Annotated Bibliography – Greg's Tree Blog

Final Project – Annotated Bibliography
Author: Greg Fletcher
Course: GEOG 560 – GIScience I: Introduction to Geographic Information Science – at Oregon State University

Authors Note

As a field arborist in Calgary, Alberta, Canada, I get to experience firsthand the local urban forest and see how people’s actions are shaping the future of this vital resource. This blog post represents my efforts, as a graduate student at Oregon State University, to better understand the role of geographic information systems (GIS) and GIScience in the field of urban forestry with the explicit goal of informing the direction and efforts of my capstone research project.

For a brief introduction to my capstone project and how GIS and GIScience will be used within the scope of that project see this blog post.

A glossary of abbreviated terms can be found at the end of this article.

[1] Baines, O., Wilkes, P., & Disney, M. (2020). Quantifying urban forest structure with open-access remote sensing data sets. Urban Forestry & Urban Greening, 50, 126653. https://doi.org/10.1016/j.ufug.2020.126653

This article presents a very current discussion regarding the possibility of bridging traditional urban tree inventory practices with remote sensing techniques to augment monitoring and assessment of the urban forest (UF) resource. The authors employ open-access satellite data in the UK, including LiDAR, and open-source/open-access software tools(QGIS, GDAL, i-Tree, LASTools, Scikit Learn, etc.). This effort aimed to establish a framework for predicting forest structure, including tree canopy height, tree canopy cover, and tree density. Demonstrating this framework’s transferability was also of key importance with work initially being conducted in greater London, UK, and further tested in Southampton, UK.

The authors have demonstrated that the techniques employed effectively generated wall-to-wall maps but are not a substitute for existing inventory practices. The framework developed provides an accessible and transferable method for rapidly generating information that can be used to model and assess the urban forest. Potential broader applications identified include mapping habitat connectivity, quantifying UF impact on public health, and air quality modeling.

My capstone requires the mapping of urban forest structure to understand the impacts of densification. This study demonstrates the validity of open-source software in both remote sensing and urban forestry, highlighting the accessibility of these methods and the ability to map urban forest structure. The authors also provide an in-depth treatment of techniques used to verify the accuracy of estimation and underscore the importance of training data for this framework’s machine learning aspect.

[2] Banzhaf, E., Kollai, H., Kindler, A. (2020) Mapping urban grey and green structures for liveable cities using a 3D enhanced OBIA approach and vital statistics, Geocarto International, 35:6, 623-640, https://doi.org/10.1080/10106049.2018.1524514

The authors in this research article describe how OBIA based remote sensing techniques coupled with population statistics are combined with mapping both grey and green infrastructure while providing a social context for Leipzig, Germany. This study’s key objectives were to analyze how grey and green infrastructure varies spatially and how this spatial variation in urban structure influences access and inequalities. It is thought that by combining LULC analysis with population, statistics, value is added, which will enable the development of healthier, resource-efficient, climate-adapted cities with a more equitable distribution of urban resources. Some key findings from this research are as follows:

Quantification of grey and green infrastructure at fine spatial scale provides an opportunity to visualize urban structures in light of population statistics to detect linkages and inequalities.
Underscores the need for better quality input data LiDAR/point cloud data would have increased classification accuracy.
Need for the examination of alternative classification methodology.
An interesting and surprising result of the study – obesity is not simply correlated to SES, which is in contrast to other studies.
Transdisciplinary work requires a common language and understanding. The language of remote sensing is jargon-heavy and very technical making it not easily understood or communicated to the uninitiated.

The authors provide a unique perspective by linking LULC mapping with population statistics. This is an exciting methodology, although not directly translatable to my capstone project, that gives insight into the transdisciplinary focus required to better understand the urban environment’s socio-ecological nature.

[3] Blaschke, T. (2010). Object based image analysis for remote sensing. Isprs Journal of Photogrammetry and Remote Sensing, 65(1), 2–16. https://doi.org/10.1016/j.isprsjprs.2009.06.004

The author in this article presents a comprehensive (although now historic) literature review examining the history and methodology of object based image analysis (OBIA) in remote sensing. It is argued that OBIA represents a significant trend in remote sensing and GIScience and serves to bridge these fields while providing improvements upon pixel-based image processing. The author attributes the recent rise in OBIA techniques and the corresponding increase in published research as a consequence of the availability of high-resolution satellite data and the advent of commercial software capable of object based image analysis. Some key findings from the author’s review are:

OBIA is a compelling image processing technique, and it can be superior to pixel-based processing methods.
Creating objects from geospatial data through image processing mimics human-based perception and analysis.
OBIA provides for the integration of spatial and spectral information, which can further enhance LULC analysis.
OBIA research is moving towards a focus on automation and generating geographic information allowing image-objects to be more easily integrated into a GIS.
OBIA can generate multiscale information (single trees to forest stands) and attenuate the modifiable areal unit problem to some degree.
Author questions whether GEOBIA can be classified as a paradigm in reference to Kuhn’s discourse on the nature of paradigms.

