Policy Significance Statement
As NASA, NOAA, and other government agencies embrace open-access policies for satellite data, this paper illuminates the barriers that still exist in accessing that data. Through this paper, we illustrate via a case study how there still exists significant barriers to using satellite data by non-remote sensing experts, particularly for those with little to no prior experience with big data. We then recommend more robust documentation, the creation of walkthroughs, the expansion of cloud computing resources, and reaching out directly to non-remote sensing experts to help remove the barriers encountered in the case study.
1. Introduction
In the current era of data-driven decision-making, Earth observation satellites play a crucial role in providing critical information. Hundreds of government-operated Earth observation satellites are currently in orbit, and many government agencies have embraced open data policies, looking to enable researchers to use the data for the broadest possible set of research issues and applications. Recent efforts, such as the United States’ “Year of Open Science,” declared in 2023, underscore the importance of open science in accelerating discovery, fostering innovation, and driving more equitable outcomes (House, Reference House2023). In the same spirit, NASA’s Earth Science Division has developed a strategy for equity and environmental justice that seeks to ensure “the investment the nation has made in NASA satellites and science benefits people across the U.S. and helps them make informed decisions about the very real challenges they face in their communities” (NASA Science, 2021). The strategy emphasizes a better understanding of data accessibility, identifying barriers to environmental justice, and enabling the integration of physical and social science via NASA datasets (NASA Science, 2021). These efforts recognize that although adopting an open data policy is an important first step, an open data policy on its own is not sufficient to enable the widespread adoption of satellite data.
In this paper, we present a case study of an interdisciplinary research group at the Georgia Institute of Technology (Georgia Tech) that set out to use satellite data for a social science research project. While one of the three professors advising the group had familiarity with the space industry and Earth observation satellite offerings from a policy perspective (Borowitz), none of the members had prior experience analyzing satellite data. This makes the group’s effort a strong example of the challenges and experiences faced by non-remote sensing experts when working with satellite data. Yet, the group still had considerable advantages compared to the average person. All members were pursuing or had already obtained postsecondary degrees, many in fields related to satellite data, such as computer science. Furthermore, their affiliation with Georgia Tech allowed them to easily consult with GIS and satellite data experts both within the university and at organizations like NASA. Despite these advantages, the research group encountered numerous major obstacles. These obstacles highlight the need for further efforts to make satellite data genuinely accessible to non-remote sensing experts.
In this case study, we document the steps the group went through to achieve the goals of the project and identify specific challenges and barriers faced at each stage. We discuss how these barriers may extend to other projects and user groups, particularly those involving non-experts. We conclude by offering recommendations for policymakers to improve the effectiveness of ongoing efforts to expand data use and improve equity in this area along with analyzing the role remote-sensing experts have in the modern ecosystem.
2. Background and assessment of the status quo
Before discussing the case study, it is necessary to define what is meant by satellite data. This term refers to data directly acquired from satellite sensors or derived from such data. We will follow the data level definitions provided by the NASA Earth Observing System Data and Information System (EOSDIS) (SMAP, 2015). Level 1 data consists of raw satellite data, with basic corrections for sensor anomalies and geolocation. Level 2 data involve further processing to derive specific variables or account for atmospheric effects, while Level 3 data is fully processed and aggregated. Level 4 products use satellite data to provide predictions or high-level insights and can even incorporate satellite data with other sources. Currently, most non-remote sensing experts primarily interact with satellite data through level 4 products. A common use case for level 4 products is studies predicting the impact of climate change (Fox et al., Reference Fox, White and McClean2011; Crane-Droesch, Reference Crane-Droesch2018).
