Water leaving the snow is not necessarily water just melted above the collector
A collector beneath a snowpack records a sudden pulse of water. It is tempting to call the pulse today’s surface melt. But that water may have travelled sideways, paused within the snow, refrozen and melted again, or entered as rain. The collector has measured an exit pathway. The origin and timing of that water need their own explanation.
Snow lysimeters help connect snow processes with water delivered at the snowpack base. Their usefulness depends on keeping the collecting footprint, observation interval and routing assumptions visible. This guide builds that connection through original calculations, then deliberately changes the conditions to show where a simple interpretation stops working.
Use supplied records for the exercises. This is not an installation guide or an invitation to work in hazardous winter terrain. The learning goal is to distinguish measured collection, inferred local drainage, surface melt production and catchment runoff without treating them as interchangeable.
Hydrology subject guide · Learning that transfers into life · How X Works hub
Choose your reading route
Route 1: Follow the measured water · Chapters 1–3
Route 2: Build the storage account · Chapters 4–5
Route 3: Read space and timing · Chapters 6–8
Route 4: Test a model honestly · Chapters 9–10
Route 5: Transfer and report · Chapters 11–12
Full chapter index · Independent worked case · Questions and practical checks
Chapter index
Follow the measured water · Chapters 1–3
Define the collector boundary first
Convert volume to an equivalent depth carefully
Separate surface melting from basal drainage
Build the storage account · Chapters 4–5
Choose solid and liquid stores without double counting
Allow rain and previously stored water into the explanation
Read space and timing · Chapters 6–8
Distinguish geometric area from contributing area
Understand why layers can redirect liquid
Keep collection-system timing separate from snow timing
Test a model honestly · Chapters 9–10
Use arrays without assuming every pan is independent
Compare the model output with the correct observation
Transfer and report · Chapters 11–12
Perform a changed-context transfer check
Report a result another reader can audit
CHAPTER 1 OF 12
1. Define the collector boundary first
A snow lysimeter collects water crossing a defined area within or beneath snow. For this guide, the target is basal outflow: liquid water leaving the bottom of a snowpack and reaching a collector. The collection point is not automatically the point at which the water first formed as melt. Naming the interface prevents an apparently straightforward record from answering the wrong question.
Draw a horizontal collection area and the snow above it. Add a surface where melting or rainfall supplies liquid, a changing liquid store within the snow and possible lateral routes. Then draw the ground beneath the collector. Water recorded by the collector is not simultaneously a measurement of subsequent soil infiltration, groundwater recharge or river discharge.
State whether the reported quantity is collected volume, equivalent depth over the pan, or an area-scaled flux estimate. All can be useful. Problems begin when one label silently replaces another. A precise bottle volume remains a local observation even when a graph is given the much broader title catchment snowmelt.
CHAPTER 2 OF 12
2. Convert volume to an equivalent depth carefully
The basic conversion is depth = volume divided by area. One litre spread over one square metre corresponds to one millimetre of water. A collector with area 0.25 m² receiving 2 litres therefore records 8 mm of collection depth over its own footprint. The calculation does not yet establish that a square metre of the surrounding landscape released 8 litres.
Keep dimensions on the working line. Two litres is 0.002 m³; dividing by 0.25 m² gives 0.008 m, or 8 mm. If the area was accidentally entered as 25 m², the computed depth would be 0.08 mm, a hundredfold error. Decimal accuracy cannot repair an incorrect area conversion.
For a rate, specify duration as well. Eight millimetres collected over four hours averages 2 mm/h during that interval. The same volume over twenty-four hours averages one-third of a millimetre per hour. Neither average establishes the maximum short-duration rate unless the record also resolves how the water arrived.
CHAPTER 3 OF 12
3. Separate surface melting from basal drainage
Surface melt is production of liquid water by phase change near the snow–air interface. Basal outflow is liquid water reaching the bottom boundary. Storage and routing can separate them in time and space. Kattelmann’s discussion of lysimeters and snowmelt models makes this distinction central: comparing the wrong model output with the collector can create a misleading test.
