COMMENT

I would like to add a bit of historical perspective to the work Messrs. Bennett and Harms report on in their very interesting paper. The paper mentions that “empirical relationships between crop yield and evapotranspiration have been used for decades to examine the effects of water stress on potential yield reductions” (p. 160). While this seems indeed to be the case, there is an account of such use of this relationship that predates by several years the earliest of the paper's references supporting that statement (Stewart et al., 1977). The account was published in the proceedings of an ASCE Irrigation and Drainage Conference held in 1971 (Egli, 1971a).

At that conference, I described the structure and functioning of the core of a planning model whose development had been funded by what was then Canada's Department of Energy, Mines and Resources (Egli, 1971b). The simulation model's central function was to produce information that would help in deciding on the optimal combination of reservoir size and extent of irrigated area. This optimum depends on the project's socio-economic objective. Where the primary objective is to ensure the irrigator's welfare, as is common in industrialized countries, the occurrence and magnitude of water shortages are minimized by limiting the irrigated area. Where the objective is to maximize overall crop output and provide employment for as many as possible, as is typical for developing countries, more frequent and more severe shortages due to irrigating a larger area are accepted. In the first case, the project benefits are concentrated in the hands of the irrigator, in the second case, they can be considered as essentially accruing to society. To provide the basis for finding the appropriate balance between the outcomes for the two classes of beneficiaries, the planning model had to allow simulating both the financial effects experienced by the individual irrigator and the total production of crops on the command area, and this for any combination of reservoir size and irrigated area.

For the model to be able to generate credible results, it needed to integrate findings and methods of agronomy, hydrology and economics. Developing an operational link between the irrigation inputs (water and labor) and crop production – with the requisite irrigation depending on the weather and the soil properties, and crop production on the weather and the actually applied irrigation water – turned out to be the greatest challenge by far. Irrigation experiments traditionally focused on determining the amount of water necessary to satisfy atmospheric demand and maximize crop yield. This did not lead to the development of production functions that would allow estimating crop yield at lower levels of irrigation, whether due to a water shortage or practiced to avoid exceeding the economically optimal amounts of applied water and related labor. From experiments with dry-land crops, agronomists had been aware for some time of the significant correlation between dry-matter production and evapotranspiration (e.g., Hanks et al., 1969). The potential of the ET-Yield relationship for use in irrigation planning or scheduling was indicated by the publication of a proposal to undertake field experiments specifically aimed at producing the necessary data (Stewart and Hagan, 1969).

Scouring the crop-research literature for sets of concomitant ET and yield data, I located five papers reporting on experiments with 10 crops, seven of them grown in Alberta (Hobbs et al., 1963; Krogman, 1967). The polynomial yield response functions I fitted to each crop's set of data had all the shape one would expect, as well as satisfactory values of the indicators for goodness of fit generated by the regression analysis. Not surprisingly, the regression equation for the dry-matter yield of repeatedly harvested grass was close to linear and had the best fit. What did surprise me was the also relatively good fit of the functions for crops whose primary economic product is grains or roots.

To demonstrate the model's functioning, I ran it for three concurrent time series of climatic and hydrologic data recorded from 1931 through 1960: daily precipitation and monthly sunken-pan evaporation at Swift Current, and monthly flow of the South Saskatchewan River at Saskatoon. The model component representing crop production incorporated a 5-day-interval soil-moisture budget. It generated as output for every simulated growing season the yield of each of eight crops assumed to grow on the predominant three soils of the first 43,000 acres irrigated from Lake Diefenbaker, as well as the amount of water applied and number of irrigations practiced on each soil. The component producing as output the hydrologic effects of varying reservoir size and irrigated area featured a reservoir model reflecting the properties of Lake Diefenbaker and a command area of which the aforementioned soils were assumed to occupy a third each.

Since the earliest attempts to numerically simulate the behavior of hydrologic systems (e.g., Linsley and Crawford, 1960), the effort of structuring and writing the computer program constituting the simulation model has inevitably led to the identification of processes and relationships that needed to be known better if the model was to generate output of an acceptable degree of reliability. The need to develop a procedure supported by experimental data that generates a realistic crop-yield estimate as a function of the applied amount of irrigation water is no exception. A generalist water-resources planner, I was at the time admittedly not too sure about the reasonableness of using the proposed ET-Yield relationship and method of dealing with a water shortage. Therefore, I suggested that this question (as well as that of regional transferability of ET-Yield functions) be addressed by people who know more about plant physiology than I do. While it was hardly my model study that gave the impetus for this to actually happen, I'm very happy to see 40 years later that the ET-Yield concept has become accepted as a viable basis for estimating the yield of irrigated crops and is being put to good use.