Abiotic Mechanisms Drive Enhanced Evaporative Losses under Urban Oasis Conditions

The oasis effect refers to the impact of advected energy on the surface energy balance leading to enhanced evapotranspiration. In this study, we utilize a 1‐yr record of water, energy, and carbon dioxide (CO2) fluxes to study the occurrence and signature of the oasis effect in an irrigated turf grass of an arid urban region. Days with the oasis effect are selected using readily available air temperature and relative humidity and include excessive heat warnings. During oasis days, higher evaporative cooling is demonstrated throughout the day, especially for late afternoons when it can exceed net radiation. Evaporative enhancements are linked to abiotic mechanisms, such as soil and irrigation water evaporation, since plant productivity is unaltered. Nighttime evaporative losses and CO2 releases are also enhanced during oasis days. Our findings show how the oasis effect impacts the water, carbon, and thermal conditions of urban parks.


Introduction
Irrigation allows the maintenance of vegetation which supports evaporative cooling, significantly reducing air and surface temperatures and enhancing atmospheric water content (e.g., Gober et al., 2010;Myint et al., 2013;Song & Wang, 2015). Sharp thermal and moisture contrasts are often created between irrigated areas and their surrounding landscapes (Chow et al., 2012;Ko et al., 2016). An example of this is the establishment of irrigated turf grasses in urban parks and golf courses which are surrounded by land cover types with less or no irrigation. These sharp contrasts in irrigation have the potential for creating an "oasis effect" whereby advected energy from hotter and drier surrounding areas from all directions can influence the irrigated site (Warner, 2004). A separate impact of irrigated sites on downwind areas is also expected (e.g., Motazedian et al., 2020). Identifying the presence of the oasis effect and the mechanisms leading to its consequences on the surface energy balance is critical in urban parks of arid and semiarid regions since these provide ecosystem services related to heat amelioration (Harlan et al., 2006), biodiversity (Cook & Faeth, 2006), cultural amenities (Park, 2017), aesthetics (Yabiku et al., 2008), and real estate prices (Larson & Perrings, 2013).
Continuous, in situ measurements of the surface energy balance in urban parks are generally lacking, thus limiting our understanding of the effects of irrigation on water, energy, and carbon dioxide (CO 2 ) fluxes. Most prior studies are limited to short periods that do not capture the strong seasonal variations that often exist in radiative forcing, outdoor water use, and turf grass management (e.g., Chow et al., 2011;Colter et al., 2019;Day et al., 2002;Pérez-Ruiz et al., 2020;Sproken-Smith et al., 2000;Templeton et al., 2018). In addition, the atmospheric conditions leading to the oasis effect are not well understood, in particular whether or not readily available data are sufficient for its identification. It is also unclear if a direct relationship exists with excessive heat warnings (EHWs) during which these sites serve as thermal refugia (e.g., Brown et al., 2015). Since urban parks and golf courses often occupy large surface areas in arid and semiarid regions, it is possible that their interactions with the atmosphere can also have regional consequences on the urban heat island (Buyantuyev & Wu, 2010), the CO 2 dome (Koerner & Klopatek, 2002), and the precipitation regime (Shepherd, 2006), among others.
In this study, we address this knowledge gap through the use of the eddy covariance (EC) method to measure water, energy, and CO 2 fluxes at an irrigated urban park in a desert region. EC measurements provide high-temporal resolution data that are useful for quantifying turbulent fluxes and identifying linkages between them that reveal the role played by urban ecosystems (e.g., Baldocchi et al., 2001). The study period allows investigating seasons when irrigation and turf grass management vary considerably, including EHW periods during the North American monsoon (see http://www.weather.gov/psr/HeatSafety). Through ancillary measurements, we aim to understand how the surface energy balance is altered under the oasis effect due to the omnidirectional effects of neighboring urban sites on the golf course. In particular, we address if changes in evapotranspiration driven by the oasis effect are linked to abiotic or biotic mechanisms through an analysis of coincident CO 2 fluxes. van Bavel et al. (1963) argued that enhanced plant transpiration explained these changes but did not benefit from simultaneous water, energy, and CO 2 fluxes for exploring the underlying mechanisms.

