Plain Language Summary

Irrigation impacts how land and atmosphere interact, both locally and globally. Therefore, it is important to understand the effects of irrigation practices and improve how water resources are managed. Advanced models such as land surface models now include irrigation. However, missing information on irrigation and other data make it difficult to evaluate and improve irrigation in these models. This study looks at how irrigation and crop yield are represented in the Community Land Model (CLM5). The model was tested using observed data and its own prediction of irrigation, and then compared to observed soil moisture and yield in two apple orchards. The model was able to predict changes in soil moisture caused by irrigation. Simulated irrigation was different in timing and amount from the observed one. This could be improved by adding more details to the irrigation routine. Next, irrigation and crop yield were studied in the entire region. Both were sensitive to changes in climate caused by the diverse landscape. A small reduction in irrigation did not negatively affect yield while halving the irrigation caused it to decrease noticeably. These findings show that land surface models like CLM5 can be useful tools for managing irrigation and water resources.

1 Introduction

Irrigation plays a vital role in sustaining global food production by providing a reliable water supply to agricultural systems, especially in semi-arid or arid regions [McLaughlin and Kinzelbach , 2015]. With a growing global population and increasing food demands, irrigation contributes significantly to ensuring food security by enabling higher crop yields and reducing the vulnerability of agricultural systems to climate change [McDermid et al. , 2023; Mueller et al. , 2012]. On the other hand, poor management of irrigation water has led to the depletion of groundwater resources [Dangar et al. , 2021;Scanlon et al. , 2012; Wada et al. , 2010] and water use conflicts in many regions [Cai et al. , 2003; Eshete et al. , 2020; Gurung et al. , 2006]. Apart from quantitative and qualitative effects on water resources [García-Garizábal et al. , 2012; Y Zhang et al. , 2022], irrigation substantially impacts biogeophysical and biogeochemical processes at the land surface through alteration of the hydrological cycle or energy budget. This has subsequent effects on climate [DeAngelis et al. , 2010;Erb et al. , 2017; Ferguson and Maxwell , 2012; Gordon et al. , 2005; Sacks et al. , 2009]. The multidimensional role of irrigation calls for increased efforts in effective irrigation management and irrigation impact studies using large-scale approaches. This is crucial not only to meet food demands and mitigate future increases in climate change induced water stress, but also to understand its interactions and feedback mechanisms within the Earth system [Elliott et al. , 2014; McDermid et al. , 2023].
Modeling can be a powerful tool to simulate complex interactions in agricultural systems, evaluate different irrigation and climate scenarios, and provide decision support for water resources management [Blyth et al. , 2021; Pongratz et al. , 2018]. This necessitates comprehensive modeling frameworks that combine field-scale representations of crop growth and irrigation with a more holistic assessment of the impacts of irrigated agriculture on water resources and climate at larger scale [Bin Peng et al. , 2020]. Process-based crop models include a range of crop parameterizations that provide a unique way to study crop growth processes in response to irrigation practices by using physical and biological principles. However, their main purpose is to simulate yield at the field scale, often over a single growing season, while lacking the interface with the land surface, soil, and climate [Cheng et al. , 2020]. Land surface models (LSMs), on the other hand, provide a more holistic representation of the land-atmosphere interactions to capture the feedback mechanisms between irrigation, vegetation, hydrological processes, and climatic conditions beyond the field scale [Blyth et al. , 2021]. Conversely, they often lack more detailed physiological and genetic representations of crops and irrigation management [Lombardozzi et al. , 2020; B Peng et al. , 2018]. This limits the ability of LSMs to reliably simulate yield and irrigation water withdrawals leading to poor model performance and biases in related processes such as carbon, energy, and water fluxes over intensively irrigated regions [Leng et al. , 2015;Lombardozzi et al. , 2020; Ozdogan et al. , 2010; Z Zhang et al. , 2020].
In recognition of the important role of human land management, efforts to advance the representation of crops and irrigation in LSMs are ongoing [Pokhrel et al. , 2016]. Various land surface models such as ORCHIDEE, the Community Land Model (CLM), and Noah-MP have since added crop modules [Levis et al. , 2012; Liu et al. , 2016; Smith et al. , 2010]. New crop representations have been developed to improve crop growth and management processes [Boas et al. , 2021; B Peng et al. , 2018] or to add new crop types [Olga Dombrowski et al. , 2022; Fader et al. , 2015;Fan et al. , 2015]. Rather simple irrigation schemes are generally incorporated based on soil moisture thresholds [de Vrese et al. , 2016; Ozdogan et al. , 2010; Sacks et al. , 2009], while more recent developments include the integration of irrigation techniques [Leng et al. , 2017; Yao et al. , 2022], irrigation water withdrawal from different sources [Leng et al. , 2017; Xia et al. , 2022], and water availability limitation [Yin et al. , 2020]. These studies, however, were performed at river basin, county, or global level with coarse spatial resolutions between 10 and 100 km. Simulated irrigation was validated against rather uncertain statistics like total yearly irrigation water withdrawals, without considering specific irrigation practices. Crop and irrigation data at higher spatial (<5 km) and temporal (e.g. daily or sub-seasonal) resolution is needed to evaluate the representation of local irrigation schedules in LSMs and support irrigation management decisions. However, data to reliably constrain and further develop implemented irrigation schemes is often lacking, e.g. irrigation amount and timing along with continuous soil moisture (SM) observations [Lawston et al. , 2017].Lawston et al. [2017] first evaluated the sprinkler irrigation scheme of the NASA Land Information System LSM with point and gridded SM observations at 1 km resolution. While the model could not capture the field scale heterogeneity and overestimated irrigation amounts, it captured well the seasonal variability and regional average SM dynamics. The authors did however use a prescribed crop phenology (green vegetation fraction) and did not examine the effect of irrigation on crop yield. A recent study examined the effect of different irrigation setups on maize yield and two water use efficiency definitions using the dynamic crop and irrigation scheme of the Noah-MP LSM [Huang et al. , 2022]. They found that modeled crop yield was sensitive to irrigation quantity and timing (in which crop growth stage irrigation was applied) and based on these results recommended an optimal SM threshold to trigger irrigation. While the authors lacked data to accurately assess the irrigation amount and crop yield, their work presents a first use of a LSM to study the effects of deficit irrigation on crop growth, yield and water use efficiency.
The work presented here builds upon previous studies to continue the evaluation and improvement of irrigation representations in LSMs combining local irrigation, SM, and yield observations. In particular, this study applies CLM version 5, with a recent extension to represent deciduous fruit trees, to model irrigation and crop growth in a Mediterranean catchment. Specifically, we aim to: (1) evaluate the existing irrigation scheme of CLM5 and enhance its flexibility to account for local irrigation management practices; (2) assess whether the model can reproduce soil moisture dynamics and crop growth in irrigated apple orchards using the enhanced model capability; (3) examine the potential to improve regional irrigation management by modeling the effect of different irrigation scenarios on crop yield and water use efficiency at the catchment scale.

