top of page

Real Ice: Field Test Report

Nome, Alaska – 21 – 27 February 2023


This was the first field trip for Real Ice, closing a first iteration of scientific review, data analysis, definition of re-icing approach, detailed design and prototyping of a device, followed by testing in target conditions. As such it was a crucial 0-to-1 moment for the company, moving us to a new phase where we took several scientific models and engineering principles to actual practice and demonstration.

It was an eye opener in many dimensions, in some cases confirming our hypotheses, in others adding many practical learnings and suggesting many possible improvements to our approach.

Most of the primary objectives of the trip were met, despite difficult weather conditions in the initial 2-3 days, and we laid the foundation to move to the second iteration, which would involve a more detailed definition of the approach and objectives, ideally supported by scientific modelling, the design and prototyping of a second device (hydrogen fuelled) and testing during the 2023-24 winter, with much broader objectives (area coverage, duration of water pumping, snow making).

As a final point, from a human perspective, this trip confirmed that we can find broad support from local populations for the ambitions of our enterprise, and the positive message we are trying to pass regarding the importance to preserve and restore the Arctic sea-ice seemed well received in most cases.

Large area flooded after 1 hour of pumping, and starting to freeze, on Day 1


The following table lists the primary and secondary objectives of the Nome Field Test. The results are detailed in the next section.

Research Papers

Our approach and ice-making methods were primarily inspired by the following scientific papers:


The “flooding water pump” was tested successfully: when properly immersed with good space under and around the strainer, it delivered the expected flow, and the duration of the battery charge was as expected, around 1 hour per full battery (700 Wh). We only witnessed minimal reduction in energy capacity for the batteries in cold weather, and in realistic use.

Pump with directional pipe extracted after flooding an area.

The directional output pipe was the most efficient one, pushing the water at a reasonable distance from the ice hole, and guaranteeing limited return of water into the hole. Additionally, the direction of the flow didn’t affect the shape of the flooded area, with flooding tending to be driven more by the surface profile and snow depth and tending to form a circular or slightly elliptical shape.

Directional pipe with pump anchored to a support (outside of picture) and standing about 50cm above the ice level.

The 360’ delivery instead had insufficient pressure in any given direction, and the water tended to stagnate around the ice hole, and most of it would flow back in. The 360’ pipe also seemed to force the pump to use more energy, possibly because of a more restricted output, having to push on a top plate. The 360’ pipe was tested both sitting on an inflatable support or directly on a circular board. In both cases the water, pushed only 20-30 cm away from the pipe, seemed to dig at the top of the ice hole, creating a large gap under the support, and allowing a significant amount of water to flow back into the hole.

360’ pipe with circular board sitting directly on the ice.

We were able to flood significant snow thickness (mostly in the 10-20 cm range, with some deeper areas) and witnessed the transformation of the snow cover to icy slush, and then bare ice.

Clearly this was done in the “wrong” season (end of winter, rather than beginning) so there was no attempt to measure the long-term impact of exposing bare ice, and the reduction of albedo from snow to bare ice was expected and clearly not optimal at the end of winter.

Also, this field test didn’t include any large extent re-icing, and all tests were limited to 1-3 hours per session, covering up to 700m2 and snow flooding for thickness of up to 20cm.

Bare ice forming, after flooding an area.

The snow making tests were unsuccessful: the pump had too little power to displace water high enough in the air to create snow, even at very low temperatures. It also seemed to struggle with ice formation into the hole, which seemed to limit the output pressure and flow.

Areas of Improvement

Flooding the snow was also achieved successfully, and provided plenty of practical feedback, which can be used to improve the effectiveness of this method:

  1. While the pump continued to displace water and flood the snow, no observable freezing was observed, even after durations of 2-3 hours, and temperatures below -20C. Freezing would only happen when the water stopped flowing, and depending on the temperature, it would be fairly quick: superficial freezing will only take 5 minutes at around -20C, on a large pool (200-300 m2) with water depth of 5-15cm, depending on the existing snow.

    • This is the most critical issue: we could think of flooding for a long time (potentially days) and then freezing the whole area once the pump moves to another area.

    • Or we could alternate pumping and freezing (no water flow) but we would need to solve the problem of the pump and potentially the ice hole freezing when idle, which would require generating heat and/or water movement while idle.

  2. By only allowing superficial freezing (keeping the pump idle for, say, 15 minutes) when the flooding restarted, the new flow of water would completely melt the superficial ice. Sufficient freezing (several hours, depending on ambient temperatures) should be allowed before flowing more water on the same area.

