Home > Free Essays > Tech & Engineering > Other Technology > Geographical Information System and Machine Learning

Geographical Information System and Machine Learning Research Paper

Exclusively available on IvyPanda Available only on IvyPanda
Updated: Feb 6th, 2022


There is no doubt that GIS is an integral part of the modern technical basis for Earth exploration. If one looks at the study of Burrough and McDonnel (1998), one can see that the authors understand GIS as a tool for creating, storing, and exploring information. Thus, the field of science and technology in which this tool finds application includes an environmental study of territories, their mapping, and reliable assessment (Longley et al., 2005). However, it is fair to admit that GIS technologies have undergone fundamental changes in recent decades, making it possible to use them in other areas as well. In particular, hydrological research is a promising area of practical application (Chow et al., 1988; Clark, 1998). For thousands of years, floods have been known to be a severe environmental problem that threatens not only human industry but also natural terrestrial ecosystems.

This situation necessitates the prediction of potential flooding sources, for which Digital Elevation Models (DEMs) are appropriate. Specifically, such models make it possible to evaluate complex geographical forms of catchments by studying river systems, stream headings, and basin-specific land use, soil, and climate information (Siart et al., 2009). This means that, according to Wang et al. (2011), GIS is widely used for analytical work on the quantitative and subjective effects of floods and runoff. Moreover, the use of spatial models addresses the challenges of predicting and investigating sources of coastal flooding (Sarker and Sivertun, 2011; Zerger and Wealands, 2004). Finally, in addition, it should be recognized that the combination of traditional GIS forms with LiDAR technologies can modify practical approaches and improve procedures for obtaining and processing results (Merwade et al., 2008; Gallegos et al., 2009).

Hydrologic Models

Technical solutions are essential to critically assess the potential for inundation, especially in coastal areas. Until recently, the range of technologies used was limited to hydrological models, such as the reproductive model. This model has been particularly relevant in the context of the effects of climate change: in other words, it has made it possible to identify problems in the early days to support complex management decisions (Messner and Meyer, 2006). However, according to Djokic and Maidment (1993), today’s world dictates new rules and the GIS environment can be an excellent alternative for obtaining more in-depth knowledge. Therefore, finding solutions to integrate hydrological models and spatial GIS technologies is a priority. It is known that both sides benefit from such a compromise, so implementation should be initiated as soon as possible (Xu et al., 2001).

The objective of this study is to develop this problem by creating an integrated GIS and LiDAR communication model as a unique solution for determining surface runoff and estimating the risk from inundation in built-up coastal areas for CoW. It should be recognized that this model is unique to modern technology, as a review of existing literature does not allow for highlighting earlier uses of this combination. For example, Pourali et al. (2014b) used conceptual hydrological models along with computer modeling of inundation-sensitive zones for CoW. Based on this, it seems clear that improved models will provide more valuable and complete information on the morphological structure of water basins. It should be further noted that this advanced methodology greatly simplifies the modeling of potential flood threats and provides a logical basis for flood flow estimation. In addition, the use of this combination extends the range of possible threat sources to atmospheric climate conditions, including precipitation.

Machine Learning Techniques

In modern science, there is an increasing trend towards the digitalization of classical research methods. Thus, the introduction of machine learning technologies has the potential to improve the practice of hydrological research on floods significantly. It should be recognized that floods often have sufficient destructive power to cause economic and health damage to settlements (Tehrany et al., 2019). Numerous attempts at analysis have been made to forecast and early warning of developing flooding. However, it is fair to say that because of the lack of knowledge about the factors and conditions of flooding, no one-size-fits-all solution has been developed. Nevertheless, machine learning methods can be used to qualitatively assess the spatial correlations between factors and their significance for mapping. Traditionally, such approaches include studies of texture, spectral and structural features of the region (Kuffer et al., 2016). Among the known learning models that have already found applications in the hydrological analysis are “Decision Tree” (DT) and “Support Vector Machine” (SVM). Although analysis of both pixel and object classifications using machine learning algorithms has already been performed for both DT and SVM, previous studies did not reveal any statistical difference between the two types of classification (Tehrany et al., 2014; Duque et al., 2017; Duro et al., 2012; Wieland et al., 2016).

