assess the risk of coastal habitats to pollutant exposure. Using this approach, we have established a framework to visualise the exposure of water column and benthic habitats to land-based pollutants.

The System of integrated Environmental and Economic Accounting- Ecosystem Accounts (SEEA EA) has been adopted as standard by the United Nations Statistical Commission in March 2021. The INCA (Integrated system for Natural Capital Accounting) project has developed an operational approach to implement in practice the SEEA EA framework for ecosystem services Supply and Use Tables (SUT). Once these accounts are available, there could be many ways to process such information. Three main cases can be identified: (i)               indicators which are derived from descriptive statistical data – they are characterized by the fact that any practitioner can use data without any further processing. A range of information can be extracted by the tables as they are; (ii)              indicators which are derived through combining and processing descriptive statistical data – data extracted by SUT need to be further processed to obtain the desired outcome. The degree of complexity of each indicator can greatly vary. The outcome obtained is “final”; (iii)            indicators which are derived through analytical work based on statistical data and methods – data extracted by SUT need to be further processed to obtain the desired outcome. In this case the outcome is not an indicator per se, but it represents an input for further computation. Skills concerning the tools/models where the ES accounting input will be used is a pre-condition. We will provide examples for these three cases with a special reference to the use of earth observation as input data to process SUT and Sustainable Development Goals (SDGs) as final set of indicators.

For years there have been indicators to assess the development of countries based on socio-economic statistics. One example is the HDI (Human Development Index) which ranks countries according to human, economic, health and educational data. The environment had no place in these indicators. However, the following hypothesis have been verified by multiple studies: the higher the HDI, the greater the pressure on the environment. Tourism is a pillar of the modern economy and a significant vector of human development. The number of international tourists hit the 1 Billion bar in 2020 and is forecasted to rise to 1.8 Billion in 2030 (despite the halt related to Covid-19), making it crucial to find efficient ways to handle this growth and preserve the fragile destinations. We present here the implementation of a sustainable development indicator for the tourism industry, the TSDI (Tourism Sustainability Development Index). The TSDI combines human development and ecological factors. The ecological factors include EO satellite data, especially air quality from Sentinel-5p, water quality from Sentinel-3 and vegetation health from Sentinel-2. The use of remote-sensing data is a key differentiator : space data is a reliable and objective measurement that can be systematically and homogeneously applied anywhere. Space data is combined with other sources (CO2 emissions, biodiversity data…) to provide a complete picture of the environmental state of an area. The human development factor includes urbanization and tourism activity as well as classical human development indicators already included in the HDI such as education index and life expectancy. The TSDI includes a notion of boundary enforcing the idea that if an area is doing well along one dimension, it is not allowed to do worse on other parameters, limiting biases and favoring areas where all sustainable development factors are under control.

The impact of tourism on the environment has been recognised since 1980 by the OECD. The effects on the environment are multiple: water and air pollution, pollution of sites by waste disposal, noise pollution linked to traffic, reduction of natural and agricultural areas, destruction of fauna and flora, degradation of landscapes, etc. UNWTO stated that 95% of tourists gather in 5% of the world. We present here the concept of an environmental thresholds system able to compute the environmental carrying capacity of a territory, either on the short term (for tourism activities) or on the longer term. Measurements of the environmental state are used to define capacity gauges for managing tourist flows in a site, beyond which there is a risk to the environment. Satellite earth observation, in-situ sensor and modelling data are used to quantify the state of the environment in the area of interest. We focus the quantification on air quality which varies on the same temporal scale (hour by hour) as human movement, on the evolution of vegetation health that follows a slower pace, and on the quantity of water available in the territory in nominal consumption versus consumption in tourist periods. Thresholds are defined using established benchmarks, such as the WHO air guidelines for air quality or common practices. Regarding human flow data, these are captured by mobile phones. These data allow the identification of the number of people in the area of interest and their profile - day-tripper, resident, transient or visitor. This categorisation, offered by mobile data providers (for instance Orange in France) allows a standard profile of the study area to be established. Thus identifying and quantifying any fluctuation around this standard situation allows us to isolate the contribution (or subtraction) of human movements on our indicators.

talk

The Sustainable Development Goal (SDG) 11 of the United Nations (UN) aims at renewing and planning human settlements in a way that offers opportunities for all, including access to essential services, housing and energy, green public spaces, transportation while reducing the use of the resources and impact on the environment. In this context, accurate, reliable and frequent information is needed to comprehensively characterize human settlements. To this purpose, the increasing availability of Big Earth data from satellites and related analytics tools has recently opened novel opportunities. However, in the last few years, this has led to the generation of several global layers, primarily focusing only on delineating the actual settlement extent, often with limited quality. To overcome this limitation, the German Aerospace Center (DLR) in collaboration with ESA and the Google Earth Engine team has been generating the World Settlement Footprint (WSF) suite. In particular, the WSF is an unprecedented collection of open-and-free global datasets aimed at advancing the understanding of urbanization at the planetary scale, which have been derived exclusively employing publicly available data as Sentinel-1, Sentinel-2, and Landsat imagery, as well as the ALOS World 3D and Copernicus GLO-30 Digital Elevation Models. At present, the suite includes: i) the WSF2015 and WSF2019, two binary mask outlining the 2015 and 2019 settlement extent at 10m resolution; ii) the WSF2019-Imperviousness, a 10m resolution layer estimating the settlement percent impervious surface; iii) the WSF-Evolution, which outlines the settlement growth at 30m spatial resolution on a yearly basis from 1985 to 2015; iv) the WSF3D, a 10m resolution product estimating the height of built-up areas and v) the WSF2019 population, which estimates the number of inhabitants per pixel at 10m resolution. All of them exhibited remarkable quality as quantitatively assessed by extensive dedicated validation campaigns supported by crowdsourcing.

The complexity and variability of the built environment makes it difficult to create accurate and consistent urban ecosystem accounts using moderate-resolution data. High-resolution (i.e., 1 m) land cover products are some of the most reliable datasets available. However, most of these data are only available for select cities, so national-scale urban ecosystem accounting in the U.S. relies on the National Land Cover Database (NLCD, 30 m). The U.S. NLCD currently provides the only nation-wide tree canopy cover (TC), land cover and impervious surface datasets, making it an essential resource for natural capital accounting. Previous efforts have succeeded in assessing and improving the accuracy of these data using a decision tree model to evaluate NLCD TC in 27 U.S. cities. As ecosystem services that underpin accounts can vary depending on regional considerations, we have developed a refined analytical model covering more U.S. ecoregions, expanding from 27 to 78 cities. In our new model, hyperparameter tuning reduced error by 5.5%; removing some inputs without nation-wide coverage and low variable importance reduced error a further 11.1%. With our new model, TC corrections have improved with an overall R2 of 0.80 (previous models R2 was 0.765). Modeled TC corrections address important limitations in NLCD TC. For example, our predictive maps show that NLCD TC underestimates TC in urban residential and riparian areas. Preliminary corrected model results show an increase in tree canopy cover in arid urban areas, where NLCD TC greatly underestimated TC. Improved accuracy of TC in hot, arid cities and riparian areas can improve our estimates of ecosystem services, including cooling benefits sometimes underestimated by tens of millions of dollars using the currently available NLCD TC. Preliminary results show the utility of incorporating high-resolution and NLCD products to create a more accurate TC dataset for urban ecosystem accounting in the United States.