This highly cited article provides background context for my capstone project regarding OBIA, which is now a prevalent methodology for analyzing HSR remotely sensed data. Although much has transpired in OBIA since this article was written, it has deepened my understanding of the foundations of this methodology.

[4] Cimburova, Z., & Barton, D. N. (2020). The potential of geospatial analysis and Bayesian networks to enable i-Tree Eco assessment of existing tree inventories. Urban Forestry & Urban Greening, 55, 126801. https://doi.org/10.1016/j.ufug.2020.126801

This article examines the use of geospatial techniques and machine learning to augment existing municipal tree inventory data in Oslo, Norway to facilitate a more meaningful valuation of the UF using i-Tree Eco. The authors employ high-resolution LiDAR data and additional spatial data sets to estimate missing tree inventory attributes. The resulting efforts enabled the researchers to improve record suitability from19% to 54%. Several key findings from this research are as follows:

Machine learning was successfully used to infer unobserved characteristics, although a cautionary note about the accuracy of value estimation using interpolation techniques is expressed.
Species recognition would have further boosted record suitability to over 9 percent.
The study demonstrates the value of geospatial techniques and machine learning to bolster tree inventory data in a time-efficient and cost-effective manner. These techniques can further be used to bridge tree inventory data gaps in areas that lack accessibility (private land).

This article’s value for my capstone research is that it provides a framework to consider for employing remote sensing techniques to augment existing tree inventory data for ecosystem service estimation. Also, when one considers how much of the UF is composed of trees on private land, this framework could provide a potential mechanism for incorporating data unavailable into ecosystem service models.

[5] Grippa, T., Lennert, M., Beaumont, B., Vanhuysse, S., Stephenne, N., & Wolff, E. (2017). An Open-Source Semi-Automated Processing Chain for Urban Object-Based Classification. Remote Sensing, 9. https://doi.org/10.3390/rs9040358

The authors provide a compelling argument for creating and adopting an open-source toolchain for urban LULC mapping by examining two case studies with disparate urban geographies (Burkina Faso vs. Belgium). This research aimed to contribute to the development of a transparent, adaptable, and transportable solution that is both customizable and freely available. In this study, the authors demonstrate that open-source software, data, and tools are valid, freely available alternatives to commercial offerings for processing remotely sensed data for urban LULC analysis. The development and subsequent re-injection of this framework back into the open-source community provides the potential for increased uptake of a sophisticated remote sensing toolchain by removing barriers such as cost and technical acumen. This framework’s open-source nature also allows researchers and users to vet, replicate, and modify as required, unlike commercial offerings.

My capstone requirement implies that a LULC change detection analysis needs to be leveraged either from the work of others or performed by me. This research provides a methodological model that could be employed to perform change detection analysis if efforts are required. Also of great potential stemming from this research effort is that the necessary toolchain to perform LULC analysis is freely available and adaptable in the form of open-source software and jupyter notebooks.

[6] Horning, N., Fleishman, E., Ersts, P. J., Fogarty, F. A., & Zillig, M. W. (2020). Mapping of land cover with open-source software and ultra-high-resolution imagery acquired with unmanned aerial vehicles. Remote Sensing in Ecology and Conservation, 6(4), 487–497. https://doi.org/10.1002/rse2.144

The authors describe recent experiments using unmanned aerial vehicles (UAV) capable of capturing ultra-high-resolution (UHR), sub-decimeter spatial resolution, imagery for land cover mapping of the sagebrush/shrub-steppe ecosystem of Central Nevada (Great Basin). The experiments’ goal was to test 4 methods of automated classification data processing utilizing machine-learning techniques and open-source software. The objective was to develop an automated system that can generate a data source for land cover mapping that exceeds trained humans’ accuracy, thus being faster, more accurate, and more objective. The authors determined that Automated classification of UHR imagery is still difficult. Although progress has been made, some critical insights as follows:

The importance of training data is not to be underestimated, and methods to select and evaluate training data is critical to accuracy when working with machine learning.
Systematically comparing workflows is not straightforward given the differences between low altitude UHR imagery and more coarse high-resolution imagery and the algorithms/workflows employed.
Open-source software is viable for image processing, with many recent advances being available in this format.

