Session 1 · Foundations

A river basin as a water budget

Before building any model, look at the river. This session introduces the four numbers a rainfall–runoff model works with, shows how river flow is turned into the same units as rain, and uses thirty-four years of Leaf River data to write the basin's water budget.

Time about 25 minutesNeeds nothing
Watch this session first: a 7-minute narrated video with English subtitles. The page below goes further and lets you try each step yourself.

You will be able to

The catchment and its gauge

Rain that falls on land runs downhill. Every point on a river therefore collects water from a definite area: all the land from which water flows to that point. That area is the point's catchment (also called a watershed, drainage basin or river basin). Its boundary is the line of high ground, the divide, where rain on one side drains towards the point and rain on the other side drains elsewhere.

The point itself is usually a stream gauge, a station where the water level is recorded continuously and converted to discharge, the volume of water passing per second. The U.S. Geological Survey (USGS) runs thousands of them. Our gauge, USGS 02472000, sits on the Leaf River near the town of Collins in southern Mississippi. Its catchment covers 1924 km², about the size of a square 44 km on a side: low hills (mean elevation 123 m), 67% forest, and loamy soils.

Why this river

The Leaf River is a classic test case for rainfall–runoff models. Sorooshian, Duan and Gupta (1993) used its historical record to show that the SCE-UA search algorithm can reliably find the best settings of the Sacramento model, a conceptual model of the same family as CREST. Many studies of model calibration have used it since. Its record is complete: not a single day of flow is missing in water years 1981–2014.

Four numbers per day

A daily rainfall–runoff model sees the world as four numbers per day, each one an average over the whole catchment.

SymbolNameUnitsWhere it comes from here
Pprecipitation (rain and snow)mm/dayDaymet, a gridded daily dataset built from weather stations, averaged over the catchment
Tair temperature, daily mean°CDaymet, as above
PETpotential evapotranspirationmm/daycomputed from temperature and the sunlight reaching the top of the atmosphere at that latitude and date (the Oudin et al., 2005, formula)
Qdischarge (river flow)mm/dayUSGS gauge record, converted to a depth (next section)

The first three are the model's forcing: they drive it. The fourth is what the model must predict. PET needs a word of explanation. It is not a measurement of evaporation. It is an estimate of how much water the air could take up from a surface that never runs dry: a thirst, set by sunshine and warmth. How much is actually evaporated depends on how much water the land can supply, and that is one of the things the model has to work out.

Key idea

The basin-averaged daily series are a simplification. Rain varies within the catchment and within the day. A daily, catchment-averaged ("lumped") model accepts this loss of detail in exchange for being simple and fast. dCREST is such a model.

River flow as a depth

Rain is measured as a depth: 10 mm of rain means that a layer of water 10 mm thick fell everywhere. River flow is measured as a volume per second, for example 50 m³/s. To compare the two, spread the river's daily volume evenly over the catchment and ask how thick a layer it makes:

\[ Q\;[\text{mm/day}] \;=\; \frac{Q\;[\text{m}^3/\text{s}] \times 86\,400\ \text{s/day}}{A\;[\text{km}^2] \times 10^6\ \text{m}^2/\text{km}^2} \times 1000\ \text{mm/m} \;=\; \frac{86.4\; Q\;[\text{m}^3/\text{s}]}{A\;[\text{km}^2]} \]

For the Leaf River, \(A = 1924\) km², so 1 m³/s is 0.0449 mm/day, and 1 mm/day of flow is 22.3 m³/s. From here on, all water amounts in this course (rain, evaporation, storage, flow) are depths in millimetres, so they can be added and subtracted directly.

Try itConvert river flow for the Leaf River

Type in either box. A flood of 1 000 m³/s at the Leaf River gauge is a layer of about 45 mm over the whole basin in one day.

Thirty-four years of the Leaf River

The chart shows one water year at a time. Hydrologists count years from 1 October to 30 September, so that each winter sits in a single year; water year 1990 runs from October 1989 to September 1990. Rain hangs down from the top as blue bars, and the river's flow rises from the bottom as a dark line. Drag across the chart to zoom, double-click to zoom out.