Under the new Census of Environment program, Statistics Canada has begun implementing ecosystem accounts following the SEEA EA framework. Preliminary work has been started on urban ecosystem extent and condition accounts. Delineating the urban extent is important to track how the physical urban form is changing over time and the implications for both the people who live there and the surrounding environment. Existing spatial boundaries of urban areas are based on administrative data and tend to include other land uses and land covers within the boundaries. In response to the need for an improved urban extent, Statistics Canada is developing “Contiguously Settled Area” boundaries. These boundaries, based on Landsat-derived land use data, will represent areas of consolidated settlement and will exclude tracts of natural land on the periphery of the core settled area. As urban areas develop, natural land covers are replaced with artificial surfaces used in the construction of urban elements. Vegetation and natural components in urban environments provide essential ecosystem services such as improving urban air quality, mitigating urban heat island effects, reducing or delaying storm water runoff, providing wildlife habitat and providing recreational opportunities and aesthetic benefits. Monitoring the condition of urban areas is important for the people who live there and depend on services provided by urban ecosystems. To that end, Statistics Canada used the normalized difference vegetation index (NDVI) from MODIS as a measure of urban vegetation condition in Canada’s urban statistical boundaries, known as population centres. This presentation will describe Statistics Canada’s approach to measuring urban extent and condition from EO data, current limitations, and planned methodological improvements.

Air quality has significantly degraded over the last century. Human activities and growing population in urban areas impose significant environmental pressure on surrounding ecosystems including the degradation of surrounding air quality. This degradation has a significant impact on human health but also on ecosystems, causing stress on the fauna and flora that compose them. The understanding of the air quality drivers and their dynamics is the key to further develop efficient risk mitigation strategies and preserve the ecosystem health. The satellite data obtained through the European Copernicus programme (i.e. through Sentinel-5P) and the Copernicus Atmosphere Monitoring Service’s (CAMS) European air quality forecast allow us to highlight the concentration of pollutants like NO2 and particulate matter near the surface of the earth with a resolution of 0.1 degrees (around 10km at the equator level). This paper proposes a downscaling approach from CAMS air quality modelled datasets (10km*10km) to improve the original resolution to a 1km*1km resolution. The objective is to move from city level to neighbourhood level in urbanised areas to support the identification of air quality drivers, link it to human activities and better evaluate the impact on surrounding ecosystems. Our model is fed with a probability map built using land cover information and augmented with meteorological data (temperature, precipitation and wind) and topographic data (i.e. digital elevation model). The model is then validated against in-situ measurements from openly available stations. This approach allows the identification of the major drivers to NO2 concentrations in a given area, and subsequently allows training a high resolution model to downscale the satellite measurements.   The generalisation of the developed approach is explored where we apply the same methodology on different geographical locations and assess the performance by comparing with available in-situ validation measurements.

The GAUSS project aims to provide specific demonstrations of the use of Earth Observation together with in situ data coming from a range of digital sensor networks to meet key needs of selected National Statistical Offices (NSOs). One key need that NSOs have is to provide data on the extent and quality of ecosystems at a local level. This is important for assessing natural capital statistics, but also has a key impact on the wellbeing of people living in these areas. Better means of defining such ecosystems and of providing faster and more accurate data on the current conditions are much needed by NSOs. To meet this need, IGIK has been working in collaboration with Statistics Poland to build a suitable service providing data on the conditions of green areas at comune level. The service is based on data from Sentinel 2, Sentinel 3 and MODIS, together with local data on land use and boundaries. It outputs data on ecosystem extent and condiiton in a format suitable for use by NSOs for their mandated reporting. It has been created for Poland but is easily transferable to other countries. GAUSS is also working with EUROSTAT and selected National Statistical Agencies to define a roadmap for future adoption of EO data in the collection of national agencies. Ecosystem accounting is a key interest of such agencies and so the project offers a key forum for discussing how this can be adopted in practice. GAUSS is funded by ESA and is undertaken by a consortium of NOA, FMI, IGIK and Evenflow. Overall the project looks at 4 different uses cases covering statistics on air quality, green indicators, snow extent and inland water bodies.

Urban ecosystems are a complex mosaic of land covers and changes often occur on such fine scales that they may not be detectable using medium to high resolution satellite imagery.   Ecosystem extent accounts rarely report the statistical confidence in detecting changes. The complexity and dynamics of urban environments makes change detection extra challenging. In this study, we used direct change mapping with a Random Forest model and Sentinel-2 imagery to explore the interplay between classification accuracy and size of basic spatial unit for change detection in the built zone of Oslo, Norway between 2015 and 2019.   We show how this type of change mapping can be used to address one of the main purposes of ecosystem extent accounts - namely to detect significant change in ecosystems. We argue that classifying changes rather the differences between opening and closing balances of ecosystem extent increases change detection capability. The Oslo accounts show a pattern of densification and construction across most land use classes between 2015 and 2019, with the biggest changes from tree canopy to low vegetation (125 ha), and for low vegetation to impervious surfaces (106 ha). Our results indicate that Sentinel-2 based extent accounts are able to detect change for urban areas over a 4 year period. Quantifying the classification accuracy at different scales also allows urban planners and managers to determine the level of accuracy that they are willing to accept in ecosystem accounts, and balance it with the types of changes they are interested in detecting. We argue that ecosystem extent change classification could be a standard complement to reporting opening and closing ecosystem extent balances .

talk

Covering 35% of Europe’s land area, forest ecosystems play a crucial role in safeguarding biodiversity and mitigating climate change. Yet, forest degradation continues to undermine key ecosystem services that forests deliver to society. We provide a spatially explicit assessment of the condition of forest ecosystems in Europe based on the recommendations of the System of Economic Environmental Accounting - Ecosystem accounting. We show that the condition across 44 forest types averaged 0.58 on a scale from 0 to 1 where 0 represents the condition of a degraded ecosystem and 1 represents a reference condition based on observations in primary forest or forest in protected sites. This is based on seven ecosystem condition variables: water content based on the normalized difference water index, soil organic carbon, species richness of threatened forest birds, tree cover density, forest productivity based on the normalized difference vegetation index, forest connectivity, and landscape naturalness. Six indicators are directly or indirectly underpinned by remotely sensed earth observation data. Forest productivity and connectivity are comparable to levels measured in undisturbed or least disturbed forests. One third of the forest area was subject to declining condition, signalled by a reduction in soil organic carbon, tree cover density and species richness of threatened birds. Our findings suggest that forest condition is affected by human and natural disturbances, and that forest ecosystems will need improvements in management and an extended period of recovery to achieve a condition that comes closer to reference conditions.

Ecosystem condition is not a straightforward concept, often requiring consideration for a range of both abiotic and biotic conditions. However, the use of reference information to determine ecosystem condition, and thus determine restoration potential as well as climate impacts on certain ecosystem types is well established in the literature. A fundamental consideration when selecting and using reference information is to consider the temporal and spatial aspects which result in natural variation. However, approaches and requirements are often site-specific and differ depending on the condition variable at hand. We use a spatial approach to estimate ecological condition for forests in Norway by using the Potential Natural Vegetation concept and applying it to satellite-based NDVI. We used reference areas (i.e. protected areas) to define a reference NDVI state, and built a Random Forest regression model that could predict potential NDVI based on climatic and edaphic characteristics. This model was then used to predict the potential NDVI (pNDVI) outside of the reference areas. The ecosystem condition is then defined as the difference between the pNDVI and the observed NDVI for the entire forest extent. In some ecosystem types and for some management purposes, it may be more useful to compare a region of interest with temporal reference from the same region, i.e. to calculate ecosystem condition based on the magnitude of increase/decrease in a specific variable over a period of interest. We used this approach for mountain ecosystems in Norway. Here, the indicator was calculated as the slope of a linear regression of annual NDVI values during the growing season against year. For mountain ecosystems, the reference condition was defined by no significant trend, that is, the reference value for the indicator is 0. However, this reference condition may vary depending on the ecosystem type considered.