An analysis of the applicability of existing solutions, both in terms of classical models and more innovative ones, was the initial point in determining the goals of scientific research. In particular, with the development of computer technology, it can be expected that the introduction of machine learning methods in hydrological practice will yield only positive results. There is an established viewpoint on this issue. The academic environment is characterized by an abundance of opinions that the use of innovative models produces comparatively better results than the traditional methods of research (Marjanović et al., 2011; Tehrany et al., 2015; Althuwaynee et al., 2016; Tien Bui et al., 2016c). The range of such work is not limited to hydrological analysis of coastal zones only. A large amount of thematic literature refers to the practical application of machine learning for flood, landslide, bushfires, and land subsidence prevention (Tehrany et al., 2013; Tien Bui et al., 2016b; Reid et al., 2015; Lee and Park, 2013). Moreover, a wide range of algorithmic machine learning methods is widely used for operational flood control. This includes already known CVM and DT, as well as artificial neural networks, genetic code programming, adaptive network-based fuzzy inference system, and some other methods (Kitsikoudis et al., 2015; Saito et al., 2009). Ultimately, this leads to the idea that the use of innovative computer technologies helps to simplify specialist work practices and provide more qualitative and reliable data on environmental threats.

However, it is worth acknowledging that it would be a mistake to assume that stand-alone methods alone could have been highly effective. In other words, neither classical methods nor the use of machine learning models alone will achieve efficiency in the context of flood studies. On the contrary, the integration of these ideas is of high value. Thus, numerous scientific works demonstrate high efficiency from the use of DT and SVM, along with well-known technologies (Pham et al., 2016; Tehrany et al., 2013). The disagreement between studies begins at the moment of discussion of which of the methods shows higher efficiency for hydrology: DT and SVM. In particular, Hong et al. (2015), Pradhan (2013), and Singh et al. (2009) found that the DT is more reliable than the SVM. In contrast, Marjanović et al. (2011), Wu et al. (2014), Tien Bui et al. (2012) found the opposite effect. Without the need for a more detailed discussion of the advantages and disadvantages of each method, it is essential to postulate that both DT and SVM have sufficient potential to improve flood modeling in hydrological expertise. For this reason, this study stops at examining the extent to which external and internal factors influence the procedure for mapping areas vulnerable to flooding.

As mentioned earlier, there is no doubt that GIS technology has enormous potential for early detection and prevention of flood consequences. However, as practice in recent years has shown, it is becoming increasingly important to find ways to harmoniously combine different technical methods and switch to spatial methods of assessment. Thus, this approach allows for solving two problems at once. First of all, using several methods reduces the possibility of errors and inaccuracies. On the other hand, the combination of techniques can significantly expand the range of the device.

For this reason, this paper explores the possibility of harmonic integration between the GIS hydrological model and machine learning methods. In other words, this study has a two-way orientation. First, it seeks to shed light on the use of GIS-based hydrological models. Secondly, it analyses differential machine learning tools and assesses their contribution to the hydrological model.

Based on the above, it should be recognized that CoW has recognized the need for urgent scientific project development to expand current hydrological directions. In particular, recognizing a deep gap in research, CoW has formed a plan to incorporate digital data from LiDAR into current hydrological spatial technologies (Baby et al., 2019). As can be seen, this is a very multidimensional and complex work, the implementation of which is possible in solving specific problems. Five objectives have been identified for the success of the project, the achievement of which is of paramount importance to the study.

  1. First, the practice of mapping water flows should be used to make decisions about land use and the development of vacant areas.
  2. Second, a flood and drainage management program must be developed to ensure that risks to human communities and natural ecosystems are reduced.
  3. Third, it is essential to assess the available data in the CoW regarding government regulation of the problem, including the Local Government Spatial Strategy v2.2.
  4. Fourth, it is necessary to assess existing technologies and IT barriers and constraints that hinder the functional development of projects for spatial information management in the Department of Environment and Primary Industry and other critical indicators (Rajabifard et al., 2002).
  5. Finally, it is necessary to achieve the conditions under which machine learning methods can be used to assess spatial correlations between potential flood causes, as well as their level of significance for the vulnerability mapping procedure.

It seems evident that a significant challenge in the implementation of this project is to model the form then it would be possible to integrate different methods and provide relevant information. To solve this problem, an interface was created, the purpose of which was to provide information about the project to all interested parties. This approach allowed to lay the foundation for this study, as it supported GIS users in project groups for decision-making.