This current account of using UAVs to map areas with UHR imagery provides a view into what is likely to be an important future avenue for data collection in conservation-oriented fields. Although not directly applicable to my capstone research, as this study focuses on rural land cover mapping, it is relevant to keep an eye to the future; as this provides perspective when considering how to deal with current challenges and possible next steps in both research and practice.

[7] Kucharczyk, M., Hay, G. J., Ghaffarian, S., & Hugenholtz, C. H. (2020). Geographic Object-Based Image Analysis: A Primer and Future Directions. Remote Sensing, 12(12), 2012. https://doi.org/10.3390/rs12122012

This paper presents a current review of geographic object based image analysis (GEOBIA) examining the theoretical foundations, best practices, methodology, and anticipated future directions. The authors’ primary objectives were to provide an accessible GEOBIA resource for novice practitioners while delivering insight and depth for those more experienced.

The authors’ review is an excellent reference providing contextual detail regarding the practice of GEOBIA while giving a very current synoptic view of both commercial and open-source offerings. Treatment of future directions, including discussion of machine learning (ML), provides insight into the benefits of ML techniques and challenges. The accompanying reference list is also of tremendous value providing a detailed reading list regarding GEOBIA.

[8] Kyttä, M., Broberg, A., Tzoulas, T., & Snabb, K. (2013). Towards contextually sensitive urban densification: Location-based softGIS knowledge revealing perceived residential environmental quality. Landscape and Urban Planning, 113, 30–46. https://doi.org/10.1016/j.landurbplan.2013.01.008

This article details the use of public participation GIS (softGIS) and its potential role in urban planning with the objective of understanding how residents perceive the local urban environment in light of urban densification efforts. The study solicited feedback, via a web-based survey, from the residents of two urban districts (Helsinki and Espoo) Finland. Feedback was classified as either positive or negative utilizing four dimensions of perceived quality (functional, social, visual environment, and atmosphere) with places being marked on a web-based map.

The authors demonstrate that location-based experiential surveys can provide a valuable contribution to urban planners as such spatially explicit, context-sensitive data is not ordinarily available. It is expressed that understanding what residents value and why, is critical when considering urban densification and renewal.

My capstone project intends to examine the impact of urban densification on the urban forest. Therefore, understanding ways to verify citizen perceptions of value regarding densification is highly relevant. The authors’ thesis of finding ways to enable urban renewal in a spatially explicit and context-sensitive way will provide value as I engage further in my research.

[9] Li, X., Chen, W., Sanesi, G., & Lafortezza, R. (2019). Remote Sensing in Urban Forestry: Recent Applications and Future Directions. Remote Sensing, 11. https://doi.org/10.3390/rs11101144

This paper summarizes the use of remote sensing in urban forestry from a current perspective by reviewing peer-reviewed literature spanning between January 2013 to March 2019. The inclusion criteria for articles to be considered for review was that articles must contain one or more of the three following themes: multi-source, multi-temporal, or multi-scale; must focus on UF or UGS; must use remotely sensed data that is airborne/spaceborne or utilizes LiDAR data.

The authors surmise that although UF can benefit from remotely sensed data, several challenges need to be addressed. Challenges are highlighted as follows:

Need for algorithms to automate multi-source and multi-scale data handling.
Modelling techniques need to better incorporate temporal/spatial resolution to better predict future outcomes based on past/current management strategies.
Multiscale analysis is difficult to perform given the complexity/heterogeneity of the UF; thus, care should be used when employing remote sensing in land use policy and decision making.

The focus of this research provides context for my capstone project by illuminating current research trends and highlighting challenges and future directions for research specific to remote sensing and UF. What is of significant importance is the thematic classification of recent research, accompanied by comparing the strengths and weaknesses of relevant remote sensing data sources.

[10] Lwin, K. K., Murayama, Y., & Mizutani, C. (2012). Quantitative versus Qualitative Geospatial Data in Spatial Modelling and Decision Making. 2012. https://doi.org/10.4236/jgis.2012.43028

The authors provide an informative view examining the quantitative and qualitative characterization of geospatial data. The general premise is that most raster datasets can be classified as quantitative due to data capture nature through mechanized instrumentation (sensors). On the other hand, Vector datasets often represent qualitative data due to the nature of human constructs captured (e.g., property boundaries or land use classifications). Although somewhat dated in a heavily technology-focused field, the article provides a relevant perspective regarding the nature and classification of geospatial data and how it should be utilized.