Imagine 10 mm water equivalent melts near the surface during an afternoon. If 6 mm remains in the snow and no other liquid source or exit exists, only 4 mm leaves the base within the interval. The surface-melt calculation and basal-outflow observation can both be correct despite their different values.
Now examine the following night with no new surface melt. If the stored 6 mm drains, the collector records water while surface melt production is zero. This delayed release is not evidence that the instrument invented water. It demonstrates why an outflow hydrograph needs a storage account before being used as an instantaneous melt-rate record.
CHAPTER 4 OF 12
4. Choose solid and liquid stores without double counting
Let the snow contain solid water equivalent I and retained liquid water L. Melting transfers water from I to L. Refreezing transfers it back. In the combined snow-water store I + L, these internal phase transfers cancel. External snowfall, rain, vapour exchange, lateral movement and basal drainage change the total.
For a simplified liquid account over one interval, ΔL = M + R + Lin − F − B − Lout, where M is melting, R rain reaching the snow, F refreezing, B basal drainage and Lin and Lout are lateral liquid transfers. Every term must use the same area and period. The equation is an accounting structure, not evidence that every term was observed.
If a total-water budget already uses the observed decrease in I + L, do not also subtract melting as an external loss. Melting changes phase without removing water from that boundary. Confusing a solid-only model variable with total water can create a false deficit even when every arithmetic operation is individually correct.
CHAPTER 5 OF 12
5. Allow rain and previously stored water into the explanation
Basal drainage during a rain-on-snow interval can contain rainwater, newly melted snow and water already held in the snowpack. A collector measures their combined exit unless additional evidence separates sources. It does not label each drop according to origin.
Suppose 5 mm of rain reaches a snowpack, surface melting supplies 3 mm, retained liquid storage decreases by 2 mm and refreezing is negligible. With no lateral exchange, basal output is 10 mm. Calling all 10 mm freshly melted snow overstates the supplied melt production by more than a factor of three.
Source identification requires more than a volume residual. Suitable tracers or other independent observations may constrain origin, but their own timing and mixing assumptions matter. A rainfall amount larger than the measured output does not prove all output was rain either: some rain can enter storage while older water is displaced and leaves.
CHAPTER 6 OF 12
6. Distinguish geometric area from contributing area
A pan has a geometric footprint. The water arriving there may originate from a different area because liquid can converge or diverge along snow layers. Webb and colleagues’ lysimeter-array study documented substantial spatial and temporal variability in basal drainage. Its value here is the demonstration of a routing problem, not a universal representative area to copy into every forest.
In a fictional example, 1 m² of surface supplies 10 litres and all of it converges onto a 0.25 m² collector. The collector-area depth is 40 mm, even though the source-area equivalent is 10 mm. Both values follow from the declared geometry. The collector has not violated mass conservation by exceeding the melt depth above its own vertical column.
Reverse the routing: water produced over the pan’s overhead area partly leaves sideways before reaching the base. The collector can then underrepresent local surface production. Without observing or modelling these lateral connections, dividing collection by pan area does not uniquely recover the melt produced directly above it.
CHAPTER 7 OF 12
7. Understand why layers can redirect liquid
Snow is not always a uniform porous column. Layers, interfaces and changing wetness can affect how liquid moves. Preferential pathways can concentrate drainage in some places while nearby collectors remain quiet. An observation of strongly uneven outflow therefore invites a routing explanation before an assumption of equally uneven surface energy input.
A useful conceptual model is a roof feeding several gutters. A large discharge at one gutter need not mean more rain fell directly above that gutter. The analogy does not describe snow physics completely, but it makes the accounting distinction visible: source distribution and exit distribution are linked by pathways, not simply by vertical alignment.
Eiriksson and colleagues investigated lateral flow using several kinds of field evidence. Their research shows why hillslope connections deserve attention. Do not turn that result into a claim that every snowpack has the same pathways or routing speed. The relevant question remains which connections are supported at the site and during the observation period.
CHAPTER 8 OF 12
8. Keep collection-system timing separate from snow timing
The recorded time may include movement through snow and through the collection apparatus. A drainage tube, temporary storage or tipping mechanism can delay or discretise the signal. Instrument timing and environmental travel time are different parts of the evidence chain.
