Irrigating urban green space for cooling benefits: the mechanisms and management considerations

Lobell et al (2009) modelled the impacts of irrigation on climate in eight major irrigated regions in the world using the Community Atmosphere Model 3.3. They set the soil moisture to 40% (fraction of saturation point) every half an hour if the soil moisture dropped below 40%. The soil moisture in all eight regions increased except northeast China, which had a high initial soil moisture. The increase in soil moisture varied between 2.3% (fraction of saturation point) in Indo-Gangetic Plains to 20.7% (fraction of saturation point) in Aral Sea Basin. The differences reflected the variations in soil type and rainfall regime (consequently initial soil moisture) between the regions; a more significant change was detected in drier regions. Yang and Wang (2015) modelled the impact of irrigation on the climate in mesic residential landscapes in Phoenix, USA. They applied 1.4 mm of irrigation to the top soil layer whenever the soil moisture dropped below 24% (v/v). The mean soil moisture increased from 10.7% (v/v) without irrigation to 27.6% (v/v) with irrigation. Other modelling studies have also predicted an increase in mean soil moisture with irrigation, ranging from 1.4% (v/v) (Harding and Snyder 2012, Zou et al 2014), 11.1% (v/v) (Kanamaru and Kanamitsu 2008) to 17.1% (v/v) (Gao et al 2020). None of the four observational or experimental studies identified in this review has reported the impacts of irrigation on soil moisture. It is difficult to make a meaningful comparison between studies because changes in soil moisture from irrigation is dependent on multiple factors, such as the interception loss from the vegetation canopy (in the case of sprinkler irrigation), soil type, initial soil moisture, atmospheric demand for evapotranspiration and irrigation amount. The differences in the impact of irrigation upon soil moisture reflect the variation in soil moisture conditions before irrigation, and unsurprisingly there are more significant changes (increase) in soil moisture in drier regions. This is very much related to the underlying soil type of a location and local rainfall inputs and potential evaporative outputs.

All reviewed studies reported a net storage of heat into the ground, i.e. a positive G (Kanamaru and Kanamitsu 2008, Vahmani and Ban-Weiss 2016, Yang et al 2016 and Wang et al 2019). The common findings of these studies were that irrigation did not reverse the direction of mean G over the study period and the changes in magnitude were usually small (<5 W m⁻²). However, it is still helpful to analyse the impacts of irrigation separately on daytime and night-time G, as measures of daily mean G can hide significant and dynamic changes in the diurnal pattern of G. In two modelling studies (Kanamaru and Kanamitsu 2008, Vahmani and Ban-Weiss 2016), it was predicted that irrigation would increase soil thermal conductivity because of an increase in soil moisture, which thereby would increase storage (positive G) during the day and subsequent release (negative G) of that heat at night. Vahmani and Ban-Weiss (2016) also predicted from their model that this would lead to a significant increase in night-time T_a, resulting in a net increase in daily mean T_a despite the important reduction in daytime T_a. Kanamaru and Kanamitsu (2008) also predicted from their modelling a similar diurnal variation in G, but they predicted a small reduction in daily mean T_a. The uncertainties in the predictions of G were attributed to the lack of detailed observational data regarding the response of soil thermal conductivity to the changes in soil moisture from irrigation (Kanamaru and Kanamitsu 2008).