Study Location
The study site is located in the northwest corner of the 18-hole Encanto Golf Course in Phoenix, Arizona, USA (33.48°N, −112.10°W, 334 m; Figure 1a). At the site, the Arizona Meteorological Network (AZMET; https://cals.arizona.edu/AZMET/) placed a weather station in 1988 to estimate reference evapotranspiration. We installed additional sensors inside the fenced area of the AZMET station over the period 16 March 2019 to 16 March 2020. Turf management at Encanto Golf Course includes varying treatments in fairways and rough areas, using warm season bermudagrass (Cynodon dactylon) and overseeding with ryegrass (Lolium perenne) to maintain grass cover during the cool season. Soils at the site are well-drained Mohall clay loam that have undergone several decades of soil and water treatments (Soil Survey Staff, 2020). While turf grass is nearly continuous in cover, a few scattered palms are found in rough areas and trees line nearby streets (170 m to west of EC site). The site is immersed in a low-density urban fabric consisting of single-family residences, commercial areas, streets, and parking lots. Local climate is classified as hot and arid (Köppen zone BWh) with an average annual rainfall of 203 mm/yr (AZMET data over the period 1989-2019), with small differences between warm (87 mm, 1 April to 30 September) and cool (116 mm, 1 October to 31 March) seasons. Similar conditions are expected in other urban parks and golf courses that occupy about 32 km 2 or 2.4% of the City of Phoenix area.

Surface Energy Balance Measurements
The EC method consists of an open path-infrared gas analyzer to measure H 2 O and CO 2 concentrations (LI-7500A, Li-COR Biosciences) and a three-dimensional sonic anemometer (CSAT-3, Campbell Scientific) to measure turbulent wind velocities. The EC system was installed at 5 m above the ground and aligned with the dominant wind direction (180°from North) to measure latent heat flux (λET), sensible heat flux (H), and net ecosystem exchange (NEE). A radiometer (CNR4, Kipp & Zonen) was installed at a 4 m height to measure net radiation (R n ), while a heat flux plate (HFP01SC, Hukseflux) was buried at 5 cm depth to estimate ground heat flux (G). Sensor placement and measurement heights were selected to sample turf grass conditions in a small EC footprint obtained using the two-dimensional model of Kljun et al. (2015) at the daily scale and aggregated for each season. Figure 1b illustrates the 50% and 80% source areas during the warm season. Vegetation within the 50% and 80% footprints were classified as grass (98% and 81%), tree (2% and 8%), bare soil (<1% and 4%), and impervious surface (0% and <8%), respectively, based on a 1-m land cover classification obtained in 2010 (Li et al., 2014). Fluxes were calculated at 30-min intervals with EddyPro® 7.0.4. EC processing included a number of standard processing steps (supporting information Text S1). Missing data accounted for 21% of the study period due to maintenance and power issues. The energy balance yielded that 90% of available energy (R n − G) was measured as turbulent fluxes (λET + H), consistent with studies in different ecosystems (Wilson et al., 2002; see Figure S1).

Ancillary Data Sets
Ancillary measurements were installed to complement the AZMET records, including: (1) volumetric soil moisture at 5-, 15-, and 30-cm depths (θ, CS616, Campbell Sci.), (2) soil temperature at 2-and 6-cm depths (TCAV-L, Campbell Sci., averaged as T soil ), (3) land surface temperature (LST) estimated from longwave radiation, following Martin et al. (2019), and (4) midday albedo (a) obtained from the radiometer. We used AZMET data for rainfall (R), air temperature (T air ), vapor pressure deficit (VPD), relative humidity (RH), incoming solar radiation (R s ), wind speed and direction, and reference evapotranspiration (ET o ). Monthly estimates of water use via sprinkler irrigation were provided by the City of Phoenix based on metering. In addition, we used high spatiotemporal resolution data on the normalized difference vegetation index (NDVI) from Planet Labs (2017), to quantify turf grass conditions. PlanetScope products (3 m, daily resolution; Text S2) were obtained from a constellation of >130 active Cubesats in four spectral bands. A total of 201 cloud-free scenes (54.6% of study period) at an overpass time from 9:30 to 11:30 a.m. were converted into a linearly interpolated NDVI series (Chen et al., 2004). Bias correction was then used to adjust NDVI to match coincident values obtained from Landsat 8. PlanetScope has been applied in agricultural areas (Houborg & McCabe, 2018), but its observing capabilities in urban parks and golf courses have not been demonstrated.

