Can soil moisture sensors support smart irrigation decision making in mountain terrace agriculture?

While our planet is heating up, mountain terraces may be able to maintain agricultural production systems in a cooler environment than the agricultural plains. Mountain terraces are, however, characterised by diverse growing environments, with highly variable, stony soils, variable plant spacing and canopy cover. This limits the effectiveness of sensor-based technologies for efficient agricultural resource management. The objective of this research is to provide guidelines for informed irrigation decision making in mountain terrace orchards. Over the past four years, we have cooperated with four farmers with irrigated fruit trees on traditional dry-stone terraces in the Troodos Mountains of Cyprus. We installed soil moisture sensors at a different number of locations and soil depths, depending on the soil and terrace characteristics. In the stony soils of the apple terraces we installed sensors 3 locations and 2 depths (10 and 30 cm). In the cherry terraces, we installed 12 sensors at 4 locations and 3 depths (10, 30 and 45 cm) and 2 additional sensors at 10 cm. In the deep soils of the nectarine terraces, we installed sensors at 2 locations and 7 depths (10-70-cm depth). In the stony plum terraces, we installed sensors at 2 locations and 3 depths (10, 25 and 50 cm). We analysed the variability of the soil moisture observations and the effect of the uncertainty of the soil moisture observations on irrigation decision making. In the cherry terrace, results of more than 3600 hourly observations for 14 sensors showed that the average difference between the driest and wettest sensors amounted to 11.6% volumetric soil moisture at 10-cm depth, 6.7% at 30-cm depth and 7.6% at 45-cm depth. The maximum difference between the soil moisture sensors was observed immediately after one of the irrigation events (12 May 2022), showing a difference of 187 mm water in the rootzone between the driest and wettest set of sensors at the 3 depths. Based on the depth-weighted average of all 14 sensors, this event showed a drainage loss of approximately 38 mm of the 55-mm applied irrigation, below the 55-cm rootzone. The sensor-based irrigation advice would have suggested that the farmer should have irrigated maximum 17 mm. If the driest set of sensors at the 3 depths would have been used, the irrigation advice would have been the same, whereas based on the the wettest set of 3 sensors, the advice would have been to irrigate maximum 22 mm. This indicates that even though the the irrigation advice can have a 29% error, 33-mm drainage loss could have been saved with sensor-based irrigation scheduling for this event.