2 Materials and Methods

2.1 Study Area

Located in central Greece, the Pinios Hydrologic Observatory (PHO) covers an area of approximately 45 km2 (Figure 1). The PHO was established in 2015 to study the Pinios catchment hydrological processes and, ultimately, to support local authorities in the sustainable management of water resources [V Pisinaras et al. , 2018]. It is characterized by a Mediterranean climate with an annual precipitation of 500 to 1200 mm, and highest precipitation amounts in the winter months, annual potential evapotranspiration of approximately 1100 mm, and annual average air temperature of 15 °C [Bogena et al. , 2018]. The area displays a range of altitudes from 1500 m in the northern part down to less than 200 m in the plain. The mountainous part of the catchment features steep slopes and is covered by forests, mixed with shrubs and grassland, while the southern plain is primarily characterized by agriculture and small villages. In the plain, sandy loam soils dominate while sandy clay loam and loamy soils also occur [V Pisinaras et al. , 2018]. The PHO is located in one of the most productive agricultural areas in Greece owing, among other factors, to widespread irrigation practices that account for over 85 % of the local freshwater consumption [Panagopoulos et al. , 2018]. The main cultivation are apple and cherry orchards (i.e., ~78 % of agricultural area) that are irrigated between May and October. There are a few other rainfed fruit and nut tree orchards in the area with < 5 % coverage. Annual crops including corn, cereal (mainly winter wheat), and potato are grown on the remaining agricultural land. They are partially irrigated, depending on precipitation occurrence, but cover a negligible part of the total irrigated area. Irrigation in the orchards is typically applied through micro sprinklers and the demand is almost entirely met by abstraction from the alluvial groundwater system through water wells, most of which are privately owned. Overexploitation of groundwater in the area due to poor irrigation management practices, amongst others, has previously been reported by Panagopoulos et al. [2018] andVassilios Pisinaras et al. [2023] resulting in the decline of groundwater levels.
Within the PHO, irrigation management in two irrigated apple orchards, hereafter referred to as S09 and S10, was studied (Figure 1). Both orchards have a size of around 1.2 ha, with a mild southern slope of <5 %. The soil texture is sandy loam and sandy clay loam with a high gravel content (13-29 %) (Table 1) and many larger cobbles (>64 mm according to Wentworth [1922]), especially below 30-50 cm depth. Trees are planted in rows, oriented North-South with 3.3 m distance between rows and an in-line distance of 1.5 m (approximately 2020 trees ha-1). The trees in S09 and S10 were planted in 2013 and 2015 respectively, with a mixture of 3 to 5 different varieties. Trees are pruned to a height of 3.5 m throughout the winter season and residues are mulched back into the soil. Bud burst typically occurs in the second half of March while fruit development starts with the end of flowering in mid to late April. Harvest dates range from late August to mid-November depending on the harvested variety. Major leaf fall starts in late October and continues until mid-November, sometimes until early December. Trees are irrigated with a micro sprinkler system with a maximum flow rate of 60 L hour-1 that is installed below the canopy, halfway between the tree stems of the same row. The irrigation season typically starts in May and continues until October. Orchards are fertilized with 80 kgN ha-1 at the end of flowering in April. Pest and fungicide treatment is applied prior to flowering and after flowering until late June. The grass in the alleys is generally mowed once a month starting in March or April and mowing material is left on the ground. During periods of intense heat, the actively growing grass cover provides a cooling effect to protect the apples from heat damage.