Seasonal Variations
Figures 1c and 1d display the annual cycle in soil, vegetation, and meteorological variables during the study period. These were selected to show the water and energy input and the resulting soil and turf grass conditions in the warm and cool seasons. Net radiation (R n ) tracks the seasonal variation in solar irradiance and daily changes linked to rainfall and cloud cover. During the warm season, rainfall (R) was below average, while air temperature (T air ) was above average (8 mm and 29.2°C as compared to 87 mm and 28.8°C from 1989 to 2019), consistent with analyses indicating exceptionally hot and dry conditions (National Weather Service [NWS], 2019). This included 26 days of EHWs between 11 June to 7 September 2019. During the cool season, R was above average (177% of 1989-2019 average), including 4 days with R > 20 mm/day, which has a low probability (<2%; Mascaro, 2018). When compared to irrigation (Figure 1d inset), however, R is a negligible to minor input for the warm and cool seasons (<1% and 36% of total, respectively). Based on turf guidance (Brown et al., 2001), sprinkler irrigation occurs daily in the warm season in the entire golf course, 10.1029/2020GL090123

Geophysical Research Letters
whereas only fairways receive daily irrigation during the cool season and roughs are irrigated once per week. Irrigation increases soil moisture (θ) during the warm season down to 30-cm depth (Figure 1d), which sustains high NDVI (Figure 1c). Decreases in irrigation in the cool season are reflected in lower θ and decreases in NDVI in rough areas, whereas overseeding and irrigation in fairways promotes continued NDVI (see Text S2 and Figure S2).
In addition to wet soil conditions, frequent irrigation promotes evaporative cooling which impacts the relation between daily T air and T soil (Figure 1d). A lower T soil was noted in the turf grass for the warm season as compared to T air (28.4°C and 29.2°C, respectively). This was supported by daily estimates of LST (28.2°C), indicating the air was warmer than the irrigated turf grass during the warm season (see Table S1), in contrast to nonirrigated land covers in Phoenix (Song et al., 2017). We explore this seasonality further in Figures 2a  and 2b by comparing the surface energy fluxes for the warm and cool seasons, shown as diurnal cycles of R n , λET, H, and G (see Text S8 and Table S2). The majority of energy input (R n ) results as evaporative cooling (λET) in the warm and cool seasons (peak λET/R n of 0.70 and 0.47, respectively). Since H and G remain consistent across the two seasons (Figures 2a and 2b), reductions in R n in the cool season or during cloudy days in the warm season lead to lower λET, an indication of an energy limitation in the turf grass. It is also noteworthy that λET is positive at night during the warm season (average of 59 W/m 2 ), while H is negative (average of −43 W/m 2 ), implying that air warms the turf grass during the night (Table S2). This nighttime behavior of λET and H, in particular during the warm season, are characteristic of the oasis effect (Warner, 2004). Only a small amount of energy is partitioned to deeper soil layers (low magnitude of G), consistent with the relation between T air and T soil .

Geophysical Research Letters
VIVONI ET AL.
The influence of irrigation extends to both the warm and cool seasons, such that surface energy fluxes are dominated by λET. Nevertheless, large variations are noted in the turf grass conditions, in particular in the relation between NDVI and midday albedo (Figure 2b inset). In the cool season, rough areas decrease in NDVI and increase in a, which reduces energy inputs via the effect of albedo on R n . Seasonal differences are also apparent in CO 2 fluxes which are compared in Figures 2c and 2d as diurnal cycles of NEE and its two components: gross primary productivity (GPP) and ecosystem respiration (R eco ; Text S9 and Table S4). Irrigated turf grass exhibits diurnal behavior that reflects photosynthetic processes, with daytime CO 2 uptake (NEE < 0) and nighttime CO 2 releases (NEE > 0). As expected, the warm season has higher GPP and R eco , leading to higher CO 2 uptake, as turf grasses receive more irrigation and energy input, exhibit greater NDVI and lower a, and grow in an environment of higher T air and T soil . Warm season CO 2 uptake during the day is considerably higher than reported values from irrigated turf grasses at other sites (Pahari et al., 2018), suggesting that biological processes are at maximum capacities, in particular for the warm season. In both seasons, a negligible impact of anthropogenic emissions, such as traffic, is noted on CO 2 fluxes, in contrast to other land covers in Phoenix (Pérez-Ruiz et al., 2020).

