**4. Discussion**

### *4.1. Crop Yields*

The yields of fresh tomato fruits were similar across schemes but varied remarkably between fields owing to differences in field-level agronomic and irrigation practices and constraints. For instance, the tomato root disease in the VIS field impacted negatively on yield, as the application of insecticides protected only the aboveground biomass. Over-irrigated tomato in 2015–2016 showed high fresh yields and the higher water content of these fruits (i.e., 95% vs. 93% in water-scarce 2014–2015). The values of tomato fresh yield corresponded to the upper ranges measured by [35], reporting 20–36.8 Mg ha−<sup>1</sup> of fresh tomato yields in the UER, and were greater than the 18 Mg ha−<sup>1</sup> reported by [49] for rainfed tomato production in the Ashanti and Brong Ahafo regions of Ghana, reflecting the positive impact of irrigation. However, the lower HI of tomato observed in our study (0.21–0.3), compared to values (0.5–0.65) reported by [22] for rainfed tomato in drylands, could be partly due to over-irrigation. The excessive water use was reflected in the low field application efficiencies (30–59%) characteristic of almost all the examined fields [19].

Late planting and waterlogging, due to the lack of drainage facilities, reduced maize grain yield to only 1.2 Mg ha−<sup>1</sup> in the affected field in the VIS. This observation confirms late planting as one of the causes of sub-optimal yield levels of rainfed maize as rainfall typically declines towards the end of the rainy season. Sallah et al. [50] reported a 30% loss in maize yields due to late planting in northern Ghana. The observed range of maize grain yields in our study (1.2–2.9 Mg ha−1) was similar to that reported for the fertilized 'Obatanpa' maize variety in Ghana (1.3 to 2.7 Mg ha−1; [38]). The values of [8] for SSA (1.3–1.4 Mg ha<sup>−</sup>1) were within the lower range of our results. However, Sugri et al. [51] reported the yield potential of the 'Obatanpa' maize variety to be 5.5 Mg ha−<sup>1</sup> in Ghana. Variations in practices of soil nutrient management and often insufficient applications of fertilizer in the examined fields could also have contributed to the variability in yields. Folberth et al. [9] emphasized that even modest additions of N and P fertilizer might double maize production in most of SSA.

## *4.2. Irrigation Practice*

The examination of field-level irrigation practices during the dry season revealed inappropriate, and in turn, ineffective water application for crop production resulting in over-irrigation in both schemes, mainly due to the lack of consideration of the crop growth stages and water storage characteristics of the soil. Over-irrigation was signified by the high GIA in the water-abundant 2015–2016 season, when farmers in both schemes used more water by shortening irrigation intervals (Table 4), leading to lower water productivity than in the previous, water-scarce season. Because of the lack of appropriate irrigation scheduling and the absence of flow measuring devices in the canals, farmers applied as much water as possible to the tomato crop, and further increased the water application rate with increasing water availability in the reservoir. Faulkner et al. [33] also observed the tendency for excessive water use in response to increasing water availability and attributed this phenomenon to the lack of knowledge of efficient and effective water application at field level. Moreover, the GIAs of tomato in our study were 100–400% larger than the range of values, 274 and 852 mm, previously reported for the UER [34,35], confirming the need for water saving.

## *4.3. Improved Irrigation Scheduling*

The need to adjust irrigation schedules to local hydro-geological conditions is suggested by the modelling analysis, for example, a significant contribution of capillary rise from the groundwater was shown to satisfy the NIR of crops. The groundwater contribution to the NIR of the study crops was highly variable, reflecting the spatial variability in hydro-geological characteristics of the cropping fields. The need to account for this variability complicates the development and application of improved farmer irrigation scheduling in the UER. According to [52], there could be varying contributions of shallow groundwater (≤3 m) to the root-zone soil moisture in fine-textured soils such as those mostly found in the Bongo and Vea irrigation schemes.

The observed increases in tomato yield (i.e., 4–14%) under the improved irrigation schedule most likely resulted from the reduction of the negative effect of over-irrigation on crop yield, as the over-irrigated cropping fields showed the highest potential (14%) to increase yields under the improved irrigation schedule. The simulated magnitude of water saving in the reservoir-based irrigation schemes, which was 22–52% of the GIA under the current irrigation practices, indicates that improving irrigation schedules offers considerable potential for water saving in the dry season in the UER irrigation systems. Overall, however, the improvement of field-level irrigation scheduling alone might not be sufficient for optimizing water productivity and availability in the schemes [16]. To achieve full benefits, equipping irrigation infrastructure with discharge-measuring and dosage structures, and reparation of the decaying water conveyance and distribution sub-systems in the UER would be necessary [18–20]. These interventions to upgrade infrastructure would need to be accompanied by the training of irrigators to handle these facilities, and by further development of water management institutions towards reliably implementing advanced irrigation schedules in order to utilize the full potential of improvements.