There is a general consensus in the literature that irrigating vegetated surfaces can lead to an increase in Q_E and a concurrent reduction in Q_H. Using the Regional Atmospheric Modelling System, Adegoke et al (2003) simulated the effect of irrigation in Nebraska, USA, by keeping the upper 0.2 m of soil saturated. They estimated that mean Q_E would increase from 74.5 to 98.2 W m⁻², while Q_H would decrease from 86.9 to 79.8 W m⁻². Chen et al (2017) initiated irrigation in their modelling study only when the root-zone soil moisture availability dropped below 50% during the growing season. The root-zone soil moisture availability was defined as the ratio of the difference between the current root-zone soil moisture and the wilting point and the difference between field capacity and wilting point. Simulated irrigation led to a daily mean increase in Q_E of 2.4 W m⁻² and a reduction in Q_H of 2.1 W m⁻². Broadbent et al (2018) studied a range of daily irrigation amounts from 5 up to 30 mm in their modelling study in a North Adelaide suburb. At 3 pm local time, the model predicted that with daily irrigation of 30 mm all the available energy was consumed in evapotranspiration (Q* = Q_E) because of the nearly unlimited soil moisture supply. This evaporation from the land surface caused the T_sfc to drop below the T_a, resulting in a negative Q_H (−40 W m⁻²).

This shift in the partitioning of surface energy fluxes is confirmed by a 12year experimental study, which measured Q_E and Q_H over maize-soybean rotation fields using eddy-covariance flux tower systems in Nebraska, USA (Chen et al 2018). They measured a mean increase in Q_E in irrigated maize fields of approximately 20 W m⁻² and a reduction in Q_H of 25 W m⁻². However, the changes in the partitioning of surface energy fluxes were more subdued in the soybean fields. As such, the measured cooling effect in terms of T_a was greater in the irrigated maize fields than in the irrigated soybean fields. This comparison between maize and soybean suggests that vegetation type has an important impact on surface energy balance and the cooling effect of irrigation.

Most modelling studies predicted a reduction in daily mean T_sfc with irrigation. Vahmani and Ban-Weiss (2016) modelled the climate impacts of irrigating xeric landscapes in Los Angeles, USA. Although they predicted an increase in daily mean T_a with irrigation, because of the increased release of soil heat storage at night, their model predicted a small reduction (0.2 °C) in mean T_sfc. Kanamaru and Kanamitsu (2008) modelled the hourly differences in T_sfc between the irrigated and unirrigated crops in the California Central Valley, USA and noted that irrigation decreased the daily maximum T_sfc by 5.1 °C at 3 pm, but increased the daily minimum T_sfc by 3.1 °C at 5 am. A small daily mean reduction in T_sfc (0.5 °C) was achieved despite the irrigation causing an extended warming period (8 pm–6 am). Using an urban canopy model, Yang and Wang (2015) predicted a decrease of 4.6 °C in the T_sfc of a mesic residential landscape in Phoenix, USA when applying a daily irrigation amount of 1.4 mm. Other regional modelling studies, using the Weather Research and Forecasting model, predicted that irrigation would lead to a reduction in mean daily T_sfc of between 1 °C and 2 °C (Wang et al 2019, Gao et al 2020). Irrigation is likely to reduce daily mean and daytime T_sfc although night-time T_sfc may increase due to the increased soil heat storage during the day and subsequent release at night.

In terms of daily mean T_a, the vast majority of studies reported a cooling effect from irrigating vegetated surfaces. Broadbent et al (2018) modelled an increase in cooling effect from −0.5 °C for 5 mm d⁻¹ of irrigation to −2.3 °C for 30 mm d⁻¹ during a heatwave in Mawson Lake, North Adelaide, Australia. There was a diminishing cooling efficiency as the daily irrigation amount increased because the surface soil became saturated and the evapotranspiration rate was limited by the atmospheric demand. Chen et al (2018) measured the cooling effect from irrigating a soybean field and from a maize field in Nebraska, USA. The cooling effect was much higher in the maize field (−0.43 °C) than the soybean field (−0.09 °C); the difference was confirmed by the contrast in surface energy fluxes between the two crops (see section 4.3). The authors explained the stronger cooling effect of the maize field by crop phenology such as plant height and leaf area index). In a recent observational study, the irrigation cooling effect measured in two urban parks in Melbourne, Australia was −2 °C to −1 °C during a non-heatwave period, and the effect strengthened to −4 °C to −2 °C during the heatwave period (Lam et al 2020).