Oasis Effect
Due to its higher irrigation and energy input, we focus on identifying if there is an oasis effect during the warm season. Figures 3a and 3b present the daily variation of ET and two primary controlling factors, R s and VPD, as obtained through a linear regression analysis (Text S4 and Table S3). Wind speed or direction were not significant controls, indicating that advected energy is omnidirectional. As expected, variations in ET reflect the seasonality of solar radiation and daily changes in R s and VPD related to the North American monsoon (Vivoni et al., 2008). Daily ET from the EC method exhibited a strong relation with evapotranspiration estimates obtained as K c ET o (Figure 3a inset), where K c is a monthly varying crop coefficient (Brown et al., 2001;Text S5). This provided confidence in the use of the long-term ET o estimates at the AZMET station (2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018) to determine days with an oasis effect (yellow circles in Figure 3). To generalize the method for other sites, we utilized T air and RH as proxies for R s and VPD. We identified the T air and RH conditions during warm season days that exceeded the 90% quantile in K c ET o in the 6,352-day record. This yielded a threshold of T air that if exceeded at a particular RH indicates a day with the oasis effect, in a manner similar to the use of a heat index (Text S6 and Figure S4). Note that oasis days represent 36% of the warm season in Figure 3, including all of the EHW days in 2019, as compared to 21% in the long-term records, an indication of the exceptionally hot and dry conditions. Furthermore, the selection of oasis days generally results in high ET, R s , and VPD, as compared to the long-term averages over 2003-2018 (Figures 3a and 3b).
Oasis days represent high-temperature and low-humidity settings leading to exceptionally large daily ET. Note that oasis days have average daily wind speeds and profile-averaged soil moisture values (1.62 m/s and 0.59 m 3 /m 3 ) that are similar to nonoasis days (1.62 m/s and 0.55 m 3 /m 3 ) during the warm season. Even the wind direction has a limited control on daily ET (Figure 3b inset), with days with winds from the west (240°to 300°) exhibiting similar values (6.62 ± 0.71 mm/day) to days from all other directions (6.28 ± 1.13 mm/day), despite having a residential area to the west. Given the high CO 2 sequestration potential of irrigated turf grasses, it is surprising that oasis days with high ET are also typically periods of CO 2 release to the atmosphere (average daily NEE of 3.78 g CO 2 m −2 day −1 ; Figure 3c inset), due to higher nighttime R eco as compared to daytime GPP (see Table S4). Nonoasis days are characterized by neutral CO 2 conditions (NEE near zero) such that a balance exists between plant productivity and respiration from soils, turf grasses, and lawn residues. Inherent water use efficiency (IWUE; Text S7) indicates that sustained plant productivity per unit of evaporative loss occurs throughout the warm season such that plants are at their maximum capacity and are not influenced by the advected energy during oasis days. These results are in contrast to studies in other climates showing that irrigated turf grasses primarily absorb CO 2 (Pahari et al., 2018). Overall, this points to the oasis effect having important outcomes for CO 2 releases from urban parks in desert cities.
Features of the oasis effect are shown in Figures 4a and 4b through a comparison of the diurnal cycles of R n , λET, H, G, and NEE averaged over oasis (n = 65) and nonoasis (n = 118) days in the warm season (see Text S6 and Table S2). For reference, conditions during EHW days (n = 26) are also shown as dashed lines. As first reported by Sproken-Smith et al. (2000), the oasis effects in an urban park is characterized by a high daytime 10.1029/2020GL090123

Geophysical Research Letters
VIVONI ET AL. and nighttime λET. During oasis days, the ratio of peak λET/R n rises from 0.62 to 0.80 (28% higher), while the peak ratio of H/R n is reduced from 0.18 to 0.10 (41% lower). Another distinguishing feature is the late afternoon period when λET can exceed R n as additional energy is input via a negative H (average H of −31 W/m 2 over 5:00 to 9:00 p.m.). Energy input during oasis days is likely affected by advection from surrounding urban areas in all directions that are hotter and drier, leading to a higher R n , a more negative H, and an increase in λET (average peak differences of +56, −40, and +142 W/m 2 , respectively). This is consistent with Sproken-Smith et al. (2000), who first performed measurements in an urban park and its surroundings. The authors also indicated that separately accounting for advected energy is not feasible