Estimates of forest above ground biomass (AGB) density are a useful indicator of the bio-physical condition of ecosystems. In the context of ecosystem accounting, AGB estimates have been experimentally used to assess both the ecological condition of ecosystems, and the potential of carbon storage as the avoided flow of carbon dioxide to the atmosphere. In the ideal scenario, the data needed for ecosystem accounting should be generated annually using consistent methods and comparable measurements. Regular in-situ measurements in forest are however unfeasible; they are often time-consuming, costly and their accuracy hindered by suboptimal sampling. The use of earth observation (EO) data to achieve time-series consistency and transparency is therefore highly desirable. However, the use of the suite of space-based AGB-map products published to-date is challenging; the maps have vast differences not only in the estimated forest AGB and associated uncertainties, but also in their underlying scope, definitions, assumptions, and level of methodological transparency. With a few exceptions, no map has been released by space-agencies at regular annual intervals. These problems impede their generic use for ecosystem accounting, where the consistency of results in between the accounting periods is key for compiling accurate accounts. Of the currently available published EO-based AGB maps, the European Space Agency’s Climate Change Initiative Biomass maps holds some potential for use in ecosystem accounting – the maps have a spatial resolution of 100 m, have been published using a consistent methodology for years 2010, 2017, 2018, and will continue to be released for the years following 2020 in due course. In this study we demonstrate the use of this dataset for assessing ecological condition of forest ecosystems in Liberia. We show the change of condition and compile carbon-stock accounting tables for the 2010 – 2018 period. Finally, we discuss uncertainty of the results.

Global Forest Watch (GFW) is a free forest monitoring system that provides timely and actionable information to support the sustainable management and conservation of forest landscapes. Since its launch in 2015, more than 5 million users have visited the GFW website from every country in the world. GFW is used primarily by civil society organizations, journalists, communities, governments, and companies around the world to see where, when, and why deforestation is happening so that they can take action to address it. GFW uses hosts open and free data, which has been peer reviewed for quality. Tools, guidance materials, and analyses are provided to enable users of all technical abilities to benefit from the data. Most of the data on GFW are Earth Observation data, allowing for consistent analyses at the national or sub-national level. Forest data which is surfaced on GFW includes forest extent at 30-meters for two time periods, and annual forest loss data. In the near future, tree cover gain and loss will be available annually based on changes to tree height (including historical data back to 2000), providing information on ecosystem extent change as well as degradation and enhancement. Contextual data such as land cover, tree plantations and protected areas provide information on drivers of change. Condition of the ecosystem can be explored using information on primary forests, intact forest landscapes, forest biomass and tree height.

The Common Agricultural Policy (CAP) is responsible for allocating agricultural subsidies within the European Union. This policy fundamentally shapes the European agricultural landscape. Among these subsidies is the aid for Sustainable Mediterranean Agrosystems (AMS) for old olive trees, which is granted to olive groves that are more than 60 years old. To validate such aid, it is necessary to establish how long an olive tree has been cultivated in a specific location. Therefore, olive trees must be detected and tracked by photo-interpretation of aerial images. Traditionally, this task has been performed manually by remote sensing experts, entailing high costs. However, recent advances in Deep Learning open up the possibility to automate this task. Accordingly, we propose a novel methodology for the detection, tracking, and categorization of olive trees. First, a semantic segmentation model is used to detect olive trees at each time step. Then, the resulting segmentation masks are further refined to increase the separability between olive trees. Finally, a hyper-temporal tracking of the olive trees is performed in order to categorize each of them according to their age. The results obtained indicate that following the proposed methodology it is possible to automate the olive grove monitoring task, allowing to achieve a more equitable allocation of the subsidies issued by the CAP. Javier Lasheras (1), Christian Ayala (1), Mikel Galar (2)(1) Tracasa Instrumental, Calle Cabárceno, 6, 31621, Sarriguren, Spain, (jlasheras,cayala)@itracasa.es(2) Institute of Smart Cities (ISC), Universidad Pública de Navarra, Campus de Arrosadía s/n, 31006, Pamplona, Spain, mikel.galar@unavarra.es

Forests play crucial role in carbon cycle and climate regulation. In recent times, global CO2 emissions due to wildfires amounted to more than one and a half times emissions of European Union caused by burning of fossil fuels. To assess magnitude of release of carbon compounds stored in plant matter due to wildfires on a global scale, accurate knowledge about above-ground biomass (AGB) is necessary. There have been significant developments to generate global scale AGB maps by utilizing various remote sensing techniques and sensors (e.g. LiDAR). However, most datasets provide data only for specific investigated period and location. Hence, existing global AGB products differ significantly in spatiotemporal patterns. This work combines latest research in AGB estimation with OroraTech's active fire data product to provide operational near real-time monitoring of global CO2 release due to biomass burning. Our research objective is to create biomass inventory capable of generating continuous global high-resolution AGB maps. This study uses sparse GEDI (Global Ecosystem Dynamics Investigation) level-4A AGB data points as ground truth and matches dense Sentinel-2 data to each GEDI data footprint based on spatiotemporal constraints. Supervised machine learning methods are then applied to learn mapping of Sentinel-2 spectral channels to AGB estimates. This will allow generation of dense AGB maps solely based on Sentinel-2 data. Furthermore, we implement techniques to infer amount of CO2 equivalent from an estimate of above ground biomass. By combining these approaches with our burnt area processor, which continuously maps burnt areas based on Landsat-8/9 and Sentinel-2 data on global scale, this enables us to provide estimates of CO2 release caused by wildfires with temporal resolution of 1-3 days. Running operationally, this will generate new, detailed insights in dynamics of wildfire related CO2 release, as well as spatiotemporal development of global AGB distribution with unprecedented temporal resolution and consistency.

talk

Seagrass habitats are essential and vulnerable ecosystems with several key roles, from biodiversity hotspots to climate change mitigation. Their characteristics, health and potential values are the main core of this study. Analysis of data retrieved from habitat modelling and scientific literature will be described, and the results will demonstrate and support the potential ecosystem services that seagrass habitats might provide in the four marine sub regions of the Mediterranean Sea. First attempt of marine accounting will be also presented under the Integrated system for Natural Capital Accounting (INCA), an approach that is fully consistent with the SEEA-EA standard. The marine ecosystem services assessed, valued and accounted are: fish and biomass (for energy) provision, blue carbon and nature-based tourism. Finally, a conclusive discussion on the potential monitoring and conservation of seagrass habitats using Earth Observation (e.g. Copernicus/Sentinel-3) to rapidly and consistently map seagrass distribution and assess condition, degradation rates, and the efficacy of restoration efforts. Such integration would be pivotal for the implementation of conservation, protection and restoration actions in the framework of European legislations. Keywords: ecosystem services, natural capital accounting, marine, seagrass, Mediterranean Sea

Standardised measures of Essential Biodiversity Variables (EBV) are necessary to characterize the condition of marine ecosystems. At the global scale, satellite observations can complement and interpolate in situ observations. A potential approach is the definition of habitats by the physical, chemical or biological characteristics. This concept of dynamic biogeochemical provinces or seascapes (Platt and Sathyendranath, 1999) relates to the patch hierarchy (refs in Kavanaugh et al., 2014), where the system is viewed as a partially ordered set, and system dynamics are a result of distinct but interacting patches within the system. The early approaches of biogeographic provinces (Sathyendranath et al.,1995; Longhurst, 1995, 2007, 2010) delimited static boundaries for the ocean regions. These regions were demarcated by areas grouped by their common oceanographic characteristics. These included surface chlorophyll fields from satellite, and a mix of climatologies and in situ data collection on physical (i.e. mixed layer depth, Brunt-Vaisala frequency, photic depth, Rossby internal radius of deformation), chemical (surface nutrient fields) and biological (production and respiration). These definitions of biogeochemical provinces have evolved to become dynamic in time at the monthly scale (Reygondeau et al., 2013), allowing insights on the effects of climate change (Reygondeau et al., 2020). The definition of dynamic seascapes is a prime example of the potential of the combination of multiple sources of EO data. The application of seascapes to Natural Capital quantification is interesting for computing dynamic changes in the extent of the marine ecosystem types, related to M2 Pelagic ocean waters (M2.1 Epipelagic ) biome. The presentation will introduce the ESA funded projects BOOMS (Biodiversity of the Open Ocean: Mapping, Monitoring and Modelling) and BiCOME (Biodiversity of the Coastal Ocean: Monitoring with Earth Observation). The projects investigate innovative ways to compute seascapes from satellite Earth Observation data.

