Novel Observing Strategies Testbed: A Lightweight, Collaborative Platform for Prototyping Earth Observing Missions

Earth observation has long relied on single satellites or fixed constellations to watch the planet, but a new approach envisions a more fluid way of watching. Instead of a rigid schedule, scientists imagine networks of satellites, ground sensors, and computer models that talk to one another in real time, deciding together where and when to look. This concept, known as a distributed space mission, requires different pieces of software and hardware to coordinate perfectly, even when they are owned by different organizations and running on different computers. The challenge is building a testing ground for these ideas that is robust enough to handle real-world failures but simple enough for researchers to use without getting bogged down in complex engineering. If these systems cannot be tested effectively before launch, the risk of failure in space increases, and the potential to improve our understanding of Earth's changing climate diminishes.

A team of researchers from Arizona State University, the University of Maryland, and NASA's Goddard Space Flight Center has developed and tested a new platform designed to solve this problem. They call it the Novel Observing Strategies Testbed. Think of it as a digital playground where scientists can build and break complex satellite coordination systems before they ever leave the ground. The platform started as a basic prototype, but after running it for a while, the team realized it lacked the reliability needed for long-term, multi-institution work. They rebuilt it in two major stages, adding features to handle user identities, ensure messages are never lost, and keep the simulation running smoothly even when connections drop. The final version, tested over two separate campaigns spanning nearly 600 days of simulated time, proved that this lightweight system could coordinate complex tasks between a satellite simulator and a massive land-surface model running on a supercomputer, all while keeping time perfectly synchronized.

The core of their work involved testing how well this new platform could manage a specific scientific problem: measuring snow water equivalent, which is the amount of liquid water contained in a snowpack. This measurement is difficult because snow varies wildly in depth and density, and no single sensor can see it all. The researchers set up a scenario where a computer model of the land, running on a powerful supercomputer at NASA, worked alongside four different software applications that simulated a constellation of small satellites. These satellites were designed to use radar to see through clouds and measure snow depth. The goal was to see if the satellites could intelligently decide which parts of the snowpack to scan, based on what the land model was telling them about changing conditions.

To make this work, the researchers had to overcome five specific hurdles that had plagued their earlier version of the testbed. First, they needed a secure way for users from different universities and NASA to log in without sharing passwords. They implemented a system where users get a temporary digital pass that refreshes automatically, ensuring that only authorized people could access the simulation. Second, they needed to guarantee that no data would disappear if the internet flickered. They replaced a "best-effort" messaging system with one that stores every message in a queue until it is successfully delivered, ensuring that even if a connection breaks, no information is lost. Third, they built in automatic recovery so that if a computer lost its connection, it would reconnect and pick up exactly where it left off without a human needing to restart it. Fourth, they created a strict set of rules for how data is formatted, so that different programs speaking to each other would never misunderstand the information being sent. Finally, and perhaps most importantly, they added a master clock that could pause the entire simulation instantly, wait for the supercomputer to finish a calculation, and then resume everything at the exact same moment.

The team ran two major tests to prove the system worked. The first test covered a full snow season from October 2024 to June 2025. In this version, the system paused the simulation for a fixed amount of time to let the supercomputer finish its work. While this worked, it sometimes left the satellite simulators waiting idly if the supercomputer finished early, or it forced them to wait too long if the supercomputer was slow. The second test, covering the 2018–2019 snow season, improved on this by making the system "event-driven." Instead of waiting for a timer, the simulation paused indefinitely and only resumed the moment the supercomputer signaled that its data was ready. This change eliminated almost all the wasted waiting time, making the entire process nearly 100% more efficient.

Throughout these long tests, the system performed with remarkable stability. Over the course of the two campaigns, which together simulated 577 days of snow seasons, the platform experienced dozens of network disconnections. In every single instance, the software automatically reconnected, recovered its place in the simulation, and continued without any data loss. The coordination between the different parts of the system was incredibly fast; when the central manager told the other applications to pause, they all froze in less than a second. This speed is crucial because it means the different parts of the system are always looking at the same moment in time, preventing errors that could arise if one part was looking at yesterday's snow while another was looking at today's.

The results showed that the platform successfully coordinated a complex workflow where a land model and satellite simulators exchanged data to select the best places to observe snow. The system identified high-value areas to scan, such as deep mountain snowpacks where other sensors struggle, and adjusted its strategy as the snow melted and shifted across the landscape. The researchers found that the system could handle the heavy computational load of the land model while keeping the satellite simulations running smoothly, even when the two were communicating across different networks and firewalls. This demonstrated that it is possible to link a high-performance supercomputer, which is often locked down for security, with a flexible, cloud-based simulation network without compromising safety or speed.

This work represents a significant step forward in how space missions are planned and tested. By proving that a lightweight, collaborative platform can handle the rigors of long-duration, multi-institution simulations, the researchers have provided a tool that can be used to evaluate new observing strategies before committing to expensive development. The system is not just a theoretical concept; it has been validated in a realistic environment, reaching a level of maturity where it can be trusted to support real mission planning. The ability to pause, resume, and synchronize complex simulations across different organizations opens the door to more adaptive and efficient ways of watching Earth, ensuring that future satellites can respond to changing conditions with the precision needed to understand our planet's most critical processes.