The research leading to these results has received funding from Ministerio de Economía y Competitividad (Grant Nos. AGL-2010-20766 and AGL2015-69822), from Junta de Andalucía (Grant No. P08-AGR-04202), from the European Community’s Seven Framework Programme-FP7 (KBBE.2013.1.4-09) under Grant Agreement No. 613817. 2013–2016 “MODelling vegetation response to EXTREMe Events” (MODEXTREME, modextreme.org) and from ERA-NET FACCE SURPLUS (Grant No. 652615, project OLIVE-MIRACLE), the latter co-funded by INIA (PCIN-2015-259). Besides, ÁL-B was funded by a postdoctoral fellowship (‘Juan de la Cierva-Formación 2015’ Programme, FJCI-2015-24109) from Ministerio de Economía y Competitividad. Conflict of Interest Statement The measurements (only performed for the central trees of the replicates) used for testing the model were oil yield ( Y oil, g m -2), seasonal ET and daily transpiration ( E p, mm d -1). With regard to the former, each tree was manually harvested and the fresh yield weighed in the field. Y oil was subsequently determined from sub-samples of 5 kg of fresh fruits. Cumulative ET was determined by water balance for the whole 2005 and 2006 seasons by measuring soil water content with a neutron probe (model 503, Campbell Pacific Nuclear Corp, Pacheco, CA, United States). Eight access tubes were installed between two trees per replicate, normal to tree rows. Measurements were taken at several depths (from 0.075 to 2.65 m deep). Finally, E p was measured in 2006 with a sap-flow system device developed and assembled in the IAS-CSIC in Córdoba and described by Testi and Villalobos (2009). The system uses the Compensated Heat Pulse (CHP) method in combination with the Calibrated Average Gradient (CAG) procedure. The probes performed readings every 15 min at 4 depths in the xylem, spaced 10 mm. Six RDI, six CDI and four CON trees were instrumented with two probes per tree, at a height of 30 cm. The outputs of each probe were integrated first along the trunk radius and then around the azimuth angle. Average sap flow records for each treatment were calibrated against the estimates of E p deduced from the difference between the measured ET and soil evaporation in a period of several weeks with no rainfall events during the summer. The model of Orgaz et al. (2006) was used to calculate soil evaporation. The calibrated sap flow data have not been published so far.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. Supplementary Material Control irrigation (CON), which applied the required water to match the maximum ET, discounting rainfall. The maximum ET was estimated using the model of Orgaz et al. (2006). Daily effective precipitation ( P eff) is calculated by discounting rainfall interception by the canopy ( P int) from total daily precipitation ( P). P int is calculated using a simplified version of the model of Gómez et al. (2001) and the resulting P eff is distributed proportionally between the two soil zones as a function of the surface fractions that remain rainfed or are wetted by localized irrigation. With regard to P int, the canopy is treated as a capacitor capable of storing rain water up to a certain limit determined by canopy dimensions and leaf area index ( LAI), according to Gómez et al. (2001). The stored water is subsequently lost by direct evaporation, which is simulated based on the Penman–Monteith equation assuming a null canopy resistance. As in Testi et al. (2006), the aerodynamic resistance is deduced from the model proposed by Raupach (1994), parametrized and validated specifically for olive orchards following Verhoef et al. (1997). The direct evaporation from wet foliage prevents tree transpiration ( E p), until the intercepted water is totally lost.Values of GC, LAD, and R zx required to initialize the model were taken from dedicated measurements. A record of Y dry of the year preceding simulations was also considered. Initial L v values were taken from records measured by Moriana (2001). Statistical Analysis Overall, the results of all the aforementioned comparisons suggest that model performance is fairly satisfactory. However, further testing against experimental data taken from different environmental conditions and orchard characteristics seems highly desirable. This would help to provide additional evidence on the predictive power of OliveCan, or else to identify situations for which model accuracy could be improved through either better calibrations or reformulation of some routines. Apart from that, it should be noted that the reliability of OliveCan for estimating certain output parameters (e.g., NEE, RESP H) has not been tested specifically in the present study, which should also be the focus of future research efforts. Model Applicability de Ciencias e Engenharia de Biossistemas, Instituto Superior de Agronomia, Universidade de Lisboa, Lisbon, Portugal Considering its mechanistic approach, the vast quantity of simulated processes and its potential uses, OliveCan represents a momentous step forward in relation to previous olive growth simulation models. In this regard, OliveCan enables one to assess the combined effects of management operations and weather over crop performance for different olive orchard and soil typologies both under unstressed and water deficit conditions. Thus, the model shows potential for a broad range of research applications. For instance, OliveCan seems particularly suitable for assessing the performance of olive orchards under future climatic scenarios, as the model explicitly accounts for the multiple effects of reduced rainfall and increased environmental CO 2 and temperature for the water and carbon balances of the orchard and the development of trees.