Mangroves constitute a critical ecosystem that is under significant pressure despite providing a host of local to global ecosystem services. The mapping and monitoring of these ecosystems is of critical importance for their protection and sustainable management. To this end, the Global Mangrove Watch (GMW) was established in 2011 under the JAXA Kyoto & Carbon Initiative. The first map of global mangrove extent – 2010 baseline - was released at the Ramsar COP13 in 2018. The GMW v3.0 dataset was released in 2022, derived from time-series of JAXA L-band SAR (JERS-1 SAR, ALOS PALSAR, ALOS-2 PALSAR-2) from 11 annual epochs (1996, 2007, 2008, 2009, 2010, 2015, 2016, 2017, 2018, 2019 and 2020). The f1-score of the v3.0 mangrove extent was estimated to be 87.4% (95th conf. int. 86.2% - 88.3%). The GMW extent and change layers have been used for a number of applications including restoration potential and ecosystem service mapping. For assessing Blue Carbon services, Above Ground Biomass (by Simard et al. from 2000 SRTM) and Soil Organic Carbon (by Santerman et al, 2018) layers have been combined with the GMW v3.0 2020 extent. By combining these layers using the same mangrove extent allows them to be harmonised to accommodate assessment of total Blue Carbon within mangrove ecosystems. The GMW extent, change and Blue Carbon datasets are available in the public domain and can be accessed via the Global Mangrove Watch Platform at https://globalmangrovewatch.org.

Blue carbon ecosystems—seagrasses, mangroves, and tidal flats—provide globally significant yet vastly underestimated and impacted ecosystem services to humans, biodiversity, and economies like carbon sequestration, coastal protection, and biodiversity maintenance—the so-called natural climate solutions. Accelerating climate change, biodiversity loss, uneven levels of protection, and infancy in pertinent spatially explicit accounts and frameworks are all significantly stressing these physical and financial benefits of coastal ecosystems, necessitating cost- and time-effective contemporary technologies. Here, we present the novel coastal ecosystem accounting framework of the Global Seagrass Watch service. Our scalable ecosystem accounting framework blends modern Earth Observation advances—cloud computing, artificial intelligence, big satellite data analytics—with high-quality field datasets and other public geospatial datasets, across multi-national and multi-annual scales. We showcase the scalability, effectiveness, and confidence of our Earth Observation technological framework through its recent applications across both tropical and temperate coastal biomes. Leveraging our cloud-native framework within the Google Earth Engine cloud platform, we nationally aggregate high-resolution satellite mosaics using the open 10-m Sentinel-2 and 5-m NICFI PlanetScope image archives. These analysis-ready mosaics are then transformed into physical ecosystem accounts of seagrass extent, condition (e.g., bathymetry, water quality), and services (e.g., blue carbon stock and sequestration rate). Our showcased national spatial seagrass accounts to date cover more than 76,000 km2 of areal extent in 28 countries, encompassing around 74,000 km of coastline within three seagrass bioregions in the Western Indian Ocean, the Mediterranean, and the Caribbean. We discuss the real-world impact of our technology for climate change mitigation in Seychelles in a recent blueprint project. We also articulate current technological challenges and respective potential research and development solutions. We envisage that these resolutions will accelerate near-future opportunities and applications towards the full operationalization of Earth Observation for transparent and effective coastal ecosystem accounting, decision-making, financing, and resilience—within and beyond the 21st century.

Ecosystem extents comprise the foundation of ecosystem accounting and while terrestrial accounts have been enabled by robust methods for land cover and land use mapping, marine accounts rely on limited existing data sets that lack several desired characteristics including temporal resolution, thematic detail in high productivity regions, i.e., the coasts, and the spatial resolution to map these coastal regions. We explore the utilization of established methods for mapping coastal bathymetry with a data fusion of IceSAT-2 and optical sensors (Landsat and Sentinel-2). This semiautomated approach selects bathymetry points and trains machine learning models to create wall to wall maps of coastal bathymetry. We then delineate multiple coastal zone extents based on their depth. These zones represent the likelihood of important coastal ecosystems such as seagrass, reefs, and tidal flats which support fisheries and other ecosystem services. We demonstrate the approach in Mozambique, Senegal, and Liberia. We discuss the limitations and potential of our approach to improve marine ecosystem accounts.

Ocean Accounts are structured set of information, in the form of maps, data, statistics, and indicators, information obtained from Ocean Accounts plays an important role in ocean management as indicators of the marine resources sustainability. Given the complexity of its preparation, the implementation of Ocean Accounts in Marine Protected Areas (MPA) as pilot site is considered a solid start, and among all established MPAs, Gili Matra (Meno, Ayer, and Trawangan) has been selected as a pilot site for Ocean Accounts implementation in Indonesia. Four accounts have been prioritized to be developed, namely ecosystem extent/assets, flows to the economy, flows to the environment, and ocean governance. Assessment involved desk studies, field surveys, interviews, and image processing and analysis to present the results through map. Changes on ecosystem extent were identified by comparing the opening stock in 2015 with the closing stock in 2021 for coral reefs, seagrass, and mangroves ecosystem types. Key findings on ecosystem extents accounts: Coral reef ecosystem decreased from 259.50 ha to 247.50 ha, seagrass ecosystem expanded from 76.75 ha to 102.50 ha, and mangrove ecosystemdecreased from 21.50 ha to 10.69 ha. Fish abundance was increased by 7,704 ind/ha while fish biomass was decreased by 290.06 kg/ha. Assessment on flows to the economy was carried out in accordance with the role and functions of those ecosystems to identify the povisioning ecosystem services. Fishing and tourism gain primary economic benefit from ecosystem services found in Gili Matra MPA. Residual waste resulted from economic activity in the area was calculated to assess the flows to the environment accounts. In this assessment, both solid and liquid wastes were mainly sourced from tourism activities. Majority of solid waste 2,017 tonnes (70.4%) was disposed of in landfills, while 330,624 m3 liquid waste was managed by waste companies.

Mangroves provide a range of ecosystem services to humans and play a vital role in coastal zone protection and management. Globally, mangroves in the Niger Delta region of Nigeria are the least studied and amongst the worst degraded in the world due to continuous oil spillage, overharvesting of wood for domestic use, commercial dredging and wetland reclamation for farming and construction. Limitations to the in-depth quantitative study of the Niger Delta mangrove services and distribution result from the region’s volatility to conflicts and security concerns. While not directly accessible, the possibility of using accessible earth observation (EO) data for monitoring the spatial and temporal dynamics of the mangroves in this region is invaluable. Using the analysis-ready data (ARD) from Digital Earth Africa, we investigated the changing pattern of mangrove extent in the study area over four years from 2017 to 2021 and compared it with long-term observations from 1996 to 2021. We assessed yearly changes in mangrove cover between 2017 and 2021 using the ARD Sentinel 2 satellite imagery. We further investigated the spatial pattern of change over an extended period of twenty-five years using the Global Mangrove Watch (GMW) data, revealing significant spatial and temporal degradation. The effects of mangrove loss were identified by mapping flooded areas using the Water Observation from Space (WOfS), with some of the causes of mangrove loss linked to reduced mangrove health, urban growth, anthropogenic activities and the hydrological variations of the delta. The study recommends continuous environmental monitoring of the degradation status and policy changes promoting sustainable practices that produced some of the regeneration states quantified between 2017 and 2019. The approach of using ARD EO data to monitor ecosystems provides a framework for better-targeted conservation policies to reduce the impacts of climate change and sea level rise on the already vulnerable ecosystems.