Vahmani and Ban-Weiss (2016) used the Weather and Regional Forecasting model to predict that irrigation would lead to an increase in the daily mean T_a because of increased soil heat storage during the day; the storage would release at night, offsetting the smaller cooling benefit by day. Several other modelling studies have similarly predicted an increase in the daily minimum T_a, again as a result of stored heat releasing at night (Kanamaru and Kanamitsu 2008, Broadbent et al 2018, Valmassoi et al 2020). However, not all modelling studies have reported a night-time warming in response to irrigation (Sorooshian et al 2011, Gao et al 2020).

The impact of irrigation on the daily maximum T_a is more consistent amongst modelling studies as most of them have predicted a reduction in the daily maximum T_a with irrigation (Kanamaru and Kanamitsu 2008, Sorooshian et al 2011, Gao et al 2020). The reduction in the daily maximum T_a seems to be associated with the magnitude of the daily maximum T_a. For example, Gao et al (2020) predicted a small reduction of 0.4 °C when the maximum T_a in the unirrigated scenario was 27.9 °C, whereas Kanamaru and Kanamitsu (2008) modelled a larger reduction of 2.1 °C when the maximum T_a was 34.6 °C, and Sorooshian et al (2011) modelled a 5.1 °C reduction from irrigation when the maximum T_a was 38.0 °C. Weather conditions are clearly an important factor in determining the strength of the cooling effect from irrigation (see detailed discussion in section 4.2).

As expected, research literature unanimously indicates there will be an increase in VP or other air humidity indices with irrigation. A controlled experiment in maize-soybean rotation fields in Nebraska, USA, measured irrigation increased the mean mixing ratio by 0.52 g kg⁻¹ (∼4%) over a 12 years study period (Chen et al 2018). Geerts (2002) compared specific air humidity inside and outside of the Murrumbidgee–Coeambally–Murray irrigation area in Australia using historic data (1968–1996) from 28 weather stations. They observed that irrigation increased the mean specific humidity inside the irrigated areas by 0.9 g kg⁻¹. Similar humidity increases were predicted in other agricultural modelling studies (Sorooshian et al 2011, Yang et al 2017). For example, Harding and Snyder (2012) modelled the impacts of irrigation on climate in the Great Plains, USA by keeping soil moisture in the top 2 m at saturation. They predicted a 0.19 g kg⁻¹ (∼2%) rise in mixing ratio over the irrigated area.

Only a few studies have considered the impact of irrigation on human thermal stress. Broadbent et al (2018) modelled the impact of irrigation on human thermal stress in North Adelaide, Australia using the humidex index. Humidex integrates the effect of T_a and VP on human thermal stress into a single index. It is the dry T_a (with a negligible moisture content) at which its corresponding thermal stress level equates to that of a given combination of T_a and VP (Masterton and Richardson 1979). The 'comfortable' Humidex range is between 20 °C and 29 °C, while the 'varying degrees of discomfort' range is between 30 °C and 39 °C. Their model predicted that irrigation reduced Humidex from 36.9 °C to 34.6 °C at 3 pm for a daily irrigation amount of 20 mm, suggesting a mitigation of heat stress on humans. They noted that the background humidity in North Adelaide was so low that the rise in humidity from irrigation would barely increase human heat stress. Yang and Wang (2015) modelled the impacts of irrigation on human thermal stress using the Index of Thermal Stress. The Index of Thermal Stress measures the rate of heat dissipation that the human body needs to achieve through sweating in order to maintain thermal equilibrium with the surrounding environment (Givoni 1963). An Index of Thermal Stress above 400 W indicates a 'very hot' condition. They modelled that irrigation reduced the Index of Thermal Stress in all but one month of the year, and the greatest thermal stress reduction would be 32.5 W in June. In a field experiment, Shashua-Bar et al (2011) measured the Index of Thermal Stress in an exposed area with irrigated grass to that in an exposed area with bare soil. The irrigation kept the thermal stress level in the lawn at 'warm' for most of the day, whereas the 'hot' and 'very hot' levels persisted in the area with bare soil. However, this comparison did not only reflect the impact of irrigation because the surface type and albedo of the two sites were different.