Since the 1960s, satellite data has been employed for land cover analysis (Kugler et al., Reference Kugler, Grace and Wrathall2019). In recent decades, the number of satellites has dramatically increased due to a reduction in launch costs by over 30-fold (Tucker and Alewine, Reference Tucker and Alewine2024). This has ushered in a golden age of satellite data, with extensive metrics now available for fields outside of land cover analysis such as nighttime lights, population distributions, and greenhouse gas emissions (Borowitz et al., Reference Borowitz, Zhou, Azelton and Nassar2023; Kugler et al., Reference Kugler, Grace and Wrathall2019; Wästfelt, Reference Wästfelt, Cohn and Mark2005; Xi et al., Reference Xi, Liu, Li, Ding, Zhang, Tarkoma, Li and Hui2023). Notably, social science researchers—who are typically not experts in remote sensing—have widely embraced this surge in satellite data availability with many using satellite data in their research.
Satellite data has become increasingly popular within social science research for several reasons. First is that compared to other data sources, there is a much lower cost associated with using satellite data. For example, mapping land cover through surveys takes decades and can have gaps. NASA’s Landsat satellite data has provided global land coverage data with minimal effort (Pawlitz, Reference Pawlitz2022). Second, the uniformity of satellite data makes work that spans administrative boundaries much simpler. Researchers do not have to worry about standardizing datasets to account for regional differences in data collection. Additionally, satellite data is available at a more granular level than traditional social science metrics (Jain, Reference Jain2020). Third, satellite data offers temporal availability at much greater rates than most ground surveys. Ground surveys typically happen once every several years, while satellite data can be collected daily, allowing researchers to analyze and track temporal trends easily (Jain, Reference Jain2020).
Despite the benefits of satellite data, there are many situations in which researchers intentionally do not use it. Some cases are simply due to satellite data not being the proper fit. For example, satellite imagery is not helpful in determining the income distribution within a household (Yeh et al., Reference Yeh, Perez and Driscoll2020). A task that traditional surveys are much better suited to accomplish, as satellite data struggles to represent complex social or spatial relationships (Wästfelt, Reference Wästfelt, Cohn and Mark2005). Unfortunately, satellite data is still underused in many cases where it could be beneficial. This gap is not unique to social science; it affects all fields. Despite the abundance of satellite data, it is often not easily accessible or user-friendly, leading to its underuse (Hossain, Reference Hossain2015).
Level 1 and Level 2 data products are often freely available, but their use requires significant pre-processing. For users with little knowledge, this task can be overwhelming or deemed not worth the effort (Hall, Reference Hall2010). Consequently, there is a high demand for modeled and derived data products (level 3 and level 4) which are much easier for researchers to analyze (Kugler et al., Reference Kugler, Grace and Wrathall2019). There are some tradeoffs for the simplicity of these advanced products. One is they are specialized so they may not meet the needs of a given research effort. Another is that they can contain unrecognized errors due to issues in the underlying level 1 or level 2 data or flawed assumptions during data processing. Users who lack the knowledge to validate these data sources themselves risk generating misleading results (Jain, Reference Jain2020). Third is that oftentimes, satellite data has to be linked with data from other sources to conduct meaningful analysis (Hall, Reference Hall2010), necessitating some remote sensing skills. Many potential users lack the technical ability needed to overcome these hurdles, contributing to the underutilization of satellite data.
Although there is some awareness that barriers to data use exist, there is little literature that thoroughly documents these challenges or explores the difficulties of gaining the technical expertise required to work with satellite data. This paper aims to address this gap by identifying concrete steps to improve policies aimed at broadening data access and use beyond remote sensing specialists, using our research group’s case study as a lens.
3. Project conception
The research group featured in the case study, led by professors from the Sam Nunn School of International Affairs at Georgia Institute of Technology, focuses on understanding the significance of megaregions worldwide for sustainable development. Megaregions are large, interconnected urban areas composed of multiple cities, often cutting across provincial or national borders. While most of the literature over the last 30 years has used traditional, largely qualitative political science methods to study these entities, the research group was intrigued by the work of Florida, who used satellite data to identify the boundaries of the world’s megaregions (Florida et al., Reference Florida, Gulden and Mellander2008). Collecting consistent data across politically fragmented megaregions is enormously challenging, but satellite data offered a solution for consistent measurements within and across these regions over time. This led the group to employ satellite data products to analyze megaregions.