Try itThe record, one water year at a time
this year

Top: precipitation (bars, right axis, pointing down) and river flow (line, left axis), both in mm/day. Bottom: daily mean air temperature and potential evapotranspiration.

Three things are worth noticing in almost every year. First, flow responds to rain within a day or two, then recedes over weeks. Second, the same amount of rain produces far more flow in winter than in summer. Third, the river never stops: even after weeks without rain, some water still arrives from storage underground. A model has to reproduce all three.

The water budget

Water is never created or destroyed. Over any period, the water that arrives as precipitation either leaves again, as evaporation or as river flow, or stays in the catchment as stored water (in snow, soil and groundwater):

\[ P \;=\; ET \;+\; Q \;+\; \Delta S \]

\(ET\) is actual evapotranspiration, the water that really went back to the air; \(\Delta S\) is the change in storage over the period.

Over a single month \(\Delta S\) can be large, but over many years storage cannot keep growing or shrinking, so \(\Delta S \approx 0\) and the budget gives a way to estimate the hard-to-measure evaporation: \(ET \approx P - Q\). For the Leaf River in water years 1981–2014:

Average over water years 1981–2014mm/yrshare of precipitation
Precipitation P1515100%
River flow Q47832% (the runoff ratio)
Evapotranspiration, \(ET \approx P - Q\)103768%
Potential evapotranspiration PET1126for comparison

Two thirds of the rain goes back to the air. Actual evaporation reaches 92% of the potential: in this humid basin the land can usually meet most of the atmosphere's demand. In a drier basin, the land would run out of water and actual evaporation would fall far below its potential.

DataPrecipitation and river flow, every water year

Light bars: precipitation. Dark bars: river flow. The gap between them is, give or take a change in storage, the water that evaporated. Dry years lose a larger share of their rain to evaporation.

Why the river runs high in winter

Rain at the Leaf River is spread fairly evenly over the year, between 3.1 mm/day in October and 5.3 mm/day in February. River flow is not: it averages 3.2 mm/day in February and only 0.39 mm/day in August. The difference comes from the atmosphere's thirst. Potential evapotranspiration climbs from 1.0 mm/day in January to 5.3 mm/day in July.

DataThe average year

Monthly means over water years 1981–2014, mm/day.

In summer the air takes back almost all the rain, the soil dries out, and new rain is used to wet the soil again instead of reaching the river. In winter evaporation is small, the soil stays wet, and rain runs off. The same rain gives very different flow depending on how wet the catchment already is. That wetness is a hidden quantity: it is not in the four numbers above. Tracking it is the main job of a hydrological model, and the subject of the next session.

In Python

The cell below loads the Leaf River data (the same file the widgets use) and computes the water budget. Edit it and press Run. For example, change the years, or compute the runoff ratio of the wettest year only.

Check your understanding

A gauge records 200 m³/s. How many mm/day is that for the Leaf River?

\(86.4 \times 200 / 1924 = 0.0449 \times 200 \approx 9.0\) mm/day.

Over 34 years the basin received 1515 mm/yr and the river carried 478 mm/yr. Where did the rest go, and what assumption did we make?

It evaporated (evapotranspiration, about 1037 mm/yr). We assumed that the change in stored water over 34 years is negligible compared with the totals, and that no water leaves the basin underground past the gauge.

Why is potential evapotranspiration larger than actual evapotranspiration?

PET is what the air could evaporate if water were always available. When the soil is dry, the land cannot supply that much, so actual evaporation is lower.

February gets about a third more rain than August, yet its river flow is about eight times larger. Why so much more?

The extra rain explains only a small part. In August the atmosphere's demand is about three times larger than in February, so most rain is evaporated and dry soil soaks up the rest. In February the soil is already wet and little evaporates, so much more of the rain becomes flow. The catchment's wetness, a stored quantity, controls the response.

What to remember