My capstone project’s underlying premise will rely on urban LULC change detection, so understanding the underlying data characterization will provide fundamental grounding.

[11] Moskal, L. Monika, Styers, Diane M, & Halabisky, Meghan. (2011). Monitoring Urban Tree Cover Using Object-Based Image Analysis and Public Domain Remotely Sensed Data. Remote Sensing (Basel, Switzerland), 3(10), 2243–2262. https://doi.org/10.3390/rs3102243

This study represents an effort to generate a methodology capable of generating LULC mapping for tree cover assessment that is accurate and repeatable. The study utilized OBIA methods and open-access data and ancillary datasets of the Seattle, WA, US area. This study consisted of three case studies. Each case study utilized datasets that employed hyperspatial resolution imagery (NAIP and aerial). The difference between each case study was primarily based on collection methods and spectral resolution.

The authors demonstrate that OBIA methods can generate repeatable tree cover assessments in a temperate urban environment utilizing open-access hyperspatial data. Key findings of this study are as follows:

OBIA techniques using NAIP imagery generated higher accuracy LULC maps than more coarse pixel-based NLCD datasets.
Spectral content appeared to be more beneficial than spatial content for tree cover assessment.
Pixel size using OBIA techniques is important – understanding the size of the average tree crown could serve as a guide for required spatial resolution.

In terms of my capstone project, this paper details an example study that can be used as a model for mapping tree cover in heterogeneous urban environments, including guidance on methodology including dataset selection, classification hierarchy, and accuracy assessment.

[12] Rall, E., Hansen, R., & Pauleit, S. (2019). The added value of public participation GIS (PPGIS) for urban green infrastructure planning. Urban Forestry & Urban Greening, 40, 264–274. https://doi.org/10.1016/j.ufug.2018.06.016

The authors chronicle a case study (Berlin, DE) investigating the use of public participation GIS (PPGIS) to enhance urban green space (UGS) planning. The manuscript details the shortcomings of traditional methods employed in the public participation step of UGS planning and outlines how PPGIS provides a spatially explicit integration of public knowledge, experiences, values, and preferences into the decision-making process.

The paper provides a template to consider for linking public perception to UGS conditions and proposed changes. PPGIS appears to be well suited for greater uptake in the planning process, given the ubiquitous nature of web-based mapping platforms, the ease with which public participation can be linked with GIS, and the need to make UGS planning more relevant and effective.

[13] Reba, M., & Seto, K. C. (2020). A systematic review and assessment of algorithms to detect, characterize, and monitor urban land change. Remote Sensing of Environment, 242, 111739. https://doi.org/10.1016/j.rse.2020.111739

Reba and Seto present a review of remote sensing change detection algorithms that depart from the traditional perspective of examining technique and instead focus on the type of information provided to the user community. The researchers posit that the scientific community is producing more literature than it can consume and hence the need for systematic reviews. The objectives presented in this review are to provide the user community with a catalog of the algorithms in use based on change being measured and to highlight gaps in knowledge to the remote sensing research community. The authors highlight their key findings as follows:

Increased need to investigate and characterize urban change beyond large cities in the global north.
Urban change studies need to look beyond urban expansion and look at intra-urban change and dynamics.
Change detection research needs to increase temporal resolution to more adequately characterize change.
Remote sensing algorithms are difficult to compare; thus, a greater focus on reproducibility, replicability, and comparability of research and algorithms results is required.

This review provides a highly beneficial context to position my capstone project when looking at the authors’ key findings. The focus on the outcome of information generated rather than technique evaluation also helps to better understand and select a change detection methodology.

[14] United Nations Economic and Social Commission for Asia and the Pacific (ESCAP) (2020). Producing land cover change maps and statistics: Step by step guide on the use of QGIS and RStudio. Retrieved from https://www.unescap.org/resources/producing-land-cover-change-maps-and-statistics-step-step-guide-use-qgis-and-rstudio

This guide serves as a tutorial on using open-source software (QGIS and RStudio), open-access data from the World Bank, and open-access data from the European Space Agency. The steps in this guide allow one to produce a land cover change map and perform statistical analysis. Working through this guide serves as a practical tutorial regarding the capabilities of both QGIS and RStudio while underscoring the benefits of open-access data and open-source software tools in terms of reducing the barrier of entry to GISS.

