Semantic Workflows and Machine Learning for the Assessment of Carbon Storage by Urban Trees Juan Carrillo Daniel Garijo Mark Crowley University of Waterloo Information Sciences Institute University of Waterloo Waterloo, Canada University of Southern California Waterloo, Canada jmcarril@uwaterloo.ca USA mark.crowley@uwaterloo.ca dgarijo@isi.edu Rober Carrillo Yolanda Gil Katherine Borda United World Colleges Information Sciences Institute Waterloo, Canada Bogota, Colombia University of Southern California kbordac@gmail.com rober.carrillo@uwcim.net USA gil@isi.edu ABSTRACT and have a clear separation between workflow design and work- Climate science is critical for understanding both the causes and flow execution [4] [43]. One of the major advantages of scientific consequences of changes in global temperatures and has become workflow systems is their role in improving the reproducibility of imperative for decisive policy-making. However, climate science scientific studies. Reproducibility plays a critical role in climate studies commonly require addressing complex interoperability is- sciences due to their impact in our society [38]. In fact, due to is- sues between data, software, and experimental approaches from sues with the documentation of experiments, some climate science multiple fields. Scientific workflow systems provide unparalleled studies have been re-examined lately due to their impact in global advantages to address these issues, including reproducibility of policy-making [19] and water resources management [25]. experiments, provenance capture, software reusability and knowl- In this paper we describe the process we followed to design and edge sharing. In this paper, we introduce a novel workflow with implement reusable scientific workflows for the climate sciences. a series of connected components to perform spatial data prepa- In particular, we focus on carbon storage assessment by using ur- ration, classification of satellite imagery with machine learning ban trees, a common requirement for cities to reduce their carbon algorithms, and assessment of carbon stored by urban trees. To the emissions globally. Our contributions include the development of best of our knowledge, this is the first study that estimates carbon a library of components for preparing geospatial data by doing storage for a region in Africa following the guidelines from the coordinate transformations, the integration of machine learning Intergovernmental Panel on Climate Change (IPCC). components to classify trees in satellite images and the creation of workflows for carbon storage assessment in cities. In order to imple- KEYWORDS ment these workflows, we use the Workflow Instance Generation and Selection (WINGS) system [18], which has been successfully Reproducibility, scientific workflows, machine learning, land cover used for applications in domains ranging from Genomics [17] to mapping, carbon assessment, Sentinel-2 Geosciences [16]. The paper starts by giving an overview of previous research on 1 INTRODUCTION the assessment of carbon storage by urban trees according to the Climate science requires modeling natural and man-made processes guidelines published by the Intergovernmental Panel on Climate that are highly complex, exhibit non-linear dynamics and possess Change (IPCC), highlighting the advantages and limitations of the disparate spatial and temporal scales. Handling this complexity most recent methods. We then describe the design considerations requires a holistic approach among multiple disciplines [29], but of our scientific workflows as well as their experimental implemen- scientists from different fields may also need to use domain-specific tation and evaluation. The paper continues with a discussion of data sources, methods, and computational models. The integration results and suggested future work. of their knowledge and experiments is a challenging task [7], espe- cially when a study is expected to provide actionable insights for decision making at regional and local scale. 