Imagine water crosses the pan at 14:00 but reaches the logger’s measurement point at 14:20. A peak at 14:20 cannot be assigned wholly to snow travel delay without accounting for the collection system. Likewise, a clock offset between meteorology and outflow can create an apparent physical lag that is partly administrative.
Compare compatible intervals and preserve missing observations. An interval without a recorded tip might contain drainage smaller than the recording increment; it might also reflect a fault. A zero, a below-resolution amount and an unavailable observation should not be merged into one numerical label merely to make a continuous graph.
CHAPTER 9 OF 12
9. Use arrays without assuming every pan is independent
Several collectors can reveal spatial variability that one cannot. But counting collectors is not enough to determine the amount of independent evidence. Nearby pans can share a routed pathway, while a sampling design can overrepresent one snow setting. Area representation and correlation both matter.
Suppose a landscape is 80% open ground and 20% beneath trees. Representative class means are 5 and 15 mm of basal drainage. Its area-weighted estimate is 0.8 × 5 + 0.2 × 15 = 7 mm. Sampling five collectors in each class and simply averaging their values gives 10 mm, answering an equal-class-weight question instead.
A cluster of neighbouring high readings may identify a real concentration zone, yet it does not justify treating every high reading as an independent estimate of the basin mean. Preserve sampling locations, setting weights and the basis for spatial inference. Better precision in one pan does not remove uncertainty about the landscape it represents.
CHAPTER 10 OF 12
10. Compare the model output with the correct observation
If a model predicts surface melt, it must include a suitable routing and storage representation before its output can be compared directly with basal collection. Alternatively, compare an observation that actually measures the surface-production quantity. A model–data comparison should identify the quantity on both sides before calculating a score.
A poor hourly fit might reflect routing timing rather than incorrect total melt. A good daily total might conceal a wrong peak or wrong internal store. Use diagnostics matched to the task: interval volume, arrival timing, persistence, spatial distribution and sensitivity to missing data can reveal different failures.
Do not tune a surface-melt factor solely to erase a collector-footprint mismatch. That may improve the numerical fit while distorting the physical production estimate. An apparently better calibration can be compensating for an omitted lateral pathway. Independent storage or spatial observations help distinguish those explanations.
CHAPTER 11 OF 12
11. Perform a changed-context transfer check
Begin with a dry, cold snowpack and a warm-period melt input. Some liquid can be retained or refrozen before basal release. Then repeat the explanation for a wetter snowpack with less available retention. The same supplied surface melt can produce different outflow timing. State the changed storage assumption rather than treating the first hydrograph as a universal response.
Next retain the same water production but redirect a lateral route toward a neighbouring pan. The local collection record changes while total production over the combined area does not. This transfer demonstrates why the spatial boundary matters just as much as the time window.
Finally change the task from local drainage to river runoff. Water leaving snow may enter soil storage, groundwater or other routes before reaching the channel. An additional catchment model and observations are needed. Explaining basal drainage correctly is an important link, not a shortcut around every downstream process.
CHAPTER 12 OF 12
12. Report a result another reader can audit
A useful report says what crossed the collector boundary, over which area and period, with what recording basis. It distinguishes measured collection from assumptions about source area, snow storage and subsequent routing. Give the uncertainty a location in the chain: volume measurement, area, timing, source attribution or spatial expansion.
When the evidence cannot separate newly melted snow from rain and released storage, report combined basal outflow rather than inventing a partition. When lateral convergence is plausible but unresolved, label the collector-area depth as a local index instead of a measured vertical melt depth.
For independent understanding, close the guide and explain how a 0.25 m² collector can record 40 mm while the contributing surface receives only 10 mm of meltwater production. Then explain how outflow can continue at night without new melting. Those two answers show that pathways and stores have remained connected to the numbers.
Independent worked case and evidence checks
Consider a fictional 0.5 m² pan beneath a snowpack. Its logger records 1 litre from 08:00–12:00, 3 litres from 12:00–16:00 and 2 litres from 16:00–20:00. Total collection is 6 litres over twelve hours. All measurements refer to water that reached the collection system, not directly to the surface melt source.
