Water use of mature pecan orchards under semi-arid conditions

Abstract

Judicious and sensible water supply is vital for optimal fruit production, and as a result most orchard crops depend on supplemental irrigation, especially in areas in South Africa, where rainfall patterns are unpredictable and sparsely distributed. Through accurate quantification or estimation of crop water use or evapotranspiration (ET), the need for supplemental irrigation can be quantified. In addition, by partitioning ET into its components, a better understanding of the factors that govern water loss from an orchard can be obtained, which is critical for determining where water savings can be made. This study aimed to measure ET and its components (canopy transpiration (Tc) and soil evaporation (Es)) of a 14-year-old mixed cultivar pecan orchard in the semi-arid Northern Cape Province of South Africa. This is one of the hotter and drier pecan production regions in South Africa and was expected to differ from where most of the pecan water use research was conducted in the United States of America (U.S.A), due mainly to a longer growing season in the Northern Cape. The current data used for water management of pecan orchards are primarily based on research done in other countries or by using an empirical model to estimate water use. As different regions are characterized by its own unique climate and management practices, modeling approaches were developed that adjust pecan crop coefficient curves (Kc) to specific climatic conditions and managements practices through weather variables, thermal time, fractional canopy cover and crop height (Allen and Pereira, 2009; Miyamoto, 1983; Samani et al., 2011; Sammis et al., 2004; Taylor et al., 2015). These empirical models may not be applicable to South African growing conditions as they contain artefacts of the regions from where they were developed, potentially leading to inaccurate ET predictions (Ibraimo, 2018). In the study by Ibraimo (2018), it was highlighted that modelling pecan ET according to a four stage Kc approach (Allen et al., 1998b) yielded accurate results on a seasonal basis, but not at a monthly time step, mainly because pecan exhibit a six stage Kc curve. A second approach was tested by Ibraimo (2018), whereby a set of reference Kc were adjusted according for canopy size and growing degree days (GDD) to derive orchard specific Kc (Samani et al., 2011; Sammis, 2004). The ET estimates correlated well with actual measurements at the study site in Cullinan, South Africa, but it was further hypothesized that the method of adjusting Kc values for climate would not be transferable to hotter production areas where GDD exceeds 1500. Ibraimo (2018) [proposed that a better method could be to adjust Kc curve according to observed phenological stages and that the approach would work better in orchards whereby Es is a minor component (≤20%) of ET. By measuring the two ET components separately it is possible that this approach could be applied to a wider range of orchards, which would allow for improved estimation, as well as the contribution of Es towards total ET. Field trials were conducted over the 2018/2019 production season on a farm in the Vaalharts irrigation scheme to measure Tc and model Es separately, which was then used to obtain seasonal ET values. From the results it was observed that the application of the empirical equation of Sammis et al., (2004) for adjusting Kc values according to thermal time does not hold true in Vaalharts that has a GDD accumulation exceeding 1500 (1861 for the 2018/19 season) during the growing season. The approach proposed by Ibraimo (2018), whereby the Kc curve was adjusted according to phenological stages allowed for more accurate estimations of Kc. The method was shown to successfully estimate monthly ET of mature pecan trees in this study, when the adjusted Kc-ref values were further adjusted for canopy size as described by Samani et al., (2011). There was a slight overestimation by the model between November 2018 and January 2019, ultimately accounting for a 6% overestimation of estimated ET as compared to ET estimated as the sum of Tc and modelled Es. The performance of the model was determined by comparing the accuracy of monthly ET modelling against determined monthly ET. From this comparison the coefficient of determination (R2) value was 0.86, which is considerable to be acceptable. The Willmott index of agreement (D) value was 0.91, root mean square error (RMSE) 23.22, mean absolute error (MAE) of 13.60 and coefficient of residual mass (CRM) of 1.01. The MAE is below the threshold of 20% which indicates that the slight deviation is still within acceptable limits. Based on the positive CRM value the deviation is attributed to an overestimation of the model. This data suggests that by allowing for the adjustment of the Kc-ref curve according to local growing conditions and canopy cover, good estimates of monthly ET can be obtained. Through this method it was possible to determine the main contributing factors that drive water loss, through both Tc or Es, as well as some of the factors driving the water loss.