2 BACKGROUND: CARBON EMISSIONS AND Scientific workflows have emerged as an integrated solution to manage this challenge, as they capture the computational steps STORAGE and data dependencies required to carry out a computational ex- There is an increasing interest among scientific organizations and periment [40]. Scientific workflows ease data handling (metadata, national governments regarding Carbon emissions and their role provenance), component versioning (parametrization, calibration) in climate science [13]. The consequences of higher concentrations of carbon gases in the atmosphere are now more explicit and in- Copyright ©2019 for this paper by its authors. Use permitted under Creative Commons ternational organizations such as the Intergovernmental Panel on License Attribution 4.0 International (CC BY 4.0). Climate Change (IPCC) are leading initiatives to monitor national Sciknow ’19, November 19, 2019, Los Angeles, California, USA Carrillo, et al. efforts to lower emissions and increase carbon storage [10]. Mon- developing countries where technical capacity and resources are itoring carbon emissions is fundamental to inform government particularly limited [1]. policies in topics such as renewable energy, transportation, and manufacturing technologies. Similarly, the assessment of carbon 3 SCIENTIFIC WORKFLOW DESIGN storage is equally important to guide hazard mitigation efforts [20]. With the advantages and limitations of published methods in mind, Most studies in climate science require domain knowledge to we design a new workflow to efficiently determine tree cover for design and run the experiments [23]. But the increasing concerns urban areas We start by presenting the advantages of knowledge about the changing climate require strategies to streamline the capture systems to represent models in geosciences as scientific replication of assessment studies, the reusability of data, methods, workflows and then describe how we leverage previous research on and results. The use of scientific workflows can significantly im- carbon assessment and tree mapping to design our own workflow. prove the implementation and reproducibility of carbon assessment The workflow is designed as multiple interconnected compo- studies, with additional gains in data and model sharing as well as nents in WINGS that operate in three consecutive stages as seen knowledge capture through semantic representations. in Figure 1, data preprocessing, mapping of tree coverage, and as- sessment of carbon storage. Our workflow is based on previous work in which high resolution satellite imagery are used to produce 2.1 Carbon assessment by urban trees land cover maps over urban areas [35] [37]. However, we focus on Urban trees provide a natural and cost-effective alternative to cap- using freely available medium resolution satellite imagery from the ture and store carbon in cities. Having trees in densely populated Sentinel-2 sensors [9] to facilitate replication by other researchers. areas also improve human health and biodiversity and provide ben- efits for flood prevention and reduced cooling costs, among other benefits [26]. In 2003, the IPCC published the Good Practice Guid- ance for Land Use, Land-Use Change and Forestry [33] and in 2006 the IPCC Guidelines for National Greenhouse Gas Inventories [11]. These guidelines suggest the use of area covered by trees, shrubs, and herbaceous (perennial) plants to determine the amount of car- bon stored as biomass in settlements. However, due to the limited Figure 1: Stages of our carbon assessment workflow availability of detailed data the majority of published studies focus only on tree cover. While these two documents describe the stages Spatial data preprocessing stage involves common operations of an assessment study, including aspects such as data collection for experiments across earth sciences, such as transformation of and uncertainty estimation, they suggest governments to deal with coordinate systems and conversion between file formats. We have minor implementation details according to their technical capacity designed all these data preprocessing steps as reusable building and available resources. blocks so they can be included in other workflows. These prepro- Published carbon assessment studies use different combinations cessing operations are also known as Extract Transform Load (ETL) of data, methods, and software. These studies can be divided into tasks and are implemented using the Geospatial Data Abstraction two major groups according to the data collection approaches and Library GDAL [15]. models they use. Assessments in the first group use a statistics point We map the tree coverage using satellite image classification and sampling technique to estimate tree density from aerial imagery design multiple components to