Initially, the group modeled its work after the study by Florida et al., which used satellite data to define the boundaries of megaregions around the globe (Florida et al., Reference Florida, Gulden and Mellander2008). It is worthwhile to note that other users could encounter even more significant barriers during the project conception phase. The geographic nature of megaregions naturally lends itself to satellite data, but other uses of satellite data may not be so clear. Additionally, the research group was given a clear example showing how satellite data could be applied, specifying a particular type of satellite data and analytical approach. Other users (and the research group as they expand to other research issues in the future) may have added challenges in identifying what satellite data exists that may be relevant to the project and conceptualizing how it could be used. This is especially true for those with novel applications for satellite data.
4. Data selection
The first step for implementing the project was to find and access the night lights data. We were aware from Richard Florida’s work, as well as Dr. Borowitz’s familiarity with Earth-observing spacecraft, that night lights data was available from the VIIRS instrument on the Suomi-NPP and NOAA-20 satellites, and that historical night lights data was available from the OLS instrument on the DMSP satellite series. Despite this awareness, finding and accessing the relevant data provided more challenging than anticipated.
The team discovered that VIIRS data was available from multiple sources in various formats. While the datasets were typically accompanied by technical documentation, the information was rarely presented in a format accessible to non-specialists. The team ultimately opted not to use data provided directly from NASA, but instead a monthly and annual nighttime light composite dataset produced by the Payne Institute for Public Policy at the Colorado School of Mines (Elvidge et al. Reference Elvidge, Baugh, Zhizhin, Hsu and Ghosh2017). As a level 3 data product, this seemed to be the most analysis-ready night lights dataset available.
The team initially planned to examine changes in megaregion size and location over time, by comparing older DMSP data to more recent VIIRS data. However, it became clear that appropriate methods for comparison across these datasets were a source of significant debate in the remote sensing community, and we ultimately abandoned that effort, despite the significant relevance such analysis would have for our research as we felt it was a problem too technically complex for our experience.
It is worth noting that our team had multiple advantages in identifying relevant datasets, as we were building on earlier research and had team members familiar with existing spacecraft and their capabilities. Nevertheless, other users, such as NGO workers or farmers, often struggle to access suitable satellite data due to the disorganized nature of satellite data repositories (Hossain, Reference Hossain2012). Many potential users are unaware of existing repositories, and Google searches rarely yield close to a comprehensive overview of available options. Creating user-friendly documentation written to a non-technical audience would significantly reduce the barriers for non-technical users to find datasets that align with their needs. The documentation would better explain datasets along with allowing datasets to more frequently appear in Google search results.
5. GIS tool selection and data processing
Our research group encountered challenges in deciding how to process the selected data due in part to the lack of detailed methodology in the earlier research we were seeking to replicate (Florida et al., Reference Florida, Gulden and Mellander2008). Attempts to contact the authors were unsuccessful, requiring the group to develop its own methodology. An early challenge was identifying the most appropriate GIS software for use in the analysis. The group found few resources to help beginners consider the trade-offs among systems. Initially, the group selected Google Earth Engine for its user-friendly interface and preloaded data. However, they found Google Earth Engine insufficient in processing power and transitioned to more in-depth analysis. We would like to note that remote sensing has been moving away from QGIS to Google Earth Engine in recent times due to Earth Engine’s parallel processing ability. However, as the research group was performing simple operations on a global scale, QGIS was the better software for our specific use case.
Even though our group had a well-defined goal: use night-lights data to identify large geographic areas with contiguous, high-intensity light—conducting this task took about six months. In addition to time spent evaluating options for GIS tools, we experienced a significant learning curve in using the tool, especially when we transitioned to QGIS. For example, while an experienced GIS user would know that extracting a .gz file from a VIIRS dataset requires using a simple file extractor tool, our team found the process to be a significant hurdle.
