The way this is usually approached is to use reference evapotranspiration (ET0). Sometimes you can obtain this value daily or weekly more or less in real time from websites, sometimes run by a water company or a university research station. [e.g. A water company in Delaware near me used to provide this information, but it looks like the website is dead.] This value (say for the week) is multiplied by a crop coefficient (Kc), to get the water use for the crop for the week. This approach is sometimes sold to homeowners and farmers to determine how much to irrigate on a weekly basis. If you just want the expected water "need" for a season, you could also use normal ET0 for a location and the Kc for the crop. But then you might have to factor in a probability of e.g. temperature extremes and drought years and rainy years... There are many caveats to this process. 1) Many crops are annuals, so their Kc will change over the life of the plant. 2) Most Kc values have been calculated in arid or Mediterranean climates. It's not known that these values work reasonably for humid climates. 3) Of course you need to find a source of ET0 close enough to your location of interest. 4) The Kc should have been calculated for, say, a healthy crop. But usually lists of Kc are complied from many studies, so you'd really have to go back to the original study to see under what management goal they developed the value. 4) There are of course many other factors that play into this, like micro-climate considerations, non-uniform or imprecise irrigation, and so on.

Thanks for the very informative answer, Salvatore !

The water requirement of a crop must be satisfied to achieve potential yields. The crop water requirement is also called crop evapotranspiration and is usally represented as ETc. Evapotranspiration is a combination of two processes - evaporation of water from the ground surface or wet surfaces of and transpiration of water through the stomata of leaves. The water requirements can be supplied by stored soil water, precipitation and irrigation Irrigation is required when ETc ( crop water demand ) exceeds the supply of water from soil water and precipitation. For more details https://extension.colostate.edu--irrig

The FAO has a standard for ET which is based on the Penman-Monteith equation. This basically requires the use of a full weather station. As mentioned above, there are services that can provide weather station derived ET, there are also satelite based services as well. All of which will require the use of crop coefficients, as mentioned above. There is also a body of work that has looked at estimating ET with reduced data sets, ie models by Hargreaves, Priestley-Taylor, Turc etc, which used just solar radiation and air temperature to estimate ET. In the attached we have tried to assess the errors relative to the FAO approach.

Just a quick follow-up, the attached might be of interest which illustrates the errors associated with reduced measurement methods used to estimate ET.

Thank you Sylvester and Martin for your inputs. So ET0, ETc, the FAO-PM eqn along with crop coefficients all suggest water needs for a well irrigated crop, which is Potential ET. Hence, from a water management perspective (such as decision making of water sharing in a river basin between multiple uses and stakeholders), the PET (+ crop coeff if available) for the region can be taken as the best available estimate of agriculturalk water needs.

Thanks Martin for sharing your tech brief that compares several well known ET models with the FAO-PM. Interesting to note that the Abtew model overestimats PET in Portugal, which i think is much drier as compared to very humid subtropical South Florida where that model was originally developed, and calibrated using lysimeters of the native wetland savanna vegetation. Such models along with MODIS-16 are valuable to estimate PET for much of the world where only air temperature data is available.

Don't you think that potential evapotranspiration is a very indirect quantity to estimate water requirement? Or actual water use? You have to know the crop coefficient, the evaporation contribution, and percolation. I believe uncertainty is large, isn't it?

Hi Luca Brocca , yes you are right. Unfortunately often it comes down to issues of practicality and scale. To obtain the best idea of ET from a particular plant then you probably need to use a weighing lysimeter with a mass/water balance approach. Moving to a field-based study then soil moisture sensors can be used which often leads to "how many sensors do I need in my field" type questions to account for soil type variations etc., so an attractive alternative can be the weather station based ET and crop coefficient method where irrigations are made at a percentage of full ETc in deficit trials. Even larger area water-use (ET) estimates can be obtained from satellite data. Whenever I have looked at studies where crop coefficients are being derived, the data does appear to suffer from clearly observable uncertainty.

Yes, but there are issues of practicability. U got to understand your scope and do calibration. I recommend use of a lysimeter. See Kinama et al publications on the same.

Agricultural water demand includes not only plant water consumption, but also livestock water consumption (although it only accounts for about 10%).

Yes Luca, the uncertainty in estimating "actual " water use of a plant or farm or forest is certainly quite high. I meant in getting an estimate of water that may be needed for agriculture, from a multiple use river basin water management perspective. Like allocating 30 % for agriculture, 10%for industry, 15% for domestic, 25% for critical environmental flows and so on.

As Martin and Samuel suggest, the idea may be to use PET as a first estimate, while continually refining and customizing PET estimates for region by using a variety of field techniques. And then using the same to calibrate and fine tune remotely sensed products.

In data poor regions of the world. PET estimates are easy to come by ( ex MODIS 16 gridded datasets from U of Montana). Even though I have found these values to at times be 50-70% higher than expected using FAO-PM and local weather data, still they can provide an estimate where no other estimate exists.

Huazhen, thank you for including livestock water consumption - yes, that water typically is pumped from groundwater in the dry season.

Thanks everyone for this discussion.

Here we discussed the challenges and difficulty of quantifying in-season water and evapotranspiration.

Restricted water allocations: Landscape-scale energy balance simulations and adjustments in agricultural water applications

https://www.sciencedirect.com/science/article/pii/S0378377419310005