train machine learning algorithms, [28][32][31]. This method is easy to implement and only requires classify the image over an area of interest, produce a visualization imagery for sample areas, but the outcome is a percentage value ready tree cover map, and determine the resulting accuracy. The that does not describe the spatial distribution of trees. The second Machine Learning algorithms we implement are Random Forests group of methods use LiDAR, aerial or satellite imagery to pro- and Support Vector Machines, both available in the Orfeo Remote vide a comprehensive assessment of tree coverage, including their Sensing toolbox [21]. Random Forest [36] [5] and Support Vector spatial distribution [42][37][8][35]. However, the second approach Machines [41] [6] are well-known methods for pixel-based image requires imagery for the complete the area of study as well as the classification in Remote Sensing due to their straightforward im- configuration of more complex methods for detection or classifi- plementation and calibration, as well as its documented robustness cation of urban trees. Later in this document we introduce our and accuracy. own method, which belongs to the second group and uses freely We train the algorithms using sample points collected through available satellite imagery and a carefully designed workflow to visual inspection as described further in this document. Our work- facilitate implementation and reuse. flow generates a map that includes other land cover categories One common limitation of carbon assessment studies is the lack such as water, grass and built areas; which may additionally serve of a systematic approach to share data, models, software, and results. for other use cases in disciplines such as hydrology, planning, and Regardless of the specific data source, technology, or processing forestry, just to name a few (as seen in Figure 2). Moreover, this method, most reports only contain descriptions of the work done, map is generated in a standard format for further use in Geographic which are not enough to replicate the experiments or reuse the Information Systems or other scientific platforms. software tools [34]. The assessment of carbon storage is completed following the Turning the information from those reports into actionable IPCC guidelines to calculate carbon stored based on urban canopy knowledge becomes a cumbersome task, especially for scientists in area. In the calculation we multiply the canopy area by a conversion Semantic Workflows and Machine Learning for the Assessment of Carbon Storage by Urban Trees Sciknow ’19, November 19, 2019, Los Angeles, California, USA and especially in South Sudan [3][24], which hinder the work of humanitarian and non government organizations; therefore, any study aiming to understand the characteristics of the rural or urban territories will likely produce positive outcomes in the short and long term. The data we use includes 4000 sample points digitized through visual inspection using Google Earth high resolution satellite im- agery. The digitization is performed by trained individuals and even though this task is not implemented as one of the workflow components we follow best practices in Geographic Information Figure 2: Some applications of land cover mapping Systems [14] and complete random checks to ensure data accuracy and consistency. The specific steps are as follows: First, 4000 ran- factor to estimate carbon stored in the form of biomass. Since no dom points are generated within the city boundary and used as values are published specifically for Africa (our initial region of spatial reference to ensure that the actual sample points are spa- interest) we use a default value suggested by the IPCC. tially distributed across the city. Then, the person who is digitizing places sample points closer to the random points but over locations where examples of each land category are seen. The four land cat- egories are trees, grass, impervious, and water, with 1000 points per category to have a balanced dataset. Impervious includes all built or bare areas not covered by vegetation, water, or agriculture. Examples of impervious areas are constructions, parking lots, roads, airport runways, and rocky surfaces. Additionally, we use a multi-spectral satellite image from the Sentinel-2 sensor made available by the European Space Agency (ESA) [9]. While this image originally includes 13 bands we use only the four bands with a pixel size equivalent to 10 meters on the ground. The first three bands collect data in the visible light spectrum and the fourth in the near infrared spectrum, with the latter being particularly useful for detecting vegetation. Figure 4 shows some sample points of the four land cover cate- gories in a neighborhood close to the White Nile river in the city of Juba, overlaid on high resolution satellite imagery from Google Earth and the same area as seen on the medium resolution Sentinel- 2 image. 