  3. The pump will need a proper cofferdam to prevent water from flowing back into the ice hole. This could be achieved with a second section of pipe (e.g., metallic cylinder with 40cm diameter) that is placed just outside the rim of the ice hole. It is important that this is not inside the hole, as the water will make its way between the ice and the cofferdam.

  4. The ice hole should be optimised for the size of the pump (diameter of around 5cm larger than the pump itself) leaving enough space for the water around the pump, but not too much to require needless effort/energy to dig it, and much larger cofferdam. During the tests our ice auger was too small to create a single hole for the pump, so we had to create 3-4 holes near each other, which ended up creating a much larger hole, and sometimes leave ice formations that would impede the full pump flow.

In future versions of the device, especially when moving to AUV, the pump may not be positioned into the ice layer, but below, in which case we would be constrained by the (typically smaller) output pipe diameter.

Other Learnings

  1. The snow thickness is a crucial variable for the result of the re-icing exercise. The freezing of the snow resulting from the flooding will be completely different when we saturate the full depth of the snow (resulting in bare ice within minutes) vs. when we partially saturate it. We witnessed cases where the remaining 5-10 cm of snow top seemed enough to keep the underlying flooded water completely liquid after 15 hours (mostly overnight) at –10 … -17C temperatures.

  2. Because the snow thickness is highly variable, even in areas that are relatively flat (the wind will create small “dunes” of more or less powdery snow) the complete flooding will take a different time, even at the same distance from the pump. The water will continue to travel at the bottom of the snow layer, as the ice tends to be flatter, regardless of the snow dunes, or different top levels of the snow cover.

  3. Because of the lack of freezing during the pumping of water, basically the speed of the water through and under the snow was very similar, in very different temperatures and weather conditions: we tested in temperatures just below the freezing point, with or without wind, and in temperatures as low as -27C. In all cases the water would expand based on the thickness of snow (see other learnings below) rather than on temperature, and essentially it would continue to expand at the same “area-speed” (as expected, given the constant water flow)

  4. The water seemed to travel both horizontally and vertically, driven by the snow thickness: while the water travelled horizontally at the bottom of the snow stack, on top of the ice level, it was also visually travelled vertically in areas closer to the pump first, until the snow was fully saturated in those areas, to then expanding in both directions further away from the pump. The water seemed to move vertically even on higher dunes, moving to a level higher than the prevailing level, as if the snow was absorbing it like a sponge.

  5. The horizontal movement of the water was observed by simply stepping on the snow around the visible flood pool or removing the snow in a small canal away from the pump and the fully flooded area: even a minimal quantity of water saturating the bottom layer of the snow would make it unstable and collapse when applying the weight, often forcing the top snow into the underlying water, and exposing the latter. We had the impression that the quantity of the underlying water was sufficient to flood the full stack of snow, but for some reason it would take longer for the process to occur. But it would flood all the stack once the top part of the snow was collapsed into it (e.g., by stepping or pressing on the snow)

  6. The harsh conditions of this environment call for devices with large enough power to guarantee some protection against ice formation: a small pump would much more quickly and easily clog and block due to freezing. There should be a lower limit in power to our evaluation of power/flow/weight/cost efficiency trade-offs.

Small canal dug a few metres away from the flooded area, to show the water travelling under the snow.

Areas That Need Clarification

  1. The bare ice created at the top of existing ice through snow flooding, in some cases seemed very brittle. While there is no trace of liquid water or slush under it, through to the underlying layer of more solid ice, even when resulting from the flooding of 20+ cm of snow, the structure seemed much weaker, possibly with a considerable amount of air, and, for instance, walking on it would result in making it collapse. This may affect the conductivity of this layer (Note: Cecilia Bitz indicated that this could be due to the speed of freezing).

  2. Once an ice hole was drilled, and once we let the water freeze, after flooding the snow completely, we could observe significant water vapour coming from the water. We attribute this to the freezing process taking place, and the latent heat being released into the atmosphere because of this.

Weather summary for the week of field tests.

Summary of Tests Executed

Common Preparation and Test Execution

The pump, output pipe, battery station, ice auger, a shovel and several measurement tools were transported by pickup truck first, and then by sled to the chosen destination. The first tests were conducted near the jetty in the Nome harbour area. Subsequent tests were conducted near Fort Davis, about 3 miles from Nome. In both cases we positioned the pump at 100 – 200m from the road, changing position every day, to cover a different part of the sea-ice, and continue to observe the evolution of status from the previous days.

