Spatial analytics and deep learning in GIS
The future will be of Spatial Analytics – based on principles of Data Science, Artificial Intelligence and Deep Learning. Spatial data is voluminously available for human activity – triggered by the Big Data and cloud technology
Today, information is key to the success of a citizen, society, governments and humanity as a whole. Our government’s vision is to bring a new paradigm for governance and development with emphasis on participatory approach from communities and citizens; enable a scientific mapping of the resources, examine needs and aspirations of citizen beneficiaries and society; support sustainable development and spatial planning; assist quick and reliable monitoring of plan implementation and status of development; enable transparent systems for inclusivity of society and support real-time mapping of feed-back and redressal systems. Spatial Analytics can play an important role and support many aspects of national economic development and governance process.
Geospatial technology has made spectacular advances over the past half a century since the early 1970s when the computerisation of maps using primitive manual coding and card-punching gave rise to “store and print maps” at will. At the same time, the satellite images from ERTS (Landsat) gave details and perspectives of every part of the Earth as never seen before. India’s tryst with remote sensing also started in early 1970s as a unique way of building applications by involving user departments – called the Joint Experiments Programme (JEP) which creating a rich experience of applications and user service orientation. Way back in 1980s, we started the GIS activities and initiated the concept of Natural Resources Information System (NRIS) – a spatial system demonstrating support for decision-making process through many pilot-projects, particularly focussing on natural resources as well as planning and development . Over the years, India has developed successful activities in Imaging (series of IRS satellites, aerial platforms), Mapping (topographic and thematic), positioning and surveying and GIS (databases and applications) – both in government and commercial domain. GIS technology is widely used and a good knowledge-base has been created over the years. In fact, India has been at the forefront at some unique systems – the National Natural Resources Management System (NNRMS); the NRIS and Integrated Mission on Sustainable Development (IMSD); the National Spatial Data Infrastructure (NSDI); National Urban Information System (NUIS); Bhuvan and more recently National GIS – these were all unique India-centric systems that would have brought scientific decisionmaking and benefits to governance and citizens in a significant way. While these systems definition were successful, I see that the systems implementation has been not that successful and we have a lot of catching up to do now. However, GIS based initiatives (or projects) of the various departments of the Government; efforts at modernization of land records; various City-GISs; GIS initiatives of the states (of particular mention is the wonderful work done by Karnataka for a Karnataka- GIS which was spearheaded by KJA); GIS technology in private sector services and research activities in universities have helped bring good and operational examples of applications of GIS.
In spite of fairly wide usage of GIS as a technology, the full potential of GIS has not been exploited for decisionsupport by planners, stake holders for governance-process, decision-makers and also bring direct benefit to citizens and many others. Some of the initiatives have certainly been successful to prove GIS application potentials in a “project mode” but GIS is yet to get a “service orientation” and get assimilated to become a part of the work-process of governance, planning and nation-building in a significant manner. There is also a gap in technology development levels – we need to maintain a high-level of national capability in this important technology area and also leverage to be in the fore-front of GIS technology at the international arena. These two important aspects are at the background of many recent endeavours and discussion – we need a forward plan of action for Spatial Technology and Applications.
In 1970s, the birth of GIS was with simple data models and a monolithic computer software, performing a wide range of functions that included capture, storage, and analysis of spatial data. Over the subsequent decades into the 1990s, the set of software functions in a GIS proliferated, as did the underlying data models become complex, in the interests of supporting new data, new applications and becoming decision-oriented on mainframes to minicomputers, and ultimately to the desktop and now to handheld devices. In the 1990s, two very significant changes to this early vision of GIS set the goals for Spatial Data Infrastructures – large volumes of maps, images and spatial data had become digitalised by then; first global coverage of satellite images were available digitally and the satellite based Global Positioning System (GPS) had greatly reduced the cost of measuring location and location mapping. Yet another development was that a large number of research had been generated and large workforce and industries, apart from government and academia, were engaged in the GIS ecosystem across the world. The need to share the digitalised spatial data across user communities triggered the vision of a National Spatial Data Infrastructures. World-over large initiatives were taken up in various nations to organise SDI and National GIS systems. India too embarked on a vision to organise a NSDI in 2002 to bring a collective effort at aligning the geospatial community and various demonstrations of NSDI were taken up.
The birth of the www and internet in the 1990s boosted the “sharing” vision and also eased the technologies that were required for sharing the spatial data and in generating and storing them on the web. The commercial aspects of spatial data became a possibility and the world saw many a g-marketplace emerge for commercial spatial data availability. By the time of the 2000s the SDI vision had been over-whelmed by GIS on the Web and now GIS on the cloud. Very quickly the world saw Google Earth, Digital Earth initiatives took a great boost and was available to every citizen.