5 IMPLEMENTATION AND RESULTS The workflow is implemented as a series of software components in WINGS. Additionally, each component is designed to run a par- ticular task and configured with relevant parameters to calibrate their functionality. Components are coded as Python scripts that serve as a high level interface for us to use libraries such as GDAL Figure 3: City of Juba in South Sudan, Africa for data preprocessing and Orfeo command line tools for satellite image classification. These components are connected through intermediate datasets 4 AREA OF STUDY AND DATA that in turn are outputs and inputs for the previous and follow- Our area of study is the city of Juba in South Sudan. It is the cur- ing components. We created 14 components in total, with eight rent capital of the country and serves as its main commercial and dedicated for data preparation, five for mapping of tree coverage, transportation hub with an estimated population of nearly 386,000 and one for carbon assessment. The use of components as modu- inhabitants [2]. Juba is located in the southern region of the country lar pieces of software to accomplish specific data processing tasks and has an extension of 103 km2 according to the urban boundary creates opportunities for reusability across a variety of models in retrieved in July 2019 from Open Street Maps [22]. Figure 3 shows Earth Sciences and Geospatial technologies. the location of Juba and South Sudan in the African continent. The eight components designed for data preparation allow re- The country of South Sudan currently faces multiple issues, searchers to handle datasets in the most common file formats and including political instability [39], poor health services [27], and transform them according to the particular goals of the study. Some a lack of infrastructure, especially for storage and distribution of operations correspond to sub-setting the spatial extent of the data, water [30]. Few mapping projects have been conducted in Africa, changing its coordinate reference system, and editing the attributes Sciknow ’19, November 19, 2019, Los Angeles, California, USA Carrillo, et al. Figure 5: Workflow fragment for data preparation of sample points algorithms. Then we use the trained algorithms to perform pixel level classification of the Sentinel-2 image for the entire area of interest and output the tree cover map. Subsequently, one more component evaluates the classification accuracy. Figure 6 shows a fragment of our workflow where we use multiple components to Figure 4: Sample points and satellite imagery of Juba map tree cover. Table 1 shows the normalized confusion matrix resulting from the evaluation of the Random Forest classification algorithm. We for tabular datasets. When running the workflow for a new area see that the tree cover category is the one with the lowest accuracy of study, the task of setting parameters such as the coordinate ref- with only 54% of the trees classified correctly in the test set. This erence system is facilitated by the WINGS system, which suggest is likely the result of using medium resolution satellite imagery the most appropriate value according to the geographic region as (Sentinel-2) with a ground pixel size of 10 m, in other words, the configured by the workflow designer. Figure 5 shows a fragment of canopy area of a tree should be about 100 m2 to be easily identifiable our workflow where we use multiple data preparation components in at least one pixel, without considering boundary issues between to perform a format transformation and reprojecting a file. adjacent pixels. The grass and impervious land cover categories Mapping of tree coverage includes four components focused on exhibit a comparable accuracy of 65% and 73% consequently. While training the Machine Learning algorithms to classify the Sentinel-2 areas corresponding to these two categories show a slightly