Carrying some equipment to the target location near Fort Davis.

All tests required the creation of a hole in the ice to accommodate the submersible pump. This was done using a 4” diameter ice auger. As the pump required at least a 12” diameter hole for clearance, and as the ice was between 70-100 cm thick, we ended up requiring 3-4 holes close to each other, and creating a much larger hole, sometimes of irregular shape. This complicated the protection of the hole from water in-flow and sometimes impeded the pump, reducing its output flow.

Ice hole with makeshift wooden planks cofferdam.

The tests were executed with durations from 15 minutes to 3 hours, replacing the battery with a fully charged one roughly after every hour of use.

21st of February 2023 - Day 1

Tests were conducted near the harbour jetty, using a directional pipe. After 30 minutes of pumping, the area covered was of elliptical shape, with 19m and 15m axis. We remeasured on a regular basis, until the end of the test. We noticed that the water had travelled outside of the fully flooded area for at least 1-2m, and the water was clearly present at the bottom of the snow.

Temperature: -4C, -3C

Time elapsed: 1:02:30

Snow depth: 10cm average - strong winds were moving the powdery snow around quickly

Area covered: approximately 500 m2 (elliptical, 30m and 22m axis) of completely flooded snow

Wind speed: 20-40 mph

Day 1 - blizzard conditions made the testing complicated.

Day 2 – Day 3: no testing due to surge storm

Temperatures were hovering around the freezing point; high winds of up to 50mph and a large snowfall limited visibility and made it very difficult to move. A surge storm lifted the fast ice by around 1.5 metres, and broke it from the landlocked ice, creating a fracture and letting vast quantities of sea water enter on top of the ice along the coast. The wind pressure continued to push the broken shelf onto the landlocked ice. It took over two days to see improved weather, after which most of the sea water froze, the fast ice reconnected to the land, while massive “scars”, breaking lines and gaps were left on the ice.

Day 2 - storm surge resulting in breakage of fast ice, making it inaccessible.

24th of February 2023 - Day 4

After the storm we could finally go back to the sea ice. We chose a location near Fort Davis, as the previous one near the harbour was still “scarred” by the surge storm.

A thicker layer of snow was present on top of the refrozen fast ice shelf, and we were very careful to cross from the land supported ice to the sea supported area. We decided to test the 360’ delivery pipe first, which wasn’t successful. We tried with both inflatable support, and without it, but a large quantity of water was flowing back into the ice hole, as the delivery wasn’t pushing the water far enough, and this type of pipe made it difficult to protect the ice hole with ice/slushy snow. Essentially the water was not spreading very far, and it caused the pump to dig deeper into the ice and a lot of the water to be lost back into the sea.

Day 4 - displacement of water using the 360’ pipe.

After fully using the first battery (which lasted a shorter 55 minutes) we observed that we only managed to cover around 100 m2, and then switched back to the directional pipe from the same ice hole.

Because of the hole now being dug into the superficial ice by a few centimetres, we decided to create a cofferdam using some wooden planks we had available.

We pumped for the following 2 hours using 2 batteries. This test confirmed that the directional pump was more efficient, only losing a small amount of water back into the sea (less than 10%, in our estimation) and covering a much larger area. The snow was at least 10 cm thick, with areas of small dunes where it was up to 20 cm thick.

The approximate area flooded was a circle with a 15m radius (approximately 700 m2) with around ⅔ of the area fully flooded and exposed, and the rest (with deeper snow, or further away from the pump) flooded in the bottom part of the snow, where the top part was now very weak, and just stepping on it would reveal the water underneath.

Temperature: -7C, -10C

Time Elapsed: approximately 3 hours (55 minutes with 360’ pipe, 2 hours with directional pipe)

Snow depth: 10 - 20 cm

Area Covered: approximately 700 m2, partially flooded on the deeper snow

Wind speed: 10 mph

Day 4 - after the pump with its directional pipe was finally extracted, and the wooden planks cofferdam unmounted.

Day 4 - observing the freezing of the icy pool.

25th of February 2023 - Day 5

The main objective of the day was to observe the freezing of the sea water after pumping. We ran the pump for 30 minutes, then stopped it for 15 minutes, and finally ran it for another 30 minutes to finish using the battery.

During the pause, with temperatures near minus 20 Degrees C, the water was freezing very quickly at the surface. However, with fairly deep water (10cm or more, fully flooding the snow layer) only superficial freezing was observed. When the pump was restarted, the frozen water was completely melted again, to clearly indicate that we would need much longer to create ice that would support water flowing on top of it.