Convert each interval using the pan area. The equivalent depths are 2, 6 and 4 mm. Total depth is 12 mm. The four-hour mean rates are 0.5, 1.5 and 1 mm/h. The whole-period mean is 1 mm/h, which conceals the different interval rates.
An independent surface-energy calculation supplies a fictional surface-melt estimate of 8 mm over the same twelve hours. The collector-area depth is 4 mm larger. Do not immediately declare the model wrong. First ask whether the two quantities share a source area, whether liquid storage decreased, and whether rain or lateral water entered.
For a closed local-column illustration, let initial liquid storage be 5 mm and final liquid storage 1 mm, with no rain, refreezing or lateral exchange. Eight millimetres of melt plus the 4 mm storage release produces 12 mm basal output. The local mass balance closes without increasing surface melt.
Now retain liquid storage unchanged and introduce 4 mm rain instead. Eight millimetres melt plus 4 mm rain again produces 12 mm output. The collector record is identical, but source attribution differs. One outflow total cannot choose between these two scenarios.
Construct a third explanation: no rain and no storage change, but lateral net input of 4 mm on the pan-area basis. Again output is 12 mm. These alternative accounts show an identifiability limit. Additional evidence is needed to decide which terms actually contributed.
Change the area boundary to a one-square-metre source region feeding the half-square-metre pan. Suppose surface melt production over that region is 6 mm and all 6 litres reach the collector. The source-area depth is 6 mm while the pan-area depth is 12 mm. The numerical discrepancy is explained by convergence rather than extra production.
The hypothetical convergence factor is source area divided by pan area, or 2. But deriving that factor from the observed depth alone would be circular if source production was itself uncertain. Geometry and pathways need independent support before the ratio becomes an environmental interpretation.
Next use a total-snow-water check. The local snow column begins with 100 mm solid and 5 mm liquid. Eight millimetres melts. At the end, solid is 92 mm and liquid is 1 mm. Total storage decreases from 105 to 93 mm, a decrease of 12 mm, matching output under the closed-column assumptions.
A mistaken spreadsheet subtracts the 8 mm melt and the 12 mm drainage from the initial total, obtaining 85 mm. It has removed the phase transfer twice: once when solid became liquid and again when liquid left. Keeping solid and liquid columns separate reveals the error.
Introduce refreezing in a new interval. Start with 4 mm liquid, add 7 mm surface melt, refreeze 3 mm and finish with 2 mm liquid. With no rain or lateral flow, basal output is 4 + 7 − 3 − 2 = 6 mm. The refrozen 3 mm remains within total snow water; it is not lost to the catchment.
Now test a collection-system lag. Suppose all drainage crossing the pan during a short pulse is delayed twenty minutes before recording. Comparing the logger directly with surface forcing would combine snow routing with this apparatus delay. Correct time labels using supported system information; do not shift the series until a preferred model fits.
For a different recording issue, let a cumulative logger register only complete 0.2-litre increments. Physical drainage of 0.12 litres followed by another 0.12 litres may appear as no completed increment in the first interval and one in the second, with a remainder. A row showing zero is not necessarily proof of no physical drainage during that row.
The half-square-metre footprint turns a 0.2-litre increment into 0.4 mm. If the scientific question depends on differences of 0.1 mm over short intervals, the supplied recording system does not directly resolve that distinction. Extending the interval changes the question and can improve accumulated-volume interpretation, but it does not recover the missing fine timing.
Compare four fictional collectors with identical areas and twelve-hour depths of 2, 4, 6 and 20 mm. Their arithmetic mean is 8 mm. The median is 5 mm. Neither is universally the correct landscape estimate without a sampling argument. The high value could represent a real concentration zone, a different setting or a measurement issue.
If the four observations represented four equal-area landscape strata, their mean would be an appropriate arithmetic expansion under that design. If the high-reading setting occupied only 5% of the landscape, equal weighting would overstate its representation. Write the design before selecting the summary statistic.