During the development of the model, it became apparent that our current understanding of some of the physiological processes to be simulated was limited. For example, timing of vegetative bud break, dynamics of leaf senescence, fruit photosynthesis and the use of reserves are among the phenomena that have received less attention in the literature. Also, OliveCan is missing a sub-model aimed to properly simulate the dynamics of oil accumulation during the fruit growth period. Further research on these and other topics (e.g., alternate bearing) are clearly needed and might result in model improvements through either a more consistent parametrization or the formulation of better equations for simulating such processes. of Agronomy, Institute for Sustainable Agriculture, Spanish National Research Council (CSIC), Córdoba, Spain Make fried blue cheese stuffed olives and serve with your favorite dipping sauce (might I suggest this sriracha dipping sauce). Finally, future improvements of OliveCan might include additional sub-models for simulating nutrient uptake and the impact of pests and diseases. Apart from that, the model shows potential for being adapted to other tree species, so its interest may not be only restricted to olive researchers. Conclusion The model presented here targets the simulation of the interactions between olive trees and their environment through a detailed characterization of the water and carbon balances of the orchard as affected by weather variables, soil attributes and management operations. The generally high level of agreement found between measured and simulated data evidence the suitability of OliveCan for estimating olive orchard dynamics. These results encourage the application of the model to simulate the growth, carbon exchange and water relations of olive orchards in a wide range of research contexts, including studies on the performance of olive trees under climate change scenarios. The development of OliveCan has also highlighted significant knowledge gaps in relation to some physiological processes and the cultivar specificity of some of the parameters. Further research on these aspects may contribute to improve the reliability of the model. Author ContributionsAll authors played a significant role in the conception and development of the model. FV led out the coding, with contributions from ÁL-B, AM, OG-T, and LT. ÁL-B, LT, and FV gathered the datasets for testing the model. ÁL-B led out the writing with significant contributions from all co-authors. Funding Obviously, the comprehensive nature and the wide range of simulated processes come at the expense of both model complexity and high input requirements. The latter is likely to be its main limitation, as far as some of the inputs (e.g., soil depth, L v distribution) are not easy to measure in the field. In any case, it is noteworthy to emphasize that OliveCan has not been primarily conceived as a decision support system for farmers, but as a research tool. Further Research During the vegetative rest period and provided that fruits are not present, all the available assimilates after discounting maintenance respiration are allocated to a virtual pool of reserves. Such reserve pool is subsequently used for the growth of vegetative organs and fruits during the growth season. Fruit growth can either be source-limited or sink-limited. In the former case, the associated partitioning coefficient is fixed whereas in the latter, it is calculated as a function of the number of fruits ( FN), which in turn is modeled as a function of the number of fruits and nodes produced in the previous year. In doing so, the model may be prone to errors in the estimates of productivity and vegetative growth for a given year when performing long runs, but such errors are to be compensated if those model outputs are averaged over biennia. With regard to the vegetative organs, fixed partitioning coefficients are adopted. Whenever fruits are present, the model considers that they become the prioritary sink of assimilates, thus the vegetative partitioning coefficients are applied after discounting the fruit demand from the daily pool of assimilates. Therefore, partitioning coefficients to vegetative organs are assumed to be independent of tree size, management factors and environmental conditions, as in the model of Morales et al. (2016). As a final remark, inspired by the CERES-type models ( Jones and Kiniry, 1986), the growth of fine roots is distributed among the different layers in the two soil zones as a function of the size and water content of each soil compartment. Variables related to canopy characteristics such as leaf area index ( LAI) or GC are updated from the estimates of biomass of leaves assuming that the crowns present an spheroidal shape with constant leaf area density ( LAD) and ratio of vertical to horizontal canopy radiuses ( R zx). Similarly, the biomass of fine roots in each soil compartment is used to compute root length density ( L v) by adopting a constant specific root length ( SRL).