Day 5 - pumping water and wooden planks cofferdam.

The pump was extracted and the tests resumed after about 4 hours, in the afternoon, using the same ice hole. This time the freezing of the previous pond was deeper, and the water was flowing on top of it. We did observe and hear the cracking sound of the underlying ice, but the water didn’t melt it completely and continued to spread wider than the previous pond.

We waited for some more freezing to happen, before we dismantled the pump, which was now also showing significant freezing around the output pipe.

Temperature: -15C, -27C (getting colder in the afternoon)

Time Elapsed: approximately 2 hours (1 hour in the morning, 1 hour in the afternoon)

Snow depth: 10 - 20 cm

Area Covered: approximately 500 m2, including an area of snow not fully flooded/saturated

Wind speed: 0 - 20 mph

6th of February 2023 - Day 6

Before starting the daily operations, we observed the results from the 2 previous days:

  • The ice area created on day 4 was very solid, partially covered by a dusting of snow, but very robust and felt like the ice at the bottom of the snow, just less even on the surface.

  • The ice area created on day 5 was instead fully frozen, but much more brittle. The ice appeared to embed much more air, and would crumble when we stepped on top of it.

  • In the areas not fully iced (from day 5) where the water only partially flooded the snow (at the bottom of the stack) the water appeared to be still liquid and not frozen at all, despite the very low temperatures overnight. The thick layer of snow (up to 20 cm) seemed to have prevented it (or most of it) from freezing

Today we wanted to do some more observations about the freezing times of sea water right after pumping, and also test a smaller pump for higher pressure water displacement, potentially useful for snowmaking.

We used the usual set up with a directional pipe for 30 minutes before stopping for 15 minutes, and then continuing for another 30 minutes.

As the day before, we observed very fast freezing (visually advancing across the flooded pond) of the water surface, but then this would be completely melted when we restarted the pump for the following 20 minutes.

At this stage we extracted the pump to test the smaller pump. We immersed it in the same hole, but unfortunately we didn’t get any useful water displacement. Its power was too low and the output pipe was too narrow to fight the water freezing, and deliver any significant water at the pressure expected. We briefly tested a larger hose with the smaller pump and observed the limited flooding flow.

We spent the remaining time on the sea ice to observe the freezing of the water displaced with the larger pump.

Day 6 - water pool starting to freeze and peripheral areas partially flooded (footstep on the right showing snow depth, and water flooded at the bottom of snow stack).

Temperature: -20C, -27C

Time Elapsed: approximately 1 hour (50 minutes with larger pump, 10 minutes with small pump)

Snow depth: 10 - 20 cm

Area Covered: approximately 400 m2, including an area of snow not fully flooded/saturated

No wind

Indications for the Next Iterations

The next iterations will include:

  • A hydrogen fuel cell power station to supply the electric submersible pump, with the aim to flood large areas through extended flooding times.

  • A snowmaking pump with appropriate power should also be tested.

The changes in the method of re-icing should include:

  • Test optimal flooding time vs. freezing time to allow the freeze to happen. Ideally, we would want to flood an area “completely” before we move to a different area and allow the first one to freeze properly. But this may be impractical in the real case, where we would only want to move a pump every 1-2 weeks. Pumps may need to be relocated much more frequently (e.g., daily) if we don’t want to keep them idle for too long.

  • Find a way to protect the pump from freezing, while we wait for the ice to form on the flooded snow. The pump used in the first iteration was self-draining, but this won’t be sufficient as the water in the ice hole will start to freeze once the pump stops. This could be by placing an insulating material over the drilled hole.

  • Use the proper size of ice auger and properly sized and engineered cofferdam/insulator to protect from backflow of water into the ice hole.

  • Find ways to measure the vertical and horizontal movement of water in snow, to ascertain the risk of leaving large areas partially flooded, with water that will not easily freeze under snow.

  • Find ways to measure the structure and the conductivity of the bare ice created by flooding the snow, as it appeared much more brittle and less dense than normally grown sea ice.

  • Making snow at the end of winter will still require low enough temperatures to turn water aerosol into snow (lower than - 2C). This will also require a special pump and delivery nozzle system (to include nucleation and fine mist, plus air), not just high power/pressure.

Thank you for reading, please follow our website and social media for the latest updates at Real ice, We would also like to thank the people & town of nome, Alaska for their support during the test. We are excited to begin our next phase of testing which will see us visit two location with our hydrogen powered pump. Stay tuned!

89 views0 comments

Recent Posts

See All


bottom of page