Solar Pump for River & Canal Irrigation: Sizing a System for a Moving Water Level

The short answer: pumping from a river or a canal is not a well job with a different pipe. A borehole gives you three things a river never will: a water level that stays put, water that is already filtered, and a known distance to the water. A river gives you none of them. The level moves, by about 10 m a year on the Amazon at Manaus, and the Mekong at Stung Treng runs from roughly 2.4 m in the dry season to peaks above 12 m in September. The water carries silt, sand and debris. And the intake is often 100 to 500 m from the field you are trying to water.

So size the system at the lowest stage you will ever see, not at today’s level, and treat the delivery line as part of the pump rather than as plumbing. Two numbers still decide your model, exactly as they do for a well: the total dynamic head at that low-water line, and the flow your crop needs divided by your effective pumping hours. The difference is that on a river, the first number changes by metres while you are not looking.

Three constraints a borehole never gives you

Every river system that fails in its second season fails on one of these three, and none of them appears in a well design.

Read the first row again, because it is the one that changes the whole architecture. In a borehole, the pump goes down the hole and the water level can drop 20 m without anyone thinking about it. On a river, the water level itself moves, and your pump is standing next to it.

How far the water actually moves

This is not a theoretical worry. These are recorded stage ranges from gauged rivers, and they are why a river system cannot be designed from a photograph of the site in the dry season.

Sources: Smithsonian Atlas of the Amazon via the Aguas Amazonicas river-level compilation, which reports average annual fluctuations of 10 to 15 m between the middle Madeira and the middle Jurua and about 10 m at Manaus, with the estuary held under 2 m by tidal action; and Mekong River Commission gauge records for Stung Treng, where monthly levels sit between 2.36 and 3.23 m in February to April and annual maxima from 1995 to 1999 ran from 8.0 m to 12.2 m in August and September.

Two things in that table should change how you think about your own site.

The first is the foothill row. An Andean tributary can rise 4 to 9 m in one day. If your pump is on a fixed platform 3 m above the normal water line, it is underwater tomorrow, and the day after that it is 6 m above a falling river. Fixed elevation is not a design on a river like that; it is a bet.

The second is that the estuary row goes the other way. Where tides dominate, the swing is small but the water moves twice a day, which is the opposite problem: your intake has to be designed for reversing flow and salinity, not for a slow seasonal rise.

Design at the lowest stage, but survive the highest

Here is the rule that ends most homemade river systems, so read it twice.

You must set the intake at the lowest water level you will ever see, and you must keep the pump electronics above the highest water level you will ever see. Those two requirements point in opposite directions, and the distance between them is the stage range. On a river with a 4 m swing, that is a 4 m vertical problem before you have added a single metre of pipe.

There are only four honest answers.

Move the pump. A pump on a sliding bracket on a bank post, or on a float, rides the level. Cheap to build, but you now have a moving part in a flood, and a float tied to a bank in a fast current is a real risk to whoever maintains it.

Put the pump in the water. A submersible on a cable or a rail never has a suction lift, because the water is already around it. This is the standard answer on rivers with a large swing, and it also removes the priming problem entirely.

Take the water from the bank instead of from the river. Sink a well into the bank a short distance back, and let the river come to you through the ground. This damps the level variation and filters the water at the same time.

Take the water from the bed. A horizontal slotted pipe under the bed, an infiltration gallery, keeps the level variation and the debris out of the picture, at the cost of the most excavation of any option.

And there is a fifth answer that is not honest, which is to place a surface pump at a fixed height somewhere in the middle of the range and hope. That is the design that works in every month except the two that matter.

Five ways to take water out of a river

Look at the suction lift column and you can see why the surface pump loses on a big river. A bank-mounted centrifugal pump has a hard physical ceiling, and it is lower than most buyers expect.

At sea level, atmospheric pressure can support a water column of about 10.3 m in theory. In practice you get 6 to 7 m, and the gap is not negotiable: the pump needs a margin above its own required net positive suction head, the strainer and foot valve cost you something, and the suction hose itself consumes head. Our worked example later in this article needs 2.5 m of suction head for an 8 m hose at 13.6 m3/h, which is comfortable. Now put the pump 4 m higher than the water because of a flood mark, and you are at 6.5 m before the river has even dropped.

There is a second, less obvious penalty, and it is visible in the measured curve of DCPM50-17-110-1500-A/D, a 1500 W surface centrifugal with a 3 inch outlet. Its measured points run from 52 m3/h at 1 m of head down to 44 m3/h at 5 m, 28 m3/h at 11 m, 16 m3/h at 14 m, 8 m3/h at 16.5 m and zero at 17.5 m. From 5 m to 17.5 m the pump loses 44 m3/h of flow, which works out at 3.5 m3/h lost for every extra metre of head. This is what a high-flow surface pump is: brilliant at near-zero head, and finished by the time you reach 17.5 m.