UGS irrigation has the potential to induce a cooling effect by modifying the urban surface energy balance. Assuming that the net horizontal advective heat flux and the anthropogenic heat flux are negligible, the urban surface energy balance (all in W m⁻²) for a grass-covered surface can be expressed as:

$$\begin{equation}{Q^*} = { }{Q_{\text{E}}} + {Q_{\text{H}}} + \Delta {Q_{\text{S}}}{ }\end{equation} \tag{ 1 }$$

where Q* is the net all-wave radiation, Q_E the latent heat flux, Q_H the sensible heat flux and ΔQ_S the net storage heat flux (Oke 1988). The partitioning of Q* into Q_E, Q_H and ΔQ_S is primarily dependent on surface type and soil moisture status (Williams and Torn 2015). In the case of grass, soil moisture becomes the sole factor in the partitioning.

The theoretical cooling mechanism of UGS irrigation is depicted in figure 1. The figure describes the differences in soil moisture, surface energy fluxes and some climate variables between an unirrigated and an irrigated UGS in the daytime and night-time in summer. In the daytime, irrigation increases soil moisture and promotes evapotranspiration in the irrigated UGS (Chen et al 2018). More energy is converted to Q_E and less to Q_H, resulting in a lower (T_a, T_sfc and human thermal stress (Broadbent et al 2018). Downward G may slightly increase because a higher soil moisture is associated with a higher soil thermal conductivity, which increases the heat conduction into the soil (Kanamaru and Kanamitsu 2008). However, G is usually one order of magnitude smaller than Q_E and Q_H during the day (Spronken-Smith et al 2000), making it less influential on the daytime T_a. Direct evaporation of water from the soil surface increases, leading to a lower T_sfc (Lobell et al 2009). Irrigation may also support a lusher growth of grass, which further promotes Q_E through transpiration (Valmassoi et al 2020). The enhanced evapotranspiration from the irrigated surface also raises VP (Sorooshian et al 2011). In contrast, the unirrigated UGS lacks evapotranspiration and Q_E, causing a higher Q_H (Chen et al 2018). The resultant effects are a higher T_a and T_sfc, but a lower VP (Gao et al 2020). In the night-time, Q_E and VP remain higher in the irrigated UGS than the unirrigated UGS because of the higher soil moisture (Valmassoi et al 2020). The increased soil heat storage in the irrigated UGS from the daytime promotes the upward G at night, leading to a higher T_sfc and Q_H (Vahmani and Ban-Weiss 2016). However, it is unclear whether the night-time T_a and human thermal stress in the irrigated UGS are higher or lower than that in the unirrigated UGS because the increased Q_E tends to reduce T_a and thermal stress while the increased G and Q_H tend to increase them.

Zoom In Zoom Out Reset image size

In addition to metabolic rate and clothing insulation, human thermal stress is determined by four climate variables, namely T_a, VP, mean radiant temperature and wind speed (Fanger 1970, Höppe 1999, Bröde et al 2012). The theoretical cooling mechanism suggests that human thermal stress can be reduced by UGS irrigation (figure 2). The impacts of irrigation on human thermal stress begin with increasing soil moisture, which then modifies the partitioning of surface energy fluxes (Q_H, Q_E and G) by increasing evapotranspiration. The increased evapotranspiration is generally associated with a smaller Q_H and a larger Q_E and G. The changes in these three energy fluxes are expected induce a reduction in T_sfc and T_a. The larger Q_E also inevitably increases VP. Eventually, the changes in T_sfc, T_a and VP affect human thermal stress. The lower T_a directly reduces human thermal stress, while the lower T_sfc reduces the stress by reducing the mean radiant temperature. The enhanced evapotranspiration from irrigation can increase VP and offset part of the cooling benefit, but irrigation is likely to