To achieve SDG target number 14, Indonesia needs reliable measurements to understand its ocean condition. BPS Statistics Indonesia undertook an in-depth study titled SEEA - Ocean Accounts in-depth study in 2021 to gain a more thorough picture of the relationship between the ocean environment and economic activities and build the ocean account. SEEA - Ocean Accounts in-depth study took place in half of the area of the nation, with 850 respondents scattered over 17 provinces. Unit of statistic is establishment that engages in one of the nine marine economic activities, namely: 1. Fisheries; 2. Energy and Mineral Resources; 3. Biotechnology Industries; 4. Marine Industries; 5. Marine Services; 6. Marine Tourism; 7. Marine Transportations; 8. Marine Building; and 9. Defense, Security, Law Enforcement and Safety at Sea. One of the key findings is, marine economic activity that consumes the most natural resources is the one that deals with energy and mineral resources. The results of this in-depth study have been discussed in national and international forums including Indonesia Ocean Account Committee, PARIS21, Conservation International for Indonesia, and Ocean for Development Program forum with Statistics Norway, and other ocean accounting-related forums. The PSUT and waste account from the findings were also utilized by MMAF to complete Gili Matra Ocean Account where a number of policy applications were developed one of which is to provide feedback for the improvement of area zoning in the Gili Matra Marine Protected Area. Since March 2022, a continuation of the SEEA - Ocean Accounts in-depth study has been conducted in order to collect ocean account data from the remaining 17 provinces of Indonesia. The objective is to be able to generate national-level data for the Indonesian Ocean Account, specifically the flow to economy account and the flow to environment account.

Mangrove ecosystems are recognised as one of the nature-based solutions to a changing climate. They are most efficient in carbon sequestration among terrestrial ecosystems, storing 3-4 times more carbon per equivalent area in their above and below-ground biomass. Notwithstanding this benefit of mangrove ecosystems among several others, they are increasingly being destructed in some regions of the world. While efforts undertaken in the past years to develop global-scale mangrove extent maps have supported countries’ decision-making process on mangrove ecosystems, these maps are not up-to-date and are also not able to capture the actual extent of mangroves at a country or regional scale. For a more efficient decision-making process, studies need to be conducted to map out more recent mangrove extent at a country scale, assess the spatio-temporal changes in mangrove extent, and identify the specific areas where mangroves can be identified. The current study addresses these concerns for Ghana’s mangroves, identifying the available mangrove species, the factors leading to the spatio-temporal changes in their extent, as well as the past and current management strategies in protecting and restoring Ghana’s mangroves. We report that while Ghana’s mangrove extent has increased 55% from 1996 to 2021, the changes at the regional scale are different. The Volta and Western regions record an increase in their mangrove extent and the Central and Greater Accra regions record a decrease in mangrove extent. The study recommends that further studies estimate the carbon sequestration and coastal defence potential of Ghana’s mangroves toward climate change mitigation and adaptation. Moreover, further efforts must be undertaken by the relevant institutions in protecting and regulating the use of mangroves and broadly promoting sustainable mangrove restoration projects in Ghana.

Seagrass ecosystems offer a wide range of significant yet underestimated ecosystem services like global climate regulation and fish habitat maintenance. In order to estimate the extent, condition, and value of these services and strengthen their uptake into national policies, spatially explicit and high-confidence nationwide ecosystem accounts of seagrass extent are necessary. Within the DLR’s Global Seagrass Watch project, we leveraged modern Earth Observation advances—cloud computing, machine learning and big satellite data analytics—with big reference data to compile physical ecosystem accounts of Bahamian seagrasses. We integrated 18,881 cloud-filtered Sentinel-2 image tiles to create a four-year multi-temporal composite across more than 114,000 km2 of shallow water. We utilised a variety of pixel and object-based features, and 5500 reference data points in 20 different Random Forest machine learning architectures to map the seagrass ecosystem extent for Bahamian waters at 10 m resolution. The mapped Bahamian seagrass extent covers an area between 39,210 km2 and 46,792 km2, featuring F1-scores of 61.71% and 74.76%, respectively. Utilising region-specific seagrass carbon data and our spatially-explicit national seagrass extent, we estimate a carbon storage of up to 771 Million Mg and a sequestration rate which equals 68 times the amount of CO2 emitted by The Bahamas in 2018. Our generated ecosystem extent accounts and algorithms highlight the importance of Earth Observation applications for Ecosystem Accounting across the marine realm and underappreciated blue carbon ecosystems like seagrasses. This contemporary synergy could support more effective policy making and financial mechanisms to safeguard future coastal resilience for societies, biodiversity, and economies.

Seagrass meadows are one of the most productive ecosystems in the world and preserving and restoring them makes a significant contribution in alleviating CO2 emissions, yet at the same time these ecosystems have been in a global decline due to anthropogenic stressors. Current methods of monitoring the extent and condition of seagrass ecosystems face a trade-off between spatial coverage and spatial resolution. In-situ campaigns are essential as they provide accurate ground-truth data on the ecosystem condition, but they have a low spatial coverage and are labour and time intensive. On the other hand, aerial surveys or space-based remote sensing provide the required spatial coverage, but are limited in their ability to determine the ecosystem condition with sufficient accuracy. Combining the merits of both approaches is therefore called for. Over the last five years, planblue - a spin-out of the Max Planck Institute in Bremen, Germany - has developed the “DiveRay”, an innovative Underwater Hyperspectral Imaging system that is able to determine the amount of carbon contained in seagrass meadows (the carbon stock) as well as the condition (“health”) of the sea grass which is an indicator of its potential to sequester CO2. The DiveRay integrates underwater navigation, a Underwater Hyperspectral Imaging camera and high resolution RGB camera, and AI-driven, automated data processing. The DiveRay surveys the ocean floor with cm to mm resolution. When diver-operated it typically covers 1,000 m2 in a single dive. Planblue is currently working on an Autonomous Underwater Vehicle mounted version of the DiveRay which will extend this range. However, with seagrass ecosystems ranging from several square kilometres to over 10,000 km2 in size it is only when our “hotspot” surveys are combined with aerial surveys or space-based observations that entire ecosystem extent and condition can be monitored – another innovation we are currently working on.

Importance of plant functional traits (PFTs) for ecosystem service delivery potentials (ESDP) have been addressed in many empirical studies in recent years but mangrove ecosystem is yet to be investigated. Therefore, this study focused on the evaluation of the ESDPs from the Sundarbans mangrove forest of Bangladesh using the concept of plant functional trait characteristics of mangrove species. Species abundance data of the Sundarbans from 81 permanent sample points (PSPs), data of plant functional traits characteristics of mangrove species and ESDPs of the Sundarban mangrove species from existing literature and expert opinions for species wise ESDPs were used, and simple linear regression (SLR), contrast matrix analysis (CMA) and mixed effect model (MEM) with year and saline zone as fixed effects and PSP as random term were applied to carryout the analysis in RStudio. Overall maximum and minimum ESDPs have been increased insignificantly (p > 0.05) over time with respect to overall PFTs, but their spatial variation is significant (p < 0.001). CMA for three saline zones portrayed that changes of both maximum and minimum ES delivery potentials are highly significant for oligohaline vs polyhaline zones (p < 0.001) and significant for Mesohaline vs Polyhaline zones (p < 0.01) but for Oligohaline vs Mesohaline zones, changes of al ES delivery potentials are insignificant. In terms of salinity variation in the Sundarbans, both maximum and minimum ES have been decreased in mesohaline zone but increased in oligohaline zone followed by polyhaline zone. Similar trends have been observed for them with respect to overall functional diversity of PFTs. MEM depicted that PSP is insignificant for both maximum and minimum ESDPs whereas year is significant with Mesohaline zone but insignificant with Oligohaline and Polyhaline zones. This study could be a reference for decision-makers in formulating and adopting better management practices.

talk

Climate variability and ballooning populations are putting unprecedented pressure on agricultural croplands and their water use, which are vital for ensuring global food and water security in the twenty-first century. In addition, the COVID-19 pandemic, military conflicts, and changing diets have added to looming global food insecurity. Therefore, there is a critical need to produce consistent and accurate global cropland products at fine spatial resolution (e.g., farm-scale, 30m or better), which are generated consistently, accurately, and routinely (e.g., every year). In this regard, we produced the world’s first Landsat-derived global cropland extent product @ 30m (GCEP30) (Thenkabail et al., 2021; download @ LP DAAC). However, global food and water security requirements and recently announced natural capital accounting (NCA) initiatives require us to produce multiple cropland products at global scale at sufficiently high-resolution (30m or better). These global products include crop watering methods (irrigated or rainfed, cropping intensities, crop types, crop productivities (productivity per unit of land), and crop water productivities (productivities for unit of water). Over next 5 years, we propose to advance our work to produce some of these products, particularly global irrigated and rainfed and global major crop types. The study will make use of Landsat-8, 9, and Sentinel-2A&2B surface reflectance products already available in the Google Earth Engine (GEE) cloud, and NASA’s Harmonized Landsat Sentinel-2 (HLS) Landsat derived product (HLSL30) for 2013-present and Sentinel-2 derived product (HLSS30) for 2015-present, that together have sub-5-day global coverage at nominal 30m resolution. These GFSAD products will make significant contributions to the Earth System Data Records (ESDRs) including Climate Data Records (CDRs), and the White House initiatives such as the Natural Capital Accounting.