This example highlights a broader issue: the lack of tutorials and beginner-friendly resources for inexperienced users. To improve the accessibility of satellite datasets, NASA and other data providers should develop video walkthroughs and tutorials for importing and manipulating data in common GIS platforms. These tutorials would simplify the learning process by breaking down the complex tasks involved in working with satellite data into easy-to-follow steps. As users progress through each step, they can apply the acquired knowledge to datasets beyond those used in the walkthrough, significantly accelerating their ability to analyze their own data sources.
6. Analyzing results
The research group was successfully developed an algorithm that replicated earlier work in delineating megaregions using light-intensity information. However, the group experienced another challenge near the end of the project in interpreting these results. The group had originally been using average light data and at one point re-ran the analysis with median light data, which was also available from the Colorado School of Mines. The results were vastly different, including across key analytical outcomes relevant to the definition of a megaregion—e.g., whether certain main cities were part of a given megaregion or not.
Having such a stark difference in results depending simply on what version of a satellite data product is used, was unexpected by the research group and emphasizes the need for clarity in the appropriate use and interpretation of specific datasets. This example depicts why satellite data providers, whether universities such as the School of Mines or government institutions such as NASA and NOAA, should prioritize offering detailed explanations of what each dataset represents and how it can be used and interpreted correctly. While these explanations are not necessary for those with experience within the field, they are critical for those new to satellite data to make meaningful contributions and avoid misinterpreting data inadvertently.
This is particularly crucial for providers of level 3 and level 4 satellite data, which are often seen as more accessible for non-remote sensing specialists. An improper understanding of the datasets can lead to misleading results. Mears and Wentz demonstrated how even remote sensing experts can misinterpret level 3 and level 4 data, revealing that the incorrect analysis of lower tropospheric temperatures led to an underestimation of the global warming rate by over 25% since 1979 (Mears and Wentz, Reference Mears and Wentz2017), a discrepancy significant enough to influence policy decisions. This clearly underscores the importance of comprehensive documentation for satellite datasets: if experts can misinterpret them, non-experts are far more likely to do so.
7. Discussion
The experience of our research group repeatedly demonstrated that actions that appeared easy on the surface often had hidden barriers. For our team, identifying the relevance of satellite data to our research project was straightforward based on previous research, but for others, making these connections may be significantly more difficult. Once a project is conceptualized, accessing data can be a challenge. NASA datasets are free and available online but determining which version of a dataset should be used for a given project is complex, with little written guidance aimed at non-specialists. When it comes to analyzing data, there are multiple easily accessible, no-cost GIS tools, but there is little guidance for beginners on how to choose the right one. While there are plenty of tutorials available, few target the highly specific needs of individual users, making them less helpful. Finally, correctly interpreting the results of the analysis can be surprisingly fraught and may depend in non-intuitive ways on the specific dataset selected. Once again, documentation on this issue aimed at a non-specialist is lacking.
For remote-sensing or satellite data experts, many of the points raised in this commentary might seem trivial, and that is part of the challenge. Many of the remote sensing specialists who develop these datasets may not be aware of the questions and challenges faced by beginners attempting to access and use data. Some might note that these challenges can be solved with “just” 3–6 months of training—for example, taking a GIS analysis course. However, the need for multiple months of training is a significant barrier to the use of this data by diverse sets of users. While such a solution may be practical for a university research group like the one featured in this case study, it may be far less practical for other users. For example, journalists using VIIRS data to identify the hardest-hit areas in Puerto Rico post-hurricane Fiona (Hernández et al., Reference Hernández, Tran and Moriarty2023) would have found undergoing extensive training too large a barrier. Instead, we would suggest that NASA and other institutions interested in making data truly accessible to a wider audience invest in documentation aimed directly at new, non-specialist users for data discovery, analysis, and interpretation in a timely manner.