better satellite image using the sample points digitized through visual accuracy they are still hard to differentiate, presumably due to grass inspection. Initially, one component extracts the pixel values of patches and house rooftops with a size smaller than the area of a the satellite image for the 4000 sample locations. Next, another pixel (100 m2 ). For the water land cover category the algorithm component uses 80% of these values as training and 20% as valida- reaches an almost perfect accuracy, which is anticipated due to the tion data to train the Random Forest and Support Vector Machine significant difference in the way it reflects the light compared to Semantic Workflows and Machine Learning for the Assessment of Carbon Storage by Urban Trees Sciknow ’19, November 19, 2019, Los Angeles, California, USA Figure 7: Resulting land cover map for the city of Juba cover)-1 yr -1 , which is the value suggested by the IPCC for Tier 2a studies [11]. Multiplying the default carbon removal factor by the total tree cover area (10,519 ha) from the previous stage, we calculate that trees in the city of Juba remove 30,506 tonnes C yr -1 . This amount is equivalent to the carbon dioxide emitted by 6632 passenger vehicles per year [12]. This value along with the classifi- cation accuracy from the previous stage are valuable information for countries to prepare their carbon assessment reports accord- ing to the IPCC guidelines. The code used to create the geospatial transformation and the carbon assessment workflows is available online.12 6 CONCLUSIONS AND FUTURE WORK Figure 6: Workflow fragment for mapping tree cover In this paper we introduced our work to create a library of workflow components to perform spatial data transformations, land cover Table 1: Normalized confusion matrix mapping and assessment of carbon storage. By leveraging scientific workflows, we aim to ease the reusability of these components Predicted class in other workflows and the reproducibility and transparency of Water carbon assessment studies. Our future work will focus on two main Grass Trees Imp. areas: first, we aim to test our workflow using data from other Trees 0.54 0.18 0.28 0 locations around the globe, which requires additional training data Actual class Grass 0.32 0.65 0.14 0 points. Second, we will focus on calibration of the parameters for Impervious 0.17 0.09 0.73 0 the classifiers to improve our classification accuracy during the Water 0.01 0 0 0.99 land cover mapping stage. REFERENCES the other three land cover categories, especially in the near infrared [1] Udara Willhelm Abeydeera, Lebunu Hewage, Jayantha Wadu Mesthrige, and band (B4). Tharushi Imalka Samarasinghalage. 2019. Global Research on Carbon Emissions: A Scientometric Review. Sustainability 11, 14 (2019), 3972. The assessment of carbon storage is performed by one compo- [2] Central Intelligence Agency. 2018. The World Factbook: South Sudan. https: nent that receives the tree cover map shown in Figure 7, calculates //www.cia.gov/library/publications/the-world-factbook/geos/od.html the total canopy area and uses a carbon removal factor to determine the total amount of carbon stored as biomass in the city of Juba. 1 https://bitbucket.org/jmcarrillog/wings-geospatial-etl/ We use a default carbon removal factor of 2.9 tonnes C (ha crown 2 https://bitbucket.org/jmcarrillog/carbon-urban-trees/ Sciknow ’19, November 19, 2019, Los Angeles, California, USA Carrillo, et al. [3] Sabina N Baariu, Galcano C Mulaku, David N Siriba, et al. 2019. State of Carto- Science & Policy 79 (2018), 1–8. graphic Services among the East African Community Member States. Journal of [26] SJ Livesley, EG McPherson, and C Calfapietra. 2016. The urban forest and Geographic Information System 11, 01 (2019), 56. ecosystem services: impacts on urban water, heat, and pollution cycles at the [4] Derik Barseghian, Ilkay Altintas, Matthew B. Jones, Daniel Crawl, Nathan Potter, tree, street, and city scale. Journal of environmental quality 45, 1 (2016), 119–124. James Gallagher, Peter Cornillon, Mark Schildhauer, Elizabeth T. Borer, Eric W. [27] Peter M Macharia, Paul O Ouma, Ezekiel G Gogo, Robert W Snow, and Abdis- Seabloom, and Parviez R. Hosseini. 2010. Workflows and extensions to the alan M Noor. 2017. Spatial accessibility to basic public health services in South Kepler scientific workflow system to support environmental sensor data access Sudan. Geospatial health 12, 1 (2017), 510. and analysis. Ecological Informatics 5, 1 (2010), 42 – 50. https://doi.org/10. [28] Mark McGovern and Jon Pasher. 