Earth Observation data streams are gradually becoming rich and reliable sources of spatial information for emerging environmental accounting frameworks. This contribution provides an overview of the potential applications of continental and very high-resolution crop maps to fulfil the data needs of the UN SEEA Ecosystem Accounting (SEEA-EA) framework. While EU crop type maps will be available for 2017-2021 as the Copernicus High-Resolution-Layers Vegetated Land Cover Component (HRL-VLCC), this study focuses on a 10-m precursor derived from satellite for 2018, EU-Crop-Map. The map is based on LUCAS in-situ data and Sentinel-1 (S1) Synthetic Aperture Radar observations. Machine Learning algorithms trained on time series of S1 from the main growing season captured the crop growing. As a result, every field cropped (e.g. wheat, maize, etc.; 19 types in total) have been mapped consistently and holistically at a very fine spatial scale. High-resolution crop maps can contribute in several ways to SEEA-EA implementations, covering all three main biophysical components of the accounting system: Ecosystem extent: the main cropping systems (arable crops, permanent crops, permanent grasslands) are key subtypes of agroecosystems. Consistent spatial information about the extent and distribution of these subtypes can be extracted from the EU-Crop-Map. Ecosystem condition: the EU-Crop-Map can also reveal further relevant characteristics of agroecosystems, for example spatial-and-temporal crop diversity, including the share of monocropping in any given area. The spatial layout of crop parcels can also be an important auxiliary information in the assessment of landscape structure. Ecosystem service supply and use: the different crops and their management interact with the associated communities and abiotic ecosystem processes. This strongly influences the services that an agroecosystem can deliver. Accordingly, the functional characteristics of the mapped crops (e.g. nectar and pollen yield, the length of bare soil periods) can be relevant input data for ES models.

Agricultural landscapes, apart from delivering provisioning ecosystem services (ES), have the potential to also deliver a broad set of other regulating and cultural services, including pollination, erosion control, climate regulation, and recreation. However, as the primary goal of the agricultural sector has been to produce agricultural products and raw materials, agricultural management has not particularly aimed at sustaining the production of non-provisioning ES. Integrating the value of ES and their complex relationships into national accounting systems can contribute to better, sustainable, and resilient ecosystem management. Since its publication, the SEEA Ecosystem Accounting (SEEA EA) framework has been implemented in multiple countries around the world with the UN Statistical Commission adopting it as a statistical standard in 2021. By using multiple data sources (incl. Earth Observations (EO), Land monitoring services, and ES initiatives), our study aimed at integrating various EO-based indicators in one consistent spatially explicit ES monitoring framework for Western and Northern European landscapes. Furthermore, given that the resilience of agriculture directly depends on the quality and function of ES, we attempted to assess the effects of potential agricultural management on ES utilizing Machine Learning (ML) techniques and demonstrate the potential implementation of ecosystem physical accounts into policy-making processes. Our ultimate goal was to showcase how integrating EO, ML, and Ecosystem Accounting can contribute towards the achievement of agri-environmental policies (such as the Common Agricultural Policy) as well as EU Green Deal goals, specifically, those related to Farm to Fork and Biodiversity Strategies.

Moderate resolution remote sensing data (30-meter resolution) provides annual information required for the development of ecosystem accounts, yet can be limited in resolving some land cover/use types due to the sub-hectare scale of change. The Sahel has extreme phenological changes from wet to dry seasons and sparse tree cover which complicate mapping the fine granularity of these resources. Over the past 20 years, Senegal’s population has doubled and rapid extensification of dryland agriculture has been reported. To address this problem, we processed the available archive of commercial Worldview-2, and -3 (2-meter resolution, 2008 to 2021) multi-spectral data for central drylands of Senegal with a U-Net Convolutional Neural Network to map changes in tree cover and croplands. We then validated these maps with an independent accuracy assessment, and compared outputs with similar temporal periods of Landsat derived land cover maps to understand the amount of information lost when mapping at a moderate scale. Our assessment demonstrates the challenges of processing temporally limited, yet spatially extensive commercial very-high resolution data and how they could be used in land cover mapping for ecosystems accounts.   

According to the updated guidelines of the System of Environmental-Economic Accounting Ecosystem Accounting (2021) ecosystem services are the contributions of ecosystems to the benefits that are used in economic and other human activities. In this study, we attempt to assess the contribution of crop provision ecosystem service to agricultural production with the use of production function method. We assume that crop provision can be described and eventually captured by the following services related to agro-ecosystems, that is pollination, soil retention, pest control and green water. An economic production function model is employed to estimate the contribution of these services to total agricultural production. Finally, the contribution of crop provision is estimated as the summation of the detached services. The model is empirically applied to a panel dataset of 27 EU countries for the time periods 2000, 2006, 2012, and 2018 drawn from Eurostat. The empirical model relies on a Cobb-Douglas functional specification and is estimated using a random effects estimator that accounts for the panel nature of the dataset. Estimation results indicate that services considered in the analysis have a positive and statistically significant effect on EU agricultural production. Agricultural production is found to be more responsive to changes in soil retention and pollination services. Authors share the biggest challenges in the application of the method in ecosystem accounting, i.e. conceptualize the crop provision service and the relation to other regulating services, data availability and reliability and the impact of a disaggregation method. The results provide new insights to possibilities of disentangling the ecosystem contribution in line with the ecosystem accounting framework

According to the updated guidelines of the System of Environmental-Economic Accounting Ecosystem Accounting (2021) ecosystem services are the contributions of ecosystems to the benefits that are used in economic and other human activities. In this study, we attempt to assess the contribution of crop provision ecosystem service to agricultural production with the use of production function method. We assume that crop provision can be described and eventually captured by the following services related to agro-ecosystems, that is pollination, soil retention, pest control and green water. An economic production function model is employed to estimate the contribution of these services to total agricultural production. Finally, the contribution of crop provision is estimated as the summation of the detached services. The model is empirically applied to a panel dataset of 27 EU countries for the time periods 2000, 2006, 2012, and 2018 drawn from Eurostat. The empirical model relies on a Cobb-Douglas functional specification and is estimated using a random effects estimator that accounts for the panel nature of the dataset. Estimation results indicate that services considered in the analysis have a positive and statistically significant effect on EU agricultural production. Agricultural production is found to be more responsive to changes in soil retention and pollination services. Authors share the biggest challenges in the application of the method in ecosystem accounting, i.e. conceptualize the crop provision service and the relation to other regulating services, data availability and reliability and the impact of a disaggregation method. The results provide new insights to possibilities of disentangling the ecosystem contribution in line with the ecosystem accounting framework.

talk

The ARIES team, in collaboration with the United Nations, has developed a web-based application, called ARIES for SEEA. SEEA Ecosystem Accounting has a strong emphasis on spatial modelling, which can be time-consuming, require substantial expertise and can be very challenging in data-limited locations. To overcome these limitations, the ARIES technology automates data and model integration processes, providing transparent assembly and reporting in a faster, cheaper way than past ecosystem service modelling. ARIES for SEEA enables more rapid and standard ecosystem account production in countries with limited resources or technical expertise, based on global and national data while simultaneously providing infrastructure for countries with more advanced data and modelling capacity to better reuse their data and models and share them with the rest of the world, advancing the practice of SEEA EA globally. Semantic modelling that underlies ARIES for SEEA enables this data and model integration, but the accounts-ready data are critical to obtaining reliable and comparable results. Since the launch of ARIES for SEEA, the team has collaborated with several countries, and the greatest challenge in nations where data are available has been the harmonisation of local datasets. The information in a country can only be used to produce statistical results when the analysis is based on consistent, analysis-ready, and fully documented inputs and methodologies. Despite the technical capacity of mapping agencies, this task has proven to be challenging in many contexts, since reconciling geospatial information within National Statistical Offices (NSOs) can be difficult. To improve the interoperability of this information nationally and globally, efforts to widen the production of accounts-ready data are fundamental. NSOs and other agencies engaged in producing accounts need support to understand and produce accounts-ready data, which can then be made interoperable for their benefit and, if appropriate, that of the larger SEEA community.