In the 2000s, we see another major change – giving a twist to the SDI vision to the vision of GIS Systems of System. Creating spatial data, maps, images – which hitherto were time-consuming and costly suddenly transformed to one of rapid process and also cost effective – thus, various technologies and sensors captured location and vast amounts of data got collected very easily across the Earth – data collection which soon made a citizen a “sensor” through on always-connected device, the mobile phone. So today it is easy to create maps of phenomena such as road congestion that are valid only for a few minutes and available free to the user through mapping apps on a mobile device. Sensors in the environment are now able to create an abundance of timedependent geospatial data, making time an important element of today’s vision of geospatial technology. Similarly, contemporary technology such as GPS,
BIM (building information management), structure-from-motion, platform- mounted LiDAR, and ground-penetrating radar have made it possible to acquire threedimensional representations of geographic features. Today, the basic element of spatial data is no more x,y but <x,y,z,t,a> where x, y-a define the traditional coordinate and z are the height/depth reference in three- dimensional space, t is time, and a is an attribute of that location in spacetime. Today, this is how spatial data is characterised – giving rise to the next vision of GIS – Big Data GIS – what I call as GIS Systems of Systems. I must tell all of you that the first of concept of such a GIS Systems of Systems was defined in India – National GIS which was a comprehensive definition of India’s leadership in the global arena of GIS. The National GIS initiative, driven by Dr Shailesh Nayak and Dr Mukund Rao, was a result of exhaustive consultation and discussions within Government, industry and academia and in 2012 the final plan of action was defined. We have not made much progress from that definition stage – and many of us are realising that in 2020 we need to go back to drawingboards and define of vision ahead again.
Where does one see the GIS technology and applications go from here? It is entirely reasonable, therefore, that the geospatial community should chart a path that can help and give India a stable direction. Where are we going with this technology, and how can it address the challenges that we currently face, and will face in the coming years? With the past experience of past 40-50 years, tremendous possibilities are there that can change the way India will deal with spatial technologies and applications – it becoming just one “bit” of data and applications in the complex Big Data Systems of System exploding on the internet.
Today, geospatial technology extends into virtually all areas of human activity helping to solve problems and make decisions, predict outcomes, and discover and explain how the Earth’s environmental and social systems work and is reaching each and every citizen on this Earth. The fusing of spatial data with other data allows its users to get “insights,” by using maps, images and statistics to reveal what would otherwise not be visible, appreciated, understood and experienced by one and all. Government agencies, private sector, academia, and non-profit organisations use GIS in many aspects of their businesses and activities. Individual citizens encounter GIS in mapping and wayfinding apps, and in apps that provide information that is customized to the user’s location. Just like a human nervous system collects signals from many parts of the body and transmits them to the brain, where they can be processed and combined, leading eventually to actions, we now have a spatial nervous system of the Earth – characterising how the planet is sensed, how the resulting data are transmitted and to whom, how the various signals are conflated and stored, and how the costs of all of this are met
The future will be of Spatial Analytics – based on principles of Data Science, Artificial Intelligence and Deep Learning. Spatial data is voluminously available for human activity – triggered by the Big Data and cloud technology. Data Analytics is key to obtain Insights and what I see for the future is that Spatial Insights will drive the vision for all of us for using more spatial data, maps, images and in real-time and arrive at key answers to questions that citizens, society, government, industries, researchers may have for their contributions for the betterment of humanity and our Earth. So S-Insights or Spatial-Insights would be the way of development of this technology further. The next few years will see massive transformation, in society as a whole but also in geospatial technology. S-Insights will be the ultimate goal – but with due regard to openness, access, and engagement, the geospatial infrastructure of the future will integrate information, processes, and workflows, and capture sufficient information to support replicability across the world. It will develop on power of technology that is anticipated in the 4I revolution – keeping the conflation, resampling, upscaling, and fusion of S-Insights to bring benefit to one and all.
Data science is emerging as a major element of knowledge and profession and is core to make sense of the vast collections of big data systems. Data science is a set of fundamental principles that can easily support and guide the extraction of insights from data. In processing data for new insights, mathematics, statistics, logic and reasoning, modelling, deep learning and artificial intelligence etc will be an important capability. For data to be processed, engineering the data itself will become so important – already Data Engineering is becoming a new paradigm for preparing the vast data so that it becomes amenable to the analysis process easily and repeatedly. Outputs would be as simple as answering statements; visualisation data products; a easy to understand GIS dashboard or any product that supports achieving a more informed decision. Spatial Analysts will drive the Big Data of GIS and make it easily amenable to Deep Learning processes where the vast data itself will provide triggers to further analysis. For example, the vast collections of IRS satellite images can itself provide the spectral signature libraries to automatically classify and segment the newer images and detect changes over time – this autonomous classification algorithm can be easily derived from the existing images that are available from 1987 onwards. Similarly, the census data collection from 1950 onwards can itself provide predictive logic and models of how Indian demography can trend for next 10/20/50 years – thereby learning trend algorithms from the census data holdings. Or say, daily traffic data in a city can itself provide autonomous road-load indicators across every hour in a city and model traffic-loads on a daily basis. There can be numerous examples by which large GIS data collections can be the self-learning crucibles and be “answers” to questions that various problems that emerge in society. In fact, GIS databases are naturally characterised for Deep Learnings and artificial intelligence models because they are a structured, referenced collection of time-stamped pixels, vectors, rasters, data points making them so easy for Dl and AI models. Spatial Analytics is very important and India needs to build its capabilities and capacities in this area for the future.
This article is based on Dr K Kasturirangan’s inaugural address at International Workshop Advanced Spatial Analytics and Deep Learning for Geospatial Applications held during January 20 – 31, 2020 in Bengaluru, India.