Another productive way forward is to increase opportunities for engagement between experts and new, diverse users. Access to experts for specific questions and assistance can be invaluable. Our team benefited greatly from the expertise available through NASA SEDAC and our own university. Our team was aware of these resources due to personal experiences (Dr. Borowitz was a member of the SEDAC users working group). NASA and similar organizations could enhance these efforts by actively promoting these resources and expanding opportunities for engagement. This could involve outreach to social scientists and other non-traditional satellite data users to raise awareness of available support. If organizations providing datasets prioritize communication with users, this could further help with data selection, analysis, and interpretation. Utilizing social media advertising and other non-traditional channels could effectively help NASA achieve this end. Reddit is an especially useful tool that many go to as a source of advice.
The NASA Earthdata forum is a good attempt at connecting inexperienced users with experts. The forum functions much like any forum. It is meant to be a community for experts and others to discuss satellite data. However, a quick review of the forum shows that many user queries go unanswered, leaving users with the guidance they need. Yet another example of how the status quo does little to help newcomers even if the framework exists. To build on this, NASA could invest more effort in increasing activity on the forum. There are two key actions we would recommend NASA to take. The first step would be for NASA to have its own staff regularly respond to others’ posts on the forum. In addition, starting the practice for each dataset release to have a dedicated post on the forum. These posts would serve as a centralized hub for discussion around a dataset, allowing for more knowledge transfer to occur, as users could see what others are saying/doing with datasets. Such a system would allow for the dataset documentation to improve from the benefit of collective knowledge even if the initial documentation is lacking. Over time, this collaborative approach would naturally enhance the dataset’s usability, as community-driven feedback and shared expertise fill in gaps in the original documentation.
8. Conclusion
The case study of this Georgia Tech research group underscores the substantial barriers non-experts face when utilizing satellite data, despite the increasing availability of such data through open-access policies. While these policies represent a significant step forward, they alone do not address the technical challenges and accessibility issues that limit the effective use of satellite data by non-specialists.
To bridge this gap, it is essential to enhance the accessibility of satellite data through several key measures. First, improving documentation and providing detailed, user-friendly guides will help demystify complex data discovery, processing, and analysis tasks. Second, creating comprehensive walkthroughs and tutorials can ease the learning curve for newcomers, making advanced tools and techniques more approachable. Third, fostering direct engagement between remote sensing experts and non-specialists through mentorship and outreach is crucial. This collaboration can facilitate knowledge transfer and support, helping to overcome the learning barriers faced by those new to satellite data while empowering remote-sensing experts to better design data products for non-experts. A straightforward way to do this would be to utilize the existing infrastructure of the NASA Earthdata forum to make it a centralized hub for collaboration surrounding satellite data.
By addressing these challenges, policymakers and data providers can significantly enhance the utility and equity of satellite data, empowering a broader range of users to leverage this valuable resource for research and decision-making. Ensuring that satellite data is not only accessible but also usable is a critical step toward realizing its full potential across diverse fields and applications.
Data availability statement
The data supporting the findings of this study are openly available in the Annual VNL V2 at https://eogdata.mines.edu/products/vnl (Elvidge et al., Reference Elvidge, Baugh, Zhizhin, Hsu and Ghosh2017).
Acknowledgments
The authors thank NASA’s SEDAC and Ramachandra Sivakumar for their technical support to the research group during the project.
Author contribution
Formal Analysis, Investigation: H.G, G.R, X.L, R.P, G.C.; Writing: H.G. (Lead), M.B, G.R, X.L, R.P, G.C, B.W; Data Curation: G.R, H.G, X.L, R.P; Conceptualization, Resources: M.B, Project Administration, Supervision: M.B (Lead), B.W; Methodology: H.G, G.R., X.L, M.B; Software, Visualization: G.R, H.G, X.L; and Validation: H.G, G.R, X.L, R.P, G.C, M.B.
Funding statement
This research received no external funding.
Competing interest
The authors declare none.
Open access
Comments
No Comments have been published for this article.