2016. Canadian urban tree canopy cover and 1016/j.ecoinf.2009.08.008 Special Issue: Advances in environmental information carbon sequestration status and change 1990–2012. Urban forestry & urban management. greening 20 (2016), 227–232. [5] Mariana Belgiu and Lucian Drăguţ. 2016. Random forest in remote sensing: A [29] Alison M Meadow, Daniel B Ferguson, Zack Guido, Alexandra Horangic, Gigi review of applications and future directions. ISPRS Journal of Photogrammetry Owen, and Tamara Wall. 2015. Moving toward the deliberate coproduction of and Remote Sensing 114 (2016), 24–31. climate science knowledge. Weather, Climate, and Society 7, 2 (2015), 179–191. [6] Gabriele Cavallaro, Morris Riedel, Matthias Richerzhagen, Jón Atli Benediktsson, [30] Ayman Nassif, Ann Stewart, Millie Mutepfe, and Petros Christou. 2016. Infras- and Antonio Plaza. 2015. On understanding big data impacts in remotely sensed tructure needs in sub Saharan Africa with particular reference to South Sudan. image classification using support vector machine methods. IEEE journal of In Third Australasia and South-East Asia Structural Engineering and Construction selected topics in applied earth observations and remote sensing 8, 10 (2015), 4634– Conference: ASEA-SEC-3. ISEC Press. 4646. [31] David J Nowak, Eric J Greenfield, Robert E Hoehn, and Elizabeth Lapoint. 2013. [7] Joseph David Daron, Kate Sutherland, Christopher Jack, and Bruce C. Hewitson. Carbon storage and sequestration by trees in urban and community areas of the 2015. The role of regional climate projections in managing complex socio- United States. Environmental pollution 178 (2013), 229–236. ecological systems. Regional Environmental Change 15, 1 (01 Jan 2015), 1–12. [32] Jon Pasher, Mark McGovern, Michael Khoury, and Jason Duffe. 2014. Assessing https://doi.org/10.1007/s10113-014-0631-y carbon storage and sequestration by Canada’s urban forests using high resolution [8] Zoe G Davies, Jill L Edmondson, Andreas Heinemeyer, Jonathan R Leake, and earth observation data. Urban forestry & urban greening 13, 3 (2014), 484–494. Kevin J Gaston. 2011. Mapping an urban ecosystem service: quantifying above- [33] Jim Penman, Michael Gytarsky, Taka Hiraishi, Thelma Krug, Dina Kruger, Riitta ground carbon storage at a city-wide scale. Journal of applied ecology 48, 5 (2011), Pipatti, Leandro Buendia, Kyoko Miwa, Todd Ngara, Kiyoto Tanabe, et al. 2003. 1125–1134. Good practice guidance for land use, land-use change and forestry. Good practice [9] Matthias Drusch, Umberto Del Bello, Sébastien Carlier, Olivier Colin, Veron- guidance for land use, land-use change and forestry. (2003). ica Fernandez, Ferran Gascon, Bianca Hoersch, Claudia Isola, Paolo Laberinti, [34] Gillian Petrokofsky, Hideki Kanamaru, Frédéric Achard, Scott J Goetz, Hans Philippe Martimort, et al. 2012. Sentinel-2: ESA’s optical high-resolution mission Joosten, Peter Holmgren, Aleksi Lehtonen, Mary CS Menton, Andrew S Pullin, for GMES operational services. Remote sensing of Environment 120 (2012), 25–36. and Martin Wattenbach. 2012. Comparison of methods for measuring and as- [10] Ottmar Edenhofer. 2015. Climate change 2014: mitigation of climate change. Vol. 3. sessing carbon stocks and carbon stock changes in terrestrial carbon pools. How Cambridge University Press. do the accuracy and precision of current methods compare? A systematic review [11] Simon Eggleston, Leandro Buendia, Kyoko Miwa, Todd Ngara, and Kiyoto Tanabe. protocol. Environmental Evidence 1, 1 (2012), 6. 2006. 2006 IPCC guidelines for national greenhouse gas inventories. Vol. 5. Institute [35] Steve M Raciti, Lucy R Hutyra, and Jared D Newell. 2014. Mapping carbon storage for Global Environmental Strategies Hayama, Japan. in urban trees with multi-source remote sensing data: Relationships between [12] United States Environmental Protection Agency EPA. 2019. Greenhouse Gas biomass, land use, and demographics in Boston neighborhoods. Science of the Emissions from a Typical Passenger Vehicle. https://www.epa.gov/greenvehicles/ Total Environment 500 (2014), 72–83. greenhouse-gas-emissions-typical-passenger-vehicle [36] Victor Francisco Rodriguez-Galiano, Bardan Ghimire, John Rogan, Mario Chica- [13] Christiana Figueres, Hans Joachim Schellnhuber, Gail Whiteman, Johan Rock- Olmo, and Juan Pedro Rigol-Sanchez. 