Geophysical Research Letters
VIVONI ET AL.
during oasis conditions since it has a direct impact on all other fluxes. Interestingly, only a minor effect is noted on NEE during oasis days, with a small increase in NEE observed at night due to an increase in R eco under the warmer temperatures of oasis days. This suggests that turf grass transpiration does not adjust to the additional energy input due to the oasis effect as plants are at their maximum capacity, as confirmed by no daytime change in NEE and GPP. As a result, an abiotic process should be responsible for the higher daytime λET. Chow et al. (2014) also noted some of the oasis effect features within an urban area of Phoenix but did not attribute the observed increase in λET to specific evaporative processes.
An inspection of representative oasis days yields insights on the factors influencing the high evaporative loss. Figures 4 c and 4d show the diurnal cycles of R n , λET, H, G, and NEE for two oasis days (5 and 30 August 2019) that were also identified as EHW days. These days recorded no rainfall and capture differing wind directions (WD of 43 and 241°from North) that sample the variability in contributions from neighboring areas during the warm season. To provide context, normalized soil moisture in the shallow 5 cm sensor and maximum wind speed are shown (see Text S10). Note the late afternoon periods when is λET greater than R n due to the energy input from H. In contrast to this consistent feature of the oasis effect (Warner, 2004), some oasis days also exhibit large increases in λET during the midafternoon as shown in Figure 4d. These are attributed

Geophysical Research Letters
VIVONI ET AL.
to large increases in wind speed, in particular when wind directions are from the west (240 to 300°), and are coincident with midafternoon decreases in H, but no discernable changes in NEE. We attribute these short-lived increases to a more localized impact of wind gusts from the residential area to the west. Interesting behavior occurs in response to irrigation as depicted through a delayed rise in the shallow soil moisture. At night, irrigation via sprinkler application promotes a short pulse in λET that is associated with an increased CO 2 release (higher R eco ). This suggests that nighttime irrigation during oasis days is the primary reason for turf grass to be a net CO 2 source. Note that nighttime λET and R eco account for 36% and 52% of daily values during oasis conditions (Tables S2 and S4). Furthermore, increased λET is not linked to plant productivity such that additional energy is partitioned through abiotic mechanisms such as evaporation from direct sprinkler water, intercepted water on turf grasses, and soils that are wet down to 30 cm.

Concluding Remarks
We identify that the oasis effect is a persistent warm season feature of an irrigated turf grass which is closely linked to periods of high air temperature and low humidity associated with EHWs. While previously thought to occur, direct evidence has not been documented at the level of detail provided here. The large distance of the site to the golf course edge suggests that advected energy is input from local surroundings in all directions rather than being an edge effect (Sproken-Smith et al., 2000). Nonetheless, certain oasis days have short periods with high winds from a neighborhood to the west that briefly elevate latent heat flux. It is important to note that the oasis effect has impacts on both daytime and nighttime conditions (see Text S11). During short periods at midday and late afternoons, latent heat flux can exceed net radiation since advected energy and irrigated conditions lead to downward sensible heat flux. At night, latent heat flux increases in response to higher amounts of advected energy and often has short pulses occurring after irrigation. These features are muted or absent from days that do meet the oasis conditions based on daily air temperature and RH data.
A summary of the CO 2 budget offers novel insights on the oasis effect (see Text S11). Daytime plant productivity is unaffected by advected energy, indicating that latent heat flux increases are not related to higher turf grass transpiration. This contradicts van Bavel et al. (1963) who attributed higher evaporative losses under advected energy to an increase in grass transpiration. Instead, latent heat fluxes are higher due to the increased evaporation from soils and direct evaporation of irrigation and from intercepted water on the turf itself. Interestingly, nighttime evaporative losses occur under conditions of CO 2 release. While evaporative losses during short pulses are associated to sprinkler water application, more persistent latent heat fluxes are linked to soil evaporation and possibly to nighttime transpiration identified to occur in warm season grasses (O'Keefe & Nippert, 2018). Additional CO 2 releases during the oasis effect are likely due to higher soil and turf grass respiration upon soil wetting from irrigation.
The existence of a warm season oasis effect in arid and semiarid regions has important implications for the urban heat island, CO 2 emissions, and heat-related health hazards. Our findings suggest that the oasis effect can be identified for other parks using readily available data. When EHWs are issued, for instance, the oasis effect provides enhanced evaporative cooling in irrigated areas and their downwind locations. This enhances the suitability of urban parks as a heat mitigation strategy (Zhang et al., 2017). Furthermore, the oasis effect is anticipated to become more prevalent during the warm season under the combined effects of urbanization and climate change which are increasing EHWs.

Data Availability Statement
Data sets for the study period are available at Zenodo (Kindler et al., 2020).