[15] Zheng, Y., Lan, S., Chen, W. Y., Chen, X., Xu, X., Chen, Y., & Dong, J. (2019). Visual sensitivity versus ecological sensitivity: An application of GIS in urban forest park planning. Urban Forestry & Urban Greening, 41, 139–149. https://doi.org/10.1016/j.ufug.2019.03.010

This article provides an example of combining GIS and multi-criteria decision making (AHP in ArcMap) through a case study approach of a planning and development effort in Tianzhu Mountain National Forest Park in Fujian Province, China. The study area represents an urban forest park with two large cities in close proximity with a corresponding demand for recreation and conservation opportunities. The authors’ objective in undertaking this study is to investigate the integration of visual and ecological sensitivity through GIS and AHP to produce a map that can be used to balance development and conservation efforts in a scientific manner. This effort demonstrates that it is possible to use readily available remote sensing data to enable the integration of multi-use criteria for UGS planning and decision making.

When examining this article with my capstone’s scope in mind, I see the direct relevance to the GIS-AHP approach employed and the need to balance efforts like urban densification and urban forest quality.

[16] Zhou, W., Wang, J., Qian, Y., Pickett, S. T. A., Li, W., & Han, L. (2018). The rapid but “invisible” changes in urban greenspace: A comparative study of nine Chinese cities. Science of The Total Environment, 627, 1572–1584. https://doi.org/10.1016/j.scitotenv.2018.01.335

The authors of this study looked at nine Chinese cities to examine the spatiotemporal change of UGS in urban core areas both within and across cities. HSR imagery is used with varying spatial resolutions to compare the effectiveness of resolution on change detection. This research attempts to address a gap in urban LULC analysis, which has primarily focused on the loss of UGS at the urban-rural interface due to urban expansion. The authors posit that past research on UGS dynamics has relied on medium resolution imagery, which does not have sufficient resolution to detect and characterize rapid, fine-scale localized change.

This article provides a foundational context for my capstone that underscores the dynamic nature of the urban environment and the requirement to use remotely sensed data with adequate resolution to detect significant fine-scale change. From understanding this change, we can begin to link underlying actions that impact and benefit derived from the UF resource.

References

[3] Blaschke, T. (2010). Object based image analysis for remote sensing. Isprs Journal of 2Photogrammetry and Remote Sensing, 65(1), 2–16. https://doi.org/10.1016/j.isprsjprs.2009.06.004

[9] Li, X., Chen, W., Sanesi, G., & Lafortezza, R. (2019). Remote Sensing in Urban Forestry: Recent Applications and Future Directions. Remote Sensing, 11. https://doi.org/10.3390/rs11101144

[10] Lwin, K. K., Murayama, Y., & Mizutani, C. (2012). Quantitative versus Qualitative Geospatial Data in Spatial Modelling and Decision Making. 2012. https://doi.org/10.4236/jgis.2012.43028

Glossary

AHP Analytic hierarchy process

ArcMap Desktop GIS software from Esri https://desktop.arcgis.com/en/arcmap/

BN Bayesian networks

ESCAP Economic and Social Commission for Asia and the Pacific

Esri Software company specializing in GISS https://www.esri.com/en-us/about/about-esri/overview

GDAL Open-source software for raster and vector geospatial processing https://gdal.org/

GEOBIA Geographic object based image analysis

GIS Geographic information system

GISS Geographic information system and science

HSR High spatial resolution

i-Tree Open access software for quantifying the benefits of trees https://www.itreetools.org/

LASTools Open-source software for processing LiDAR data http://lastools.org/

LiDAR light detection and ranging

LULC land use land classification

MAUP Modifiable areal unit problem

ML Machine learning

NAIP – National Agriculture Imagery Program https://eos.com/naip/

NLCD National Land Cover Database https://www.usgs.gov/centers/eros/science/national-land-cover-database?qt-science_center_objects=0#qt-science_center_objects

OBIA Object based image analysis

PPGIS Public participation GIS

QGIS Open-source GIS software https://www.qgis.org/en/site/

RStudio Open-source for data analysis https://rstudio.com/

Scikit Learn Open-source machine learning software written in python https://sklearn.org/

SES Socioeconomic status

UAV Unmanned aerial vehicle

UF Urban forest/urban forestry

UGS Urban green space

UHR Ultra-high resolution