3.1.1.2 Irrigation season
In 2021 and 2022, the farmer irrigated every 5-7 days starting mid-May through October. Irrigation amounts per event varied strongly and averaged 14 and 25 mm in S09 and S10 respectively (upper panel of Figure 3 and Figure 4). Irrigation increased SM by up to 10 vol% in the top 5 cm and about 5 vol% at 50 cm depth. To represent the observed irrigation schedule, the CLM5 irrigation routine can be adjusted in two ways, by (1) adapting \(\psi_{\text{target}}\) or (2) tuning the\(f_{\text{thresh}}\) parameter. Figure 5 shows the effect that different values of these two parameters have on several aspects of the simulated irrigation (e.g., start of irrigation period, number of irrigation events, irrigation frequency) as well as on SM and crop yield. In both cases, a lower parameter value results in a later onset of irrigation, fewer irrigation events and lower total irrigation amounts. However, the parameters have different effects on irrigation frequency, whereby smaller values of \(f_{\text{thresh}}\) result in less frequent irrigation events while the irrigation volume per event increases (Figure 5). Changing\(\ \psi_{\text{target}}\), on the other hand, has little effect on the irrigation frequency and volume. SM in the upper 50 cm of soil increases with increasing values of both parameters. The increase is exponential for\(\psi_{\text{target}}\ \)with values ranging between 0.195 and 0.275 cm3 cm-3 and almost linear for\(f_{\text{thresh}}\) with a somewhat smaller range. Consequently, varying \(\psi_{\text{target}}\) has a more pronounced effect on yield compared to \(f_{\text{thresh}}\) for the investigated range of parameter values.
For the model run using the standard irrigation routine, we set\(f_{\text{thresh}}\) to 0.7 while leaving \(\psi_{\text{target}}\) at its default value of -34 kPa, which resulted in approximately weekly irrigation events of on average 26 mm per event, starting mid-May. This, however, could only partially reproduce the observed irrigation schedule and SM dynamics compared to using the irrigation data stream. Nevertheless, both irrigation approaches showed fluctuations of similar magnitude compared to the observed values in the upper soil. Less dynamics than observed were simulated at 50 cm depth for both irrigation approaches and both orchards. The wet bias in S10 was still persistent throughout the profile for the simulation using the irrigation data stream while simulated SM based on the default irrigation routine dropped to the range of observed values (Figure 4).
Simulated and observed total yearly irrigation were similar in S09 with the observed effective irrigation being 433 and 458 mm (75% of actual measured irrigation) and simulated amounts being 425 and 439 mm for 2021 and 2022, respectively. In S10, observed effective irrigation amounts were considerably higher than in S09, which could be expected considering the lower observed SM in S10. Compared to the observed 706 and 586 mm, for 2021 and 2022, respectively, the model applied only 393 and 388 mm, which is a result of the simulated wet bias.