Water is an essential resource for all living organisms on Earth. About 71% of the Earth’s surface is water-covered – and about 96.5% of this water is in the oceans! Surface freshwater, which is available in rivers, floodplains, ponds, lakes, and reservoirs, accounts for less than 0.5% of this resource. Human impacts can endanger these freshwater ecosystems, and water management programs are progressively needed to track changes, understand patterns, and develop public policy. Fieldwork campaigns for assessing and monitoring water quality are expensive, logistically challenging, and time-consuming. Therefore, comprehensive and routine surface water monitoring using traditional field surveys is largely unfeasible. Furthermore, it cannot adequately resolve spatial variability in larger systems.
How can remotely sensed data help monitoring water resources?
Satellite data have been used for studying and monitoring Earth resources routinely since the launch of the first Landsat satellite in 1972. Over aquatic ecosystems, remotely sensed data have spurred the development of algorithms for mapping water quality attributes (e.g., chlorophyll a, total suspended solids, absorption by colored dissolved organic matter, and Secchi disk depth) from small inland waters to the open ocean. These large-scale and repeating observations have revolutionized our understanding of surface waters (Figure 1) and are now central to assessments of aquatic biodiversity, climate change impacts, trophic states, and primary productivity.
Figure 1. Water bodies from space showing variability in water color. (A) Taihu Lake, China; (B) Amazon River, Brazil; (C) Chesapeake Bay, United States; (D) Billings Reservoir, Brazil. Sentinel-2/MSI true-color composite obtained from EO Browser (https://apps.sentinel-hub.com/eo-browser/)
However, remote sensing algorithms demand in situ datasets for calibration and validation. The collection of highly trustworthy observations, which are now becoming known as fiducial reference measurements, is a task requiring special instruments and strict adherence to measurement protocols as seen in the video below. In addition, extensive field surveys are needed to cover previously under-sampled water bodies in remote and sometimes dangerous areas, experiencing extreme weather conditions. Think about the difficulties of collecting data across an Amazon rainforest floodplain lake, located a 30-minute walk inside the dense forest, from your canoe! Or about the logistical challenges of planning surveys in coastal waters and the open ocean, from the Peruvian coast to the Bornean coast and the South China Sea! What about conducting fieldwork in the large rivers of China or in cyanobacteria dominated reservoirs in Brazil? Many previously unsampled water bodies are in developing countries where limited funding is available to create and develop methods to analyze, monitor, and understand their water resources.
Then the question is: how can we address these data needs globally?
One of the answers is the GLORIA initiative. The GLObal Reflectance community dataset for Imaging and optical sensing of Aquatic environments (GLORIA) is an international effort that includes 7,572 curated hyperspectral remote sensing reflectance (Rrs, sr-1) measurements, with co-located measurements of limnological and bio-optical parameters, such as chlorophyll a, total suspended solids, absorption by colored dissolved organic matter, and Secchi disk depth (Figure 2). This open-access dataset is an attractive resource for the aquatic remote sensing community as it is grouping efforts of a large number of institutions and providing data for 450 different water bodies worldwide with a comprehensive collection of optical water types.
Figure 2 Sampling locations available in GLORIA.
GLORIA contains observations from three decades of field surveys by individual researchers and teams. Many of us have been motivated to travel to almost inaccessible lakes because we have seen their odd colors on satellite images (Figure 3). This is exemplified in the Amazon floodplain forests, which are traversed with creeks and in which small boats are needed for portaging and accessing such lakes. Often, our field surveys resulted in sleepless nights taking care of experiments, organizing and packing the equipment, airplane travel with suspicious looking items of baggage, sickness from drinking unpalatable water, multiple days of laboratory analysis after (and before) data collection, and much more. Experimental challenges included the adaptation of ocean color protocols for use in waters of optical complexity not observed before, and the necessity of developing new strategies for setting up the equipment in relation to the location visited and the size (and type) of the boat used. Nevertheless, the people we met, places we saw, and benefits to society and the environment rewarded us. In the Amazon floodplain, the local people taught us how to live, eat, survive, and treasure their lifestyles. The teammates, scientists, and students we met on our field surveys shared their visions and insights, and we became friends. The places we saw on these field surveys were unique and the beauty of watching the sunset from a boat was priceless. Last, but not least, our water quality monitoring efforts helped to ensure clean water to the local people and communities, bringing benefits to society.
Figure 3 – Examples of freshwater field campaigns.