That steepness is why the delivery pipe in the next section matters so much, and it is also why surface pumps have a ceiling on this kind of job. Across the whole surface range in our catalogue, once you are moving 23 m3/h or more the highest maximum head available is 30 m on a 2.2 kW machine, and most high-flow surface pumps stop between 14 and 21 m: DCPM15-14-48-550 at 14 m, DCPM21-14-72-750 at 14 m, DCPM27-21-110-1500 at 21 m. Above roughly 25 to 30 m of total head at a serious flow, a surface pump stops being an option and you are choosing a submersible whether you wanted one or not. That is not a recommendation, it is a boundary in the product data, and it points the same way as the physics: on a river, put the pump in the water.

The submersible side of the catalogue does not have that ceiling. 4/6DSC36-22-110-1500 is a 1500 W unit that moves 36 m3/h at 22 m, and 4DSC15-45-110-750 does 15 m3/h at 45 m on 750 W. Compare that with DCPM27-21-110-1500, a surface pump of the same 1500 W that is rated 27 m3/h at 21 m. Same power, same head class, and the submersible is ahead by about a third on flow, because the machine is sitting in the medium it is moving instead of reaching down into it.

Riverbank wells: the filter you already own

A riverbank well is the option most people have never heard of and the one that solves the most problems at once. You sink a well into the bank 15 to 100 m back from the water, and the river has to travel through the bank material to reach it. What arrives in the well is river water that has already been through a sand and gravel filter, and the level in the well is not the river’s stage. It is a damped, delayed version of it.

The numbers from documented riverbank filtration systems are worth knowing, because they are better than most irrigation buyers expect:

For an irrigation system, the water quality gain is nice but the real prize is the stable level. That is what turns an impossible design into an ordinary one, because a well in the bank lets you use a standard borehole submersible instead of a floating contraption. Your pump sits at a fixed depth, its cable never has to survive a flood, and the head you design for does not move by 4 m between seasons.

The catch is clogging, and the same review is blunt about it: in the sediment-rich tropical rivers that look most attractive for irrigation, roughly half of documented systems suffer physical clogging of the infiltration interface. Budget for it. Design the well so the pump can be pulled and the screen can be jetted, and expect to do that work on a schedule rather than in an emergency.

Two more practical points. First, keep the well far enough back that flood scour cannot cut into it, but close enough that you are still inducing river water rather than mining the local aquifer. Second, test for salinity if you are near a tidal reach: a riverbank well in an estuary can pull brackish water, and no impeller material fixes that.

Infiltration galleries: the pipe under the riverbed

If you want the filtered-water benefit without a vertical well, the horizontal version is an infiltration gallery: a slotted or jointed pipe laid beneath the riverbed or in the bank, surrounded by a graded gravel filter, with the pump drawing from a collection point on dry land.

The published design rules are specific, and they are worth following exactly:

• Pipe diameter 75 to 300 mm, jointed or slotted.
• A graded gravel filter around the pipe, coarse near the pipe and finer outward.
• At least 1 m below the dry-season saturated zone, so the gallery never draws air at the end of the dry season.
• At least 1.5 m deep measured against river scour, so a flood does not expose it.
• Choose a bed of medium to coarse sand and avoid the inside of river bends, where deposition happens and clogging accelerates.
• Do not let a gravel bed sit in direct contact with the river water: fine particles penetrate gravel deeper than they penetrate sand, and settle where scour can no longer lift them out again.

That last point is the one people get wrong, and it is counterintuitive. A gravel surface looks like it should filter better and stay cleaner. In practice sand resists clogging better, because fines stay near the surface where the next flood can wash them away.

For a farm, the honest summary is that a gallery is the best-quality option and the most work. If your river has a sand bed, a stable low-water channel, and you can get machinery onto the bank in the dry season, it is worth it. If not, a riverbank well gets you most of the benefit for a fraction of the excavation.

What the river is carrying, and where it sits

A well gives you water with almost nothing suspended in it. A river gives you a moving load of solids, and the single most useful fact about that load is that it is not evenly spread through the water column.

At the Obidos gauging station on the Amazon, suspended sediment measured 300 to 340 mg/L close to the bed and only 50 to 70 mg/L in the upper part of the same vertical section. The mean concentration at the station, averaged over years of monitoring, is 45 to 80 mg/L. The Solimoes and Madeira tributaries, which together carry most of the basin’s sediment, run above 100 mg/L. Around 90 percent of the load is silt and clay, and the bed material at Obidos has a median diameter near 0.2 mm.

