Letter to the editor: Discussion on Friedman S.P. 2023 ‘Is the crop evapotranspiration rate a good surrogate for the recommended irrigation rate?’

A recent publication by Friedman (2023) aimed to open a discussion on the extensive reliance on the evaluated crop evapotranspiration rate for optimal irrigation recommendations. The main argument in the paper is that using estimated crop evapotranspiration to replenish the soil could either substantially over- or underestimate the optimal irrigation rate. This claim is then supported by two extreme examples: (1) extensive, low-frequency irrigation of deep-rooted crops grown in fine-textured soils during or after the rainy season where the contribution of soil water and shallow groundwater to crop water uptake is significant; and (2) intensive, high-frequency irrigation of shallow-rooted crops planted in coarse-textured soils, where deep percolation occurs. In both cases, estimates of the evapotranspiration rate as the required irrigation dose are suboptimal, and therefore, the main argument in Friedman's paper is valid.

Friedman admits that the paper does not convey new information yet presents a sound analysis to support the general message—that the optimal irrigation dose does not equal the crop evapotranspiration. The discussion of this point is important, since it is perhaps forgotten or ignored at times, and thus often the recommended irrigation rate is set as the estimated crop evapotranspiration. However, Friedman fails to support one of the main conclusions, rendering it an unfounded opinion: ‘… efforts in research and practice to evaluate crop evapotranspiration for recommending an optimal irrigation rate are not always justified’. Instead, Freidman nostalgically suggests returning to empirical ‘yield-seasonal irrigation rate production functions’ and basing irrigation decisions on ‘gained, case-specific knowledge’. Unlike near-real-time sensing-based estimations of evapotranspiration, this suggestion, by definition, cannot lead to optimal irrigation, mainly since it ignores the spatial heterogeneity in the field. Additionally, interannual variations in precipitation temporal patterns are ignored by yield-seasonal irrigation rate production functions. It is, therefore, inhibitive to precision irrigation practices that allow the application of water (and nutrients) to the plant at the right time and place and in small measured doses to provide it with optimal growing conditions. However, Friedman later rationalizes this with the unsupported claim that ‘at large, it seems that the technological developments of sensing and telemetry, data processing and artificial intelligence decision-making are running ahead, with a yet unproven conjecture that basic economic agronomic strategies can be disregarded (circumvented) when optimizing irrigation and related agriculture practices’.

As a remote sensing and precision irrigation scientist, I have difficulty with such opinions, mainly when they are not well rationalized and supported with referenced evidence. Indeed, lousy irrigation solutions and products relying on technology-based evapotranspiration estimates may exist. Yet, Friedman fails to point them out; thus, his generalization and labelling of this entire line of solutions are without merit. Returning to old production functions lacking a spatial component will keep us away from optimal irrigation. Moreover, intimate, personal knowledge of one's field is no longer adequate in the age of large-scale industrial farming, where agronomists manage extensive, dynamic lands and may not manage the same land throughout their career. I believe that the way towards sustainability is not by going backwards but by harnessing lessons like the main message in Friedman's paper to help us develop more innovative, more sophisticated ways to determine the irrigation rate. In the context of Friedman's paper, crop growth models can be regarded as ‘production functions’, which can be upscaled to account for spatial variability. Accordingly, I would like to suggest that evapotranspiration rate estimates must be supplemented by more complex models that consider water movement in the soil. Assimilating this information into a crop model, for example coupling DSSAT and HYDRUS 1-D (Shelia et al., 2018), is an exciting way forward that can well support irrigation decisions.

Moreover, in the context of spatial heterogeneity in the field, remote sensing observations can be assimilated into a crop model to improve crop model predictions in space and time and even predict water stress in advance, allowing better decision support (Berger et al., 2022; Manivasagam & Rozenstein, 2020). Therefore, improving such mechanistic approaches would make evapotranspiration estimates beneficial for determining the optimal irrigation rate under a wide range of conditions. I am a big advocate of mechanistic approaches. Still, I wholeheartedly believe that artificial intelligence solutions can be used successfully to estimate the evapotranspiration rate (e.g. Rozenstein et al., 2023) and that this estimation can then be assimilated into a comprehensive model to optimize irrigation decisions.

In conclusion, I believe that in the quest for sustainability, efforts in research and practice to evaluate crop evapotranspiration for recommending an optimal irrigation rate are justified. Moreover, contrary to what Friedman suggests, these efforts are well worth public and industrial research investment.