Field survey insights: Bio-optical lake survey in Aotearoa New Zealand
Aotearoa New Zealand has over 3800 lakes greater than one hectare in size and countless smaller ones. A very wide range of lake types and bio-geo-optical conditions can be found in a relatively small and accessible country due to its diverse geology, topography, and its latitudinal extend. To sample this diversity, and to make important satellite matchup measurements in the under-sampled southern hemisphere, Moritz Lehmann (Xerra Earth Observation Institute & the University of Waikato), led a field program to characterize as many New Zealand lakes as possible.
Between 2017 and 2021, Moritz and his team made 194 measurements of remote sensing reflectance and determined water quality in 89 lakes. The team’s home base at the University of Waikato in Hamilton in the central North Island provided a central location from which to reach many lakes in a day’s drive. More extensive logistics and equipment, including water filtration rigs and liquid-nitrogen sample storage, were needed for multi-day field trips, such as those to sample the great lakes on New Zealand’s South Island (Figure 4).
Figure 4: Example of fieldwork conducted in New Zealand. Figures A-B: Multi-day field trips require makeshift lab setups in rented accommodation and portable freeze storage. Figure C: Satlantic HyperOCR radiometer in a skylight-blocked setup under perfect conditions on Lake Ohau, Canterbury. Figures D-E: Water color determination using Munsell color plates (D) and Secchi disk depth measurement (E) in an oligotrophic lake.
This video below provides an impression of a perfect day on Lake Monowai, an oligotrophic glacial lake in the south of New Zealand’s South Island. Moritz was accompanied by Mortimer Werther, PhD student at the University of Stirling, Scotland, and Dishan De Silva, MSc student at the University of Waikato, who produced the video.
Over four years, the scientists traveled a distance of 30,000 km to reach coastal dune lakes in the subtropical north, glacial lakes in the temperate south, and riverine, volcanic, and peat lakes all over the country. About 300 hours were spent on watercraft ranging from canoes to high-powered motorboats, and a similar time was needed to complete sample analyses in the lab at the University of Waikato.
The making of GLORIA initiative
The only way to build a large, comprehensive, and difficult-to-collect dataset was to reach out to the global bio-optical research community. In 2018, Nima Pahlevan (NASA Goddard Space Flight Center & Science Systems and Applications, Inc.) was invited to co-lead a global round-robin initiative for evaluating various satellite data processing schemes over inland and coastal waters. For a thorough global assessment, a sizeable in situ dataset was deemed inevitable; hence, Nima began the data drive by making several community announcements and conducting outreach activities. The dataset expanded over the following years and was used to develop global algorithms for estimating water quality attributes, such as the concentration of chlorophyll a. As the value of the dataset was proven via highly cited publications, the desire to make it openly available to the global scientific community led to the establishment of the core team to drive its publication in October 2021. Moritz Lehmann, Daniela Gurlin (Wisconsin Department of Natural Resources) and Nima began requesting permissions for open-access publication from existing data contributors and reached out to more scientists. The final data-gathering effort took place at the Ocean Optics XXV conference in October 2022 in Qui Nhon, Vietnam, and resulted in unusual optical water types from fjords and dark, almost tea-stained lakes.
Our work has shown that the only way to build a large, comprehensive, difficult-to-collect dataset of existing measurements was to reach out to the global bio-optical research community. This means that GLORIA is a grass-roots, collaborative effort in which we were rewarded with an unmatched diversity of environments and water types (Figure 5) encompassing the clear waters of fjords (Norway), eutrophic waters of Lake Taihu (China) and Ibitinga Reservoir (Brazil), turbid waters of Amazon floodplain lakes (Brazil), and black waters of Lake Pyszne (Poland). Field surveys in these locations provided unique Rrs data taken in sometimes challenging conditions, which increases our dataset's importance and demonstrates the invaluable contributions of the participating researchers through sharing their data.
Figure 5: Locations of some sampling sites of the dataset.
The necessity of a globally representative dataset for the development of global water quality algorithms demands tenacity and endurance in data collection. We hope that our example will inspire other research groups to assemble comparable datasets and make them open access as soon as possible. That said, we recognize and thank all who contributed in some way to GLORIA: locals and guides, field personnel, lab technicians, undergraduate and graduate students, captains and skippers, and all other supporting personnel. We believe this dataset can be used to produce new and more robust algorithms for water quality monitoring and improving environmental protection on a global scale.
Further information about the GLORIA dataset can be found in in our data descriptor in Scientific Data (https://doi.org/10.1038/s41597-023-01973-y). The GLORIA dataset can be accessed at PANGAEA (https://doi.org/10.1594/PANGAEA.948492).