Two conclusions follow immediately:

• Lift the intake off the bed. A strainer sitting on the bottom sees four to six times the solids of one hung a metre higher. This one decision does more for pump life than any filter you can buy.
• Expect a second sediment peak when the water falls. On the Amazon, sediment discharge rises during the falling-water season from June to September, as material deposited on the floodplain is resuspended, producing an average increase of around 72 percent. The cleanest-looking river is not necessarily the one in flood.

For what sediment does to impellers, and why stainless or hardened components matter more in a river than in a borehole, see plastic vs stainless steel impeller and sand handling in sandy wells.

Then decide how much of it you are going to remove, and be realistic about the answer. A settling basin is sized by surface area, not by volume, and the rule is simple: the required surface area equals the flow divided by the settling velocity of the smallest particle you want to catch.

Settling velocities below come from Stokes law for quartz-density particles in water at 20 degrees Celsius, and the areas assume the ideal case so you should double them for a real basin.

Read the last column pair together and the design decision makes itself. Removing sand is cheap: a basin the size of a small table handles a 0.1 mm grain at 20 m3/h. Removing silt is not: drop to 0.02 mm and the same flow needs a basin 25 times larger, which is a pond, not a tank. So settle the sand, control the silt by keeping velocity low and by lifting the intake, and if you need genuinely clean water at low cost, buy it from the bank instead, with a riverbank well.

Then size the clean-out, because the basin is a solids trap and somebody has to empty it. At 300 mg/L, the near-bed figure from the Amazon, a pump moving 12.5 m3/h takes in about 3.75 kg of solids per hour. That is roughly 2.7 litres of wet mud per hour, or about 16 litres over a six-hour pumping day, and around 39 litres a day if you are moving 30 m3/h. Put a drain and a clean-out hatch on the basin and put the job in the maintenance calendar, because a full basin stops settling and starts passing everything through.

The 300-metre problem: the delivery line picks your pump

Here is the part that catches out almost every river project, and it has nothing to do with the river. On a borehole, the pump is usually near the tank. On a river, the pump is at the water and the field is somewhere else, and that distance is head.

Friction head in a pipe scales roughly with length divided by diameter to the power of 4.87. Diameter is in the denominator of a high power, which means diameter is worth far more than most people assume. These figures are computed with the Darcy-Weisbach equation for PE pipe with a roughness of 0.007 mm, carrying water at 20 degrees Celsius.

Three things fall out of that table.

A hose is not a delivery pipe. The same 13.6 m3/h that costs 2.3 m through 90 mm PE costs 81.7 m through a 1.5 inch hose over 300 m. That is not a rounding error, it is more head than most solar pumps can produce on their best day, and it is the single most common reason a river system that was sized correctly on paper delivers nothing in the field.

Distance and diameter trade off remarkably evenly. A 63 mm line running 60 m costs 2.6 m of head; a 90 mm line running 300 m costs 2.3 m. So halving your distance and keeping the thin pipe gets you roughly what you would have got by keeping the distance and buying the fat pipe. One of those two things usually costs nothing at all, and it is the one you should look at first.

Velocity is a second reason to go bigger. At 25 m3/h, 63 mm PE runs at 2.9 m/s, which is above the 1.5 to 2 m/s band you normally aim for, and it will scour, hammer and wear. The same flow through 90 mm runs at 1.4 m/s and behaves. If you want the full method rather than three sample rows, it is in the pipe sizing guide.

A worked example: 1.5 hectares, 300 metres from the river

Let us put all of it together. A 1.5 hectare vegetable plot on a floodplain, 300 m from the low-water channel, on a river that swings 4 m between the dry-season low and the flood peak.

Step 1, the demand. At a net 5 mm per day over 15,000 m2 the crop needs 75 m3/day. With 5.5 effective pumping hours in a tropical solar day, the system has to deliver 13.6 m3/h. If you are working out a site of your own, the method is the same as in how to size a solar water pump.

Step 2, the head. Static lift from the low-water surface to the highest hydrant is 6 m. Add the suction side and the fittings, and the two delivery pipe options separate immediately.

Now watch what that difference does to the pump choice, using the measured curve of DCPM50-17-110-1500-A/D, a 1500 W surface centrifugal:

• At 20.8 m the pump delivers nothing at all. Its measured curve reaches zero at 17.5 m. Your 1500 W machine is not “a bit short”, it is out of the water. To move 13.6 m3/h at 20.8 m you have to leave this class entirely and step up to a 3 kW machine.
• At 10.3 m the same pump reads about 29 m3/h on its curve, more than twice what the field needs, and the day’s 75 m3 is done in about 2.6 hours.