The vast amount of Earth Observation (EO) data, image processing technology and applications can create barriers to the use of EO data in ecosystem accounting. Moreover, the variety of tools and the lack of interoperability between these platforms pose a challenge and a risk of lock-in to the user. The Account Ready Data Hub (AccoRD-Hub) is an architecture for on-the-fly discovery, manipulation and validation of data (geospatial and temporal manipulation of raster, vector, and tabular datasets) in a statistical compliant way. It aims to remove as much as possible the hassle of expert knowledge to manipulate big data inputs, so that the user can focus on the compilation of ecosystem accounts. The AccoRD-Hub allows the harvest of data from different data sources and providers, but itself does not mandate to store any data. The AccoRD-HUB mainly provides processing functions and methods for data manipulation and compilation of harmonized data stacks following the FAIR (Findability, Accessibility, Interoperability, and Reusability) principles for ecosystem accounting. It relies for the data processing and storage on other platforms e.g. OpenEO. Next to EO pre-processing, the focus lays on statistics extraction for custom regions by: direct statistics extraction from various EO platforms and repositories; spatially disaggregation of statistics via EO & non-EO based proxies (statistically gridding) and then aggregation for custom regions; vector operations for different input sources; database operations (e.g. TSV format). Empathies are laid on data quality and tracking of uncertainties to support statistical reporting. Currently, the platform architecture is tested in different projects to generate multi-Tier level accounts in West-Africa and Europe. Since the harmonized output data stacks follow the STAC (spatio temporal asset catalog) specifications, established platforms like ARIES can directly access the account ready data.

Australia has invested in the development of experimental land and ocean accounts based on Earth Observation data and other related data sets. These observation-based data are providing rich information and open up new areas of policy relevant analysis. The accounts were developed using a human centred design approach in conjunction with the government department responsible for managing the implementation of Environmental-Economic Accounting. This talk will highlight some of the progress that Australia has made using earth observation data for the purposes of environmental-economic accounts. We will cover the data sources used in the compilation processes, the methods applied to construct accounts and share some of our observations to date. We will discuss the challenges of integrating large earth observation data sets that span over 30 years as well as potential applications from an end user perspective. One of the key findings from the operalisation of ocean ecosystems and environmental-economic accounts is how processes can be developed once and shared between land and ocean ecosystem accounts. The challenge for ecosystem and environment accounts is the degree of interoperability between data and systems to ensure good quality information. We will also discuss the potential application of these data from an end user perspective and the next steps for land accounts at the Australian Bureau of Statistics.

Statistics Canada has embarked on a new Census of Environment program, which will involve cataloguing ecosystems in Canada following an ecosystem accounting framework. Since ecosystems are the foundation of ecosystem accounting, there is a need to decide how ecosystems will be defined, delineated and monitored for change in extent and condition over time. In Canadian efforts to compile ecosystem accounts to date, land cover has been used as a proxy to represent ecosystems. While land cover is important, it represents only one facet of ecosystems and the SEEA EA recommends that ecosystems should be considered from a more holistic perspective that includes both biotic and abiotic components. Statistics Canada therefore proposes complementing land cover-based representations of ecosystems with two other approaches to mapping ecosystems: a multi-characteristic approach and a fully comprehensive approach. Without earth observation (EO) derived ecosystem and land cover data, a Census of Environment with ecosystem accounting at its core would be an almost impossible undertaking for a country as large and geographically diverse as Canada. Improvements and expansion in the use of EO have made significantly more environmental data available, at more frequent time intervals, with greater spatial coverage and resolution than ever before. However, it is important to recognize the important limitations of EO derived data as they are used in ecosystem accounting. This presentation will describe Statistics Canada’s proposed approach to mapping ecosystems, including a discussion of the limitations and strengths of each, with a particular focus on the role of EO data.

‘Living England’ (LE) aims to provide an up-to-date, national scale, predicted habitat map for England every two years. Led by Natural England and funded by Defra’s Environmental Land Mangement and Natural Capital and Ecosystem Assessment programmes, LE provides evidence on the extent and distribution of England’s broad habitats, supporting key policy areas including natural capital assessments and agri-environment schemes. LE brings together expertise in ecology, earth observation (EO) and data science to predict habitat classes for discrete, homogenous land segments, using a machine learning (ML) framework. Sentinel satellite imagery and derived products are combined with other open-source datasets (e.g., terrain, climate, geology) to generate segments, before implementing a series of random forest classification models. These models are specific to pre-delineated biogeographic zones (BGZs), to reduce processing loads and account for phenological variations. The models are trained, tested, and validated using ground-truth data compiled from dedicated field-based point surveys and pre-existing habitat datasets. Producing a national scale habitat map using an EO-ML approach within a public sector context presents certain technical and operational challenges. After a brief overview of LE, this talk will discuss the specific challenge of managing uncertainty in the modelled outputs. Whilst LE reports an average classification accuracy of c. 88% nationwide, significant variations exist across BGZs and habitat types. Furthermore, accuracy alone is insufficient for describing habitat mapping success in this context. To address this, alongside further acquisition of ground truth data in underrepresented areas, we are developing a bespoke, reliability scoring system for LE. This is based upon a set of normalised metrics, including F1 score, model confidence and validation data age/coverage, to determine a classification reliability score for every segment nationwide. On-going sensitivity testing and engagement with local experts is helping refine the approach, to facilitate adequate quantification and communication of mapping uncertainty for end-users.

Spatially dense global data, such as derived from remote sensing, provide a wealth of information about the world. However, to make full use of these data, new analytical approaches are necessary to statistically assess patterns and to make future predictions. Many of the questions that researchers and policymakers ask using global data involve specific hypotheses. For example, is primary production in natural habitats increasing more rapidly at northern latitudes due to more-rapid increases in temperature? Is fire frequency increasing more rapidly in rural areas with greater housing density? Are different range-management types more effective for carbon sequestration? These questions require statistical analyses that assess how changes in a variable of interest, Y, depend on one or more explanatory variables, X. In other words, the hypotheses can be conceptualized as regressions of Y on X. These are not simple regressions, however, because data often consist of time series of points taken at many locations over the globe, and for remote-sensing data the locations often number in the millions of pixels. We have developed methods that overcome the statistical and computational challenges of regression-type analyses on large spatiotemporal datasets. The suite of methods together make up PARTS (Partitioned Autoregressive Time-Series) analyses. By accounting for both spatial and temporal autocorrelation that are typically found in global data, PARTS not only gives correct hypothesis tests (correct P-values); PARTS also gives greater statistical power to test hypotheses that might otherwise be obscured by patterns in the data Y that aren't explained by variables X. The PARTS toolbox is flexible and available (https://github.com/morrowcj/remotePARTS), and can be used to analyze a wide range of different types of global assessment data.

