Machine learning for ecosystem services

Simon Willcock; Javier Martinez-Lopez; Danny A.P. Hooftman; Kenneth J. Bagstad; Stefano Balbi; Alessia Marzo; Carlo Prato; Saverio Sciandrello; Giovanni Signorello; Brian Voigt; Ferdinando Villa; James M. Bullock; Ioannis Athanasiadis

doi:10.1016/j.ecoser.2018.04.004

USGS · 70199950

Machine learning for ecosystem services

Simon Willcock·Javier Martinez-Lopez·Danny A.P. Hooftman·Kenneth J. Bagstad·Stefano Balbi·Alessia Marzo·Carlo Prato·Saverio Sciandrello·Giovanni Signorello·Brian Voigt·Ferdinando Villa·James M. Bullock·Ioannis Athanasiadis

View original source ↗Export metadata ↓

Abstract

Recent developments in machine learning have expanded data-driven modelling (DDM) capabilities, allowing artificial intelligence to infer the behaviour of a system by computing and exploiting correlations between observed variables within it. Machine learning algorithms may enable the use of increasingly available ‘big data’ and assist applying ecosystem service models across scales, analysing and predicting the flows of these services to disaggregated beneficiaries. We use the Weka and ARIES software to produce two examples of DDM: firewood use in South Africa and biodiversity value in Sicily, respectively. Our South African example demonstrates that DDM (64–91% accuracy) can identify the areas where firewood use is within the top quartile with comparable accuracy as conventional modelling techniques (54–77% accuracy). The Sicilian example highlights how DDM can be made more accessible to decision makers, who show both capacity and willingness to engage with uncertainty information. Uncertainty estimates, produced as part of the DDM process, allow decision makers to determine what level of uncertainty is acceptable to them and to use their own expertise for potentially contentious decisions. We conclude that DDM has a clear role to play when modelling ecosystem services, helping produce interdisciplinary models and holistic solutions to complex socio-ecological issues.

Explore related subjects

Ecosystem Services Article

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Simon Willcock, Javier Martinez-Lopez, Danny A.P. Hooftman, Kenneth J. Bagstad, Stefano Balbi, Alessia Marzo, Carlo Prato, Saverio Sciandrello, Giovanni Signorello, Brian Voigt, Ferdinando Villa, James M. Bullock, Ioannis Athanasiadis. 2018. Machine learning for ecosystem services. https://doi.org/10.1016/j.ecoser.2018.04.004

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related USGS reports

USGS2026

The influence of tree canopy cover data choices on urban ecosystem accounting

Because urban landscapes are heterogeneous, the methods and spatial resolution used to depict the land surface greatly influence the representation of urban features. Land cover products such as tree canopy cover (TCC) are particularly sensitive to the methodology and resolution used in their creation. Differences in TCC mapping have implications on the outcomes of ecosystem service (ES) models, including those underlying natural capital accounting. Here, we quantify the sensitivity of physical rainfall interception and local climate regulation ES models for 189 U.S. cities to TCC inputs from four TCC products: a) National Land Cover Database (NLCD), b) Enhanced NLCD TCC, c) aggregated city-specific composite, and d) global tree canopy height dataset. We find both city-level and aggregate differences in TCC estimates, from a 38% decrease to a 3% increase relative to an aggregated high-resolution product. These differences result in up to 3% overestimations and 27% underestimation of rainfall interception and 2–56% underestimation of local climate regulation ES. City size, population, and greenness in addition to climatic variation drove differences between TCC products, and this variation requires users to carefully consider the choice of input data for any planned analysis. Though high-resolution data can offer greater nuance and accuracy, more limited spatiotemporal availability can hinder their usefulness for long-term monitoring applications such as natural capital accounting. The differences found in this study provide valuable insights for making informed decisions on data inputs for use in urban ecosystem research and for contextualizing model outcomes.

Contiguous United States↗

USGS2025

Interoperability for ecosystem service assessments: Why, how, who, and for whom?

Despite continued, rapid growth in the literature, the fragmentation of information is a major barrier to more timely and credible ecosystem services (ES) assessments. A major reason for this fragmentation is the currently limited state of interoperability of ES data, models, and software. The FAIR Principles, a recent reformulation of long-standing open science goals, highlight the importance of making scientific knowledge Findable, Accessible, Interoperable, and Reusable . Critically, FAIR aims to make science more transparent and transferable by both people and computers . However, it is easier to make data and models findable and accessible through data and code repositories than to achieve interoperability and reusability. Achieving interoperability will require more consistent adherence to current technical best practices and, more critically, to build consensus about and consistently use semantics that can represent ES-relevant phenomena. Building on recent examples from major international initiatives for ES (IPBES, SEEA, GEO BON), we illustrate strategies to address interoperability, discuss their importance, and describe potential gains for individual researchers and practitioners and the field of ES. Although interoperability comes with many challenges, including greater scientific coordination than today’s status quo, it is technically achievable and offers potentially transformative advantages to ES assessments needed to mainstream their use by decision makers. Individuals and organizations active in ES research and practice can play critical roles in creating widespread interoperability and reusability of ES science. A representative community of practice targeting interoperability for ES would help advance these goals.

Ecosystem Services↗

USGS2025

The Benefits Knowledges Learning Framework: A tool for learning across diverse knowledge systems in ecosystem valuation

Sustainable and just environmental management depends on meaningful consideration of the plural values of nature, as they arise in association with diverse worldviews and understandings of well-being. To achieve value pluralism in decision-making, we must also attend to knowledge pluralism, in terms of recognizing the validity and decision relevance of a broader suite of knowledge forms that convey diverse understandings of well-being and benefit. In this article, we outline a social learning tool – the Benefits Knowledges Learning Framework – that supports expanded thinking about decision-relevant, actionable knowledge, and the associated spectrum of available opportunities to learn from these diverse knowledge forms across phases of decision-making. It does so through: 1) cultivation of reflexivity and mutual learning about the knowledge systems of diverse actors involved in the decision process; 2) identification of diverse benefits knowledge forms that are available to inform decision-making; and 3) identification of opportunities to learn from these knowledge forms. Diverse forms of benefits knowledge include both knowledge products (documentation) and knowledge practices (lived and embodied). The framework can be applied to retrospective case analysis to understand and learn from constraints and enabling factors in past decision processes. It can also be applied to assess on-going decision-making and identify current opportunities for improvement. The framework begins with a start-up phase that encourages those applying the framework to address any concerns raised by stakeholders and rightsholders and determine whether framework application is appropriate in a particular context.

Washington↗