The overall objective of this study is to design a support system for decision-making in the context of flood management. It should be emphasized that this project is being implemented in Victoria, Australia. In particular, given the objective of the project, efforts were focused on finding a harmonious mechanism for integrating the methods described. Thus, a hydrological model based on GIS has been developed, in which machine learning elements have been introduced. This approach is expected to significantly support the implementation of floodplain management strategies based on an analysis of potential risks from floods.

As with any technology with the potential to become commonplace, this project requires preliminary tests to determine the degree of effectiveness in hydrological practice. For this purpose, the CoW has been selected as a pilot area where distributed hydrological models and methods for monitoring potential flood sources both in the road, and building areas will be assessed. The technical implementation of the study is based on the use of high-resolution digital data provided by LiDAR and crowdsourcing information. It should be admitted that the number of technological solutions is continuously growing, but two technologies have been chosen from the whole variety: ArcHydro and ArcSWAT. Thus, ArcHydro is an extension of the tool for modeling based on hydrological data, ArcGIS. On the other hand, ArcSWAT is an alternative tool included in the ArcGIS package aimed at soil and water assessment. It is fair to say that the second extension is slightly more comfortable to use for those who do not have enough knowledge and GIS systems, while ArcHydro is more suitable for professionals with experience: the program provides more technical features and can be customized to the user’s needs. Additionally, it should be noted that among all the variety of GIS systems, including, for example, QGIS QSWAT, were chosen ArcGIS tools, because they are more integrated with the software language Python and geodata.

It is worth noticing that the hydrological model proposed in this study, along with mechanisms for monitoring potential flood threats, will provide an excellent solution for experienced and inexperienced users. The results obtained by the project will inform future predictions of possible coastal inundation for all categories of users regardless of experience and knowledge (Hine et al., 2017). A user with hydrological expertise refers to professional GIS specialists, along with staff from design and modeling offices. As can be seen, the proposed integration solution will benefit them in developing visualization tools. On the contrary, users without sufficient experience are, as a rule, heads, and organizers of departments who use the results to promote their company.

The ArcSWAT and ArcHydro used in this project are similar in general but show different results in practice. The diversity of approaches to the analysis of hydrological systems makes it possible to assess which model is more appropriate for further research. For example, in field tests, it was found that the ArcSWAT provided more qualitative and useful information than the ArcHydro, especially for smaller basins such as the Darebin Creek. On the other hand, ArcHydro outperforms the competition in terms of ease of use and working volumes of data for large basins like Inverloch.

Finally, it should be stressed that built-in GIS-based hydrological models, integrated with machine learning methods, can realize many methodological possibilities. This means that, along with the technical advances of computer-based devices, the era of big data has arrived in hydrology, with many breakthrough discoveries. Automatic methods become the answer to the current needs of specialists. Therefore, this study gives an overview of the most important algorithms of machine learning, which have found application in thematic literature. The main idea of the work is that there is no single method that would be effective in a single-use. On the contrary, only a comprehensive approach that combines unique methods can guarantee effectiveness.

This research paper on Geographical Information System and Machine Learning was written and submitted by your fellow student. You are free to use it for research and reference purposes in order to write your own paper; however, you must cite it accordingly.
Removal Request
If you are the copyright owner of this paper and no longer wish to have your work published on IvyPanda.
Request the removal

Need a custom Research Paper sample written from scratch by
professional specifically for you?

801 certified writers online

Cite This paper
Select a referencing style:


IvyPanda. (2022, February 6). Geographical Information System and Machine Learning. https://ivypanda.com/essays/geographical-information-system-and-machine-learning/


IvyPanda. (2022, February 6). Geographical Information System and Machine Learning. Retrieved from https://ivypanda.com/essays/geographical-information-system-and-machine-learning/

Work Cited

"Geographical Information System and Machine Learning." IvyPanda, 6 Feb. 2022, ivypanda.com/essays/geographical-information-system-and-machine-learning/.

1. IvyPanda. "Geographical Information System and Machine Learning." February 6, 2022. https://ivypanda.com/essays/geographical-information-system-and-machine-learning/.


IvyPanda. "Geographical Information System and Machine Learning." February 6, 2022. https://ivypanda.com/essays/geographical-information-system-and-machine-learning/.


IvyPanda. 2022. "Geographical Information System and Machine Learning." February 6, 2022. https://ivypanda.com/essays/geographical-information-system-and-machine-learning/.


IvyPanda. (2022) 'Geographical Information System and Machine Learning'. 6 February.

Powered by CiteTotal, free citation maker
More related papers