Our joint resolve and capability to revert worldwide losses of natural capital (NC), from land degradation to biodiversity loss, will critically determine the vitality of our economy and more fundamentally the path our civilization will take. From now on, every land-use decision must be nature–positive and there is building pressure from both ground-up and top-down for economic actors to demonstrate their commitment to nature positive approaches. Today, land stewards are unable to focus on the long-term investment in their NC because markets incentivize extraction over conservation and restoration. Capital investors, on the other hand, are increasingly desperate to become nature positive, but miss standardized, verifiable, and investable pathways for doing so. We are building an end-to-end solution bridging the gap between capital providers and land stewards, who are implementing land-use decisions on the ground . A standardized infrastructure of NC accounts is set up, building conceptually on SEEA-EA and others, allowing land-stewards to assess, monitor and market their efforts to conserve and restore the NC of their land in order to shift economic incentives. At its core this requires an automated, quantitative, EO -driven measurement, reporting and verification (MRV) system, with the capability to assess every piece of land worldwide regarding its stock of NC, e.g. biodiversity, as well as its flows of ecosystem services. This MRV is being built in a joint effort of academic and industry partners, with collaboration and openness as core principles in order to maximize impact. Going beyond the MRV, we are building bridges into markets, by creating investment pathways, for example for insetting, supply chain risk reduction, philanthropic and compliance markets. This empowers land stewards and investors to drive land restoration and nature’s protection with real financial backing and long-term outlooks, while simultaneously helping guide investment portfolios and companies towards more nature positive outcomes.

Latest advances in cloud-based Earth Observation platforms have played a significant role in the upscaling and modernization of Ecosystem Accounting (EA) workflows, allowing physical and monetary accounting of vast and remote areas throughout the globe. However, in the context of computer science potential and machine learning capabilities, there are still many areas to be explored and integrated. One of these areas is the uncertainty of machine learning predictions and its relation to noise in Earth Observation and Ecosystem Accounting Here, we demonstrate a classification model that can retrain itself based on the spatially explicit uncertainty information that it gathers through the process. More specifically, under the probabilistic rules, the model can estimate the per pixel uncertainty of a classification product and retrain itself. The model retraining takes place by ingesting more training points from areas with minimum uncertainty. The study areas are divided into two regions as the remote sensing data are also from two different sources. The first case study, which is a four-class system coastal ecosystem extent classification, is based on Sentinel-2, level 2A data and it spans the whole Bahamian shallow shelf. The second case study consists of PlanetScope imagery and a binary(seagrass and non-seagrass ecosystem extent) classification and it takes place across the entire Seychelles national scale.    Our results indicate the adjustability capabilities of the model to the noise due to better accuracy assessment of all the classes and in particular the minority ones. Moreover, in addition to the better accuracy that our workflow achieved, the final product also provides the uncertainty extent of the mapped region. Such spatially explicit information can be vital for future planning of in-situ sampling, for reducing uncertainties in multi-tier physical and monetary accounting, and more importantly, for policy making regarding the extent of marine protected areas.

This study investigates the reasons for the decrease in the water level of Beyşehir Lake and the shrinkage in the lake's surface area in recent years. For this purpose, the lake water level was determined from multi-mission satellite altimeter data, and the lake area was calculated using high-resolution optical satellite images. Data from Copernicus Global Land Service was used for multi-mission satellite altimeter data, and the lake level trend between 1993-2022 was calculated with the least squares method. European Space Agency's (ESA) Sentinel-2 high-resolution optical images were used to determine the change in the lake surface area between 2015 and 2020. These high-resolution optical images were processed with The Sentinel Application Platform (SNAP) software. The Normalized Difference Water Index (NDWI) and Modified Normalized Difference Water Index (MNDWI) were calculated based on processed optical images, and these indexes reflect the changes in water surface area. From the satellite altimeter data, a decreasing trend of 2.5 ± 0.5 cm/yr in the lake water level in the last ten years and shrinkage of approximately 8 km2 in the last 6 years from the satellite images were determined. The possibility of one of the most important reasons being drought was emphasized, and monthly average air temperature data and monthly average precipitation data were obtained from the Turkish General Directorate of Meteorology. With these data, 3- and 12-month Standardized Precipitation Evapotranspiration Index (SPEI) were calculated. Regarding these calculated drought indexes, moderate, extreme, and severe hydrological drought has been determined in the region. According to the analysis, drought is thought to be the most important reason for the decrease in the lake water level and shrinkage in the lake surface area. Key words : Remote Sensing, Lake Water Level, Satellite Altimetry, Sentinel-2

Because of their long-lasting impacts on ecosystems with socio-economic over very large areas, droughts are considered by far the most damaging of natural disasters. Considering that the magnitude and duration of the drought impacts in Europe are continuously increasing, there is an urgent information need on impacts on ecosystems for decision making and mitigation for a number of economic and societal sectors. Here we present continental scale assessment of drought impact on above ground biomass (AGB), using several EO datasets, such as the Medium and the High Resolution Vegetation Phenology and Productivity (MR-VPP and HR-VPP) and the Corine Land Cover datasets of the Copernicus Land Monitoring Service, Soil Moisture Anomalies of the Copernicus Emergency Service as well as ecosystem distribution and climatic regions datasets. Seasonal trajectories and phenology and productivity parameters are used for studying vegetation dynamics as a response to soil moisture deficits. Due to the presently short time series (2017-2021), the 10m resolution Sentinel-2 derived HR-VPP data is complemented with the MR-VPP phenology and productivity anomalies on 500m resolution to address deviations from the long-term (2000-2021) normal conditions in Europe. Drought hazard and drought pressure is indicated by long-term observations of negative soil moisture anomalies. The study also demonstrates the potential of using big EO processing data cubes, such as the Euro Data Cube (https://eurodatacube.com/), and its potential of developing a big data analytics platform. The data cube allows the monitoring of drought induced AGB loss in agricultural parcels, grasslands or forest stands on the pan-European scale, instead of addressing broad ecosystems only. Data cubes provide a unique possibility to build a processing platform, which allows the fast, efficient, quality controlled, transparent and repeatable monitoring of drought impacts on annual and intra annual bases with the highest spatial resolution existing today.

One of the most transversal ecosystem thematic accounts is biodiversity accounting since it has implications in calculating the condition, ecosystem services, and assets. One of the main ways of calculating biodiversity is using maps of species distributions. However, the scales and periodicity with which these maps are generated are inadequate to meet the needs of accounting for biodiversity at national or regional scales. For a long time, the European Union has been at the forefront of monitoring species diversity in member countries and proposing conservation measures for both species and habitats through its Habitat and Birds Directives. Within these directives, the member countries send official information on the status of the species and habitats of community interest found in these directives. This information, updated every six years, is in a resolution of 10x10 km, which is very useful on a European scale but insufficient on a national or regional scale. In the present study, we have developed a methodology that increases the resolution of the information, reaching maps of species distributions at 1x1 km. In addition, thanks to advances in remote sensing and machine learning techniques, we have been able to model the spatial distribution of species of community interest present in Spain. In the years, we have information from the directive reports and in other years. Thanks to the models' predictions from the remote sensing explanatory variables used in the modelling. This method provides robust, time-series species distributions that can be used to create repeatable, comparable, and scalable biodiversity accounts. Finally, you can be useful in developing the new monitoring of species within the Habitat and Birds Directives and in developing other accounts such as condition or ecosystem service.

The Global Environment Facility (GEF) is an international financial mechanism providing grants for generating global environment benefits. Multilateral development projects funded by the GEF provide technical support on mainstreaming environmental policies and principles into other sectoral decisions. Multilateral projects are usually set up to test or develop new technologies that improve sustainability. Earth observation technologies are important tool sets to assist countries in monitoring and assessing the status of, and changes in the natural environment, and these international projects use earth observation technologies in project formulations. In the last five years, such internationally funded projects including earth observation technologies provide greater opportunities to develop and use ecosystem service accounts in decisions. This presentation demonstrates how earth observation technologies are utilized in international projects and how the data generated by earth observation have been or can be converted to ecosystem service accounts. The last part of the presentation elaborates how to upscale local scale ecosystem service accounts to a national programme through internationally funded, more specifically GEF funded, projects.