That is the whole lesson in two lines. The pipe did not change the water requirement, the field, or the sun. It changed the pump class.

Step 3, the intake. With a 4 m stage swing, a bank-mounted surface pump placed above the flood mark faces 4 m plus its own 1.5 m of freeboard as suction lift, and after the hose and strainer you are at roughly 6.5 m against a practical ceiling of 6 to 7 m. That is not a design, that is a coin toss every dry season. So take the intake decision that the physics is pointing at and hang a submersible in the water. The table below lists the real catalogue options for a river duty point of about 10 to 14 m of head, sorted by the choice you are actually making: surface or submersible.

For this site, 4/6DSC30-19-110-1100 has comfortable margin at a 10.3 m duty point, and 4/6DSC36-22-110-1500 or 4DSC19-35-110-1500 are the next steps up if you want head in reserve for a longer hose or a raised tank.

If instead you can sink a well into the bank, the design gets easier again, because the level in the well is damped and you can use a standard borehole machine: 4DSC15-45-110-750 covers 15 m3/h at 45 m on only 750 W, and 4DSC9.5-50-110-750 or 4DSC6-67-72-750 handle smaller flows at higher heads. For the difference between the two pump families in general, see submersible vs surface pump.

Step 4, the array. Solar arrays here are specified at 1.3 times pump power or above, so a 1100 W pump wants at least 1.43 kW of panels: three 550 W modules give you 1.65 kW. Then check the series wiring against the controller’s open-circuit voltage ceiling before you buy cable, which is the mistake that destroys more controllers than lightning does.

Step 5, the buffer. Because the pump delivers 29 m3/h into a 13.6 m3/h requirement, add a tank. A 10 m3 buffer lets you irrigate in the evening, keeps the pump off its float switch, and turns an oversized pump into a non-problem.

Step 6, the silt. If you are drawing straight from the river rather than from a bank well, and your intake hangs half a metre off the bed, size a settling basin for the sand. At 13.6 m3/h the 0.1 mm row of the Stokes table needs about 0.42 m2 of ideal surface area, so build roughly 1 m2 of surface with a 0.5 m working depth. Be clear about what that buys you: it removes sand, not silt. At that surface area your overflow rate is 3.8 mm/s, which catches anything coarser than about 0.065 mm and passes the rest.

Commissioning, clogging and five mistakes that end river projects

Before you buy anything, spend one season measuring. Put a graduated post in the bank and record the level weekly, and find out from the local authority or a nearby gauging station what the historic low and high water marks were. A river system designed on one site visit is a system designed for the weather that day.

Then avoid these five, which cover most of the failures I see.

1. Designing at today’s water level. Size everything at the lowest stage on record, then check that the highest stage does not submerge your controller. If those two cannot both be satisfied with a fixed pump, the pump has to move, or the intake has to move to the bank.
2. Using suction where the pump belongs in the water. A submersible in the river costs less than the earthworks to make a bank pump survive a flood, and it removes the suction limit entirely. This is why seasonal rivers and silt loads push African systems towards submersibles as well.
3. Buying the pump before the pipe. In the worked example above, 63 mm instead of 90 mm turned a 1500 W job into a 3000 W job. Pipe is a fraction of the cost of a pump, a controller and an array, and it is the cheapest head you will ever remove.
4. Letting the strainer sit on the bed. Remember the Obidos numbers: 300 to 340 mg/L near the bed against 50 to 70 mg/L near the surface. Raise the intake and you cut the abrasion load several times over for the price of a longer rope.
5. Building an intake nobody can service. A strainer you cannot reach, a pump you cannot pull, a basin with no drain. Every one of those turns a two-hour job into a two-week outage in the middle of the growing season.

If you would rather not do the arithmetic on a site you cannot easily visit twice, the sizing tool will do the head calculation for you, or send me the numbers and I will come back with a model, a pipe size, an intake arrangement and a panel count that fit the river you actually have rather than the one in the photograph. Full working of the friction, settling and sediment figures in this article is reproducible with the project script, so you can check every number against your own site.

Planning a solar pump for a river or canal? Put your low-water level, your lift and your pipe run into the sizing tool to see your real head, or message me on WhatsApp with your low-water level, your lift from the low-water line, your distance to the field, the seasonal swing you have seen and your crop, and I will come back with a model, a pipe diameter, an intake arrangement and a panel count sized for the lowest stage rather than the day you took the photo.