2012. An assessment of the effectiveness ström, Anthony Hobley, and Stefan Rahmstorf. 2017. Three years to safeguard of a random forest classifier for land-cover classification. ISPRS Journal of Pho- our climate. Nature News 546, 7660 (2017), 593. togrammetry and Remote Sensing 67 (2012), 93–104. [14] American Society for Photogrammetric Engineering and Remote Sensing (ASPRS). [37] Johannes Schreyer, Jan Tigges, Tobia Lakes, and Galina Churkina. 2014. Using 2015. ASPRS positional accuracy standards for digital geospatial data (Edition 1, airborne LiDAR and QuickBird data for modelling urban tree carbon storage and Version 1.0., November, 2014). Photogrammetric Engineering & Remote Sensing its distributionâA case study of Berlin. Remote Sensing 6, 11 (2014), 10636–10655. 81, 3 (2015), A1–A26. [38] Michael Schulson. 2018. Science’s Reproducibility Crisis Is Being Used as Po- [15] GDAL/OGR contributors. 2019. GDAL/OGR Geospatial Data Abstraction software litical Ammunition. Wired (2018). https://www.wired.com/story/sciences- Library. Open Source Geospatial Foundation. https://gdal.org reproducibility-crisis-is-being-used-as-political-ammunition/ [16] Yolanda Gil, Ewa Deelman, Ibrahim Demir, Chris Duffy, Marlon Pierce, Suresh [39] Gunnar M Sørbø and Abdel Ghaffar M Ahmed. 2013. Sudan divided: continuing Marru, and Gerry Wiener. 2012. Designing a Roadmap for Workflow Cyberin- conflict in a contested state. Springer. frastructure in the Geosciences: From Big Data to the Long Tail. American [40] Ian J. Taylor, Ewa Deelman, Dennis B. Gannon, and Matthew Shields. 2014. Work- Geophysical Union Fall Meeting, San Francisco, CA. flows for e-Science: Scientific Workflows for Grids. Springer Publishing Company, [17] Yolanda Gil, Varun Ratnakar, Ewa Deelman, and Christopher Mason. 2012. Using Incorporated. Semantic Workflows for Genome-Scale Analysis. International Conference on [41] Phan Thanh Noi and Martin Kappas. 2018. Comparison of random forest, k- Intelligent Systems for Molecular Biology (ISMB), Long Beach, CA. nearest neighbor, and support vector machine classifiers for land cover classifica- [18] Yolanda Gil, Varun Ratnakar, Jihie Kim, Pedro Antonio Gonzalez-Calero, Paul tion using Sentinel-2 imagery. Sensors 18, 1 (2018), 18. Groth, Joshua Moody, and Ewa Deelman. 2011. Wings: Intelligent Workflow- [42] Jan Tigges, Galina Churkina, and Tobia Lakes. 2017. Modeling above-ground Based Design of Computational Experiments. IEEE Intelligent Systems 26, 1 (2011). carbon storage: a remote sensing approach to derive individual tree species http://www.isi.edu/~gil/papers/gil-etal-ieee-is-11.pdf information in urban settings. Urban ecosystems 20, 1 (2017), 97–111. [19] Kenneth Gillingham, William Nordhaus, David Anthoff, Geoffrey Blanford, [43] Katherine Wolstencroft, Robert Haines, Donal Fellows, Alan Williams, David Valentina Bosetti, Peter Christensen, Haewon McJeon, and John Reilly. 2018. Withers, Stuart Owen, Stian Soiland-Reyes, Ian Dunlop, Aleksandra Nenadic, Modeling uncertainty in integrated assessment of climate change: A multimodel Paul Fisher, et al. 2013. The Taverna workflow suite: designing and executing comparison. Journal of the Association of Environmental and Resource Economists workflows of Web Services on the desktop, web or in the cloud. Nucleic acids 5, 4 (2018), 791–826. research 41, W1 (2013), W557–W561. [20] Clair Gough. 2016. Carbon capture and its storage: an integrated assessment. Routledge. [21] Manuel Grizonnet, Julien Michel, Victor Poughon, Jordi Inglada, Mickaël Savin- aud, and Rémi Cresson. 2017. Orfeo ToolBox: Open source processing of remote sensing images. Open Geospatial Data, Software and Standards 2, 1 (2017), 15. [22] Mordechai Haklay and Patrick Weber. 2008. Openstreetmap: User-generated street maps. IEEE Pervasive Computing 7, 4 (2008), 12–18. [23] Frank P Incropera. 2016. Climate change: a wicked problem: complexity and uncertainty at the intersection of science, economics, politics, and human behavior. Cambridge University Press. [24] Ayako Kagawa and Guillaume Le Sourda. 2018. Mapping the world: cartographic and geographic visualization by the United Nations Geospatial Information Section (formerly Cartographic Section). In Proceedings of the ICA, Vol. 1. 58. [25] ZW Kundzewicz, V Krysanova, RE Benestad, Ø Hov, M Piniewski, and IM Otto. 2018. Uncertainty in climate change impacts on water resources. Environmental