Try a two-setting expansion. Let 90% of a plot have a representative drainage estimate of 4 mm and 10% have 20 mm. The weighted estimate is 5.6 mm. If an independent review changes the high-drainage area share to 20%, the estimate becomes 7.2 mm. This is a sensitivity to spatial representation, not an instrument recalibration.
Now test uncertainty in pan area. Six litres divided by 0.50 m² gives 12 mm. Dividing by 0.45 m² gives about 13.33 mm; dividing by 0.55 m² gives about 10.91 mm. These are illustrative sensitivity endpoints. Do not call them a confidence interval without an appropriate probability model.
Volume uncertainty has a different effect. With area fixed at 0.50 m², a collected volume between 5.5 and 6.5 litres corresponds to 11–13 mm. If both volume and area are uncertain, preserve their joint assumptions. Independently combining every extreme can be conservative but is not automatically a statistically justified interval.
For a final timing exercise, compare two twelve-hour records with equal 6-litre totals. One delivers all water in a single hour; the other delivers 0.5 litres each hour. Their mean rates match, but their peak delivery and potential downstream consequences differ. A daily-volume agreement does not validate rapid-response timing.
Try one final diagnostic without changing the data. A learner obtains the correct 12 mm but calls it groundwater recharge. The conversion is sound; the quantity label is not. Repair the pathway from basal snow drainage through soil and groundwater before changing the arithmetic. Another learner reports 1.2 mm because litres were converted incorrectly. That needs a unit repair instead. Calling both mistakes carelessness would conceal the different missing connections and lead to less useful practice.
A student should now be able to write a bounded conclusion: the pan collected 6 litres, equivalent to 12 mm over its footprint, during the specified period. Surface melt production, source fractions, contributing area and catchment runoff remain separate estimates unless the additional observations and models support them. That smaller statement is the reliable foundation for a larger explanation.
Worked comparison
| Quantity | What the example establishes | What remains separate |
|---|---|---|
| 6 L over a 0.5 m² pan | 12 mm pan-area collection depth | Contributing surface area |
| 8 mm melt plus 4 mm released storage | 12 mm basal outflow in a closed model | Unique source partition in the field |
| 8 mm melt plus 4 mm rain | The same basal output is possible | Proof that all output is melted snow |
| 6 L routed from 1 m² into 0.5 m² | Different source- and pan-area depths | A conservation failure |
| Same total in one hour or twelve hours | Equal volume, different timing | Equal peak response |
Questions and practical checks
Does a snow lysimeter directly measure surface melt?
Not necessarily. A basal collector records water arriving at the snowpack base after storage, refreezing and routing. Match the observation to the interface actually measured.
Can outflow exceed melt calculated above a pan?
Yes under some conditions, including stored-water release, rain or lateral convergence. The difference requires a compatible water balance before being assigned to an error.
Is a zero logger row the same as zero drainage?
Not always. Recording increments, delays, data gaps and faults must be distinguished from a supported physical zero.
Will more pans automatically give a basin mean?
No. Locations, area representation and shared pathways matter. An array can reveal variability without automatically resolving catchment expansion.
What should an independent check target?
The uncertain link: collection calibration, storage change, source composition, timing or spatial representation. More readings of one term do not identify every unobserved term.
Continue with the connected subjects
Follow snow storage and delivery
Snow water equivalent: density and basin expansion
Snow-pillow loads: retained meltwater and station support
Canopy snow: water supplied to the ground from temporary storage
Connect forcing and catchment accounting
Snowmelt models: temperature indices and energy limits
Frozen soil: basal snow drainage is not automatically infiltration or recharge
Water-budget closure: residuals and identifiability
Apply the connected model
Rain–snow partition: distinguish the incoming phase estimate from water collected below snow
Sources and scope
Kattelmann: snowmelt lysimeters in model evaluation
Webb and colleagues, 2018: spatial and temporal variability in a snow-lysimeter array
Eiriksson and colleagues, 2013: lateral snow flow at hillslope and catchment scales
The numerical cases in this guide are original fictional teaching examples, not measurements from a real site. This is an educational explanation, not a field safety instruction, engineering approval or an official certification. For a real project, use qualified practitioners, current product documentation and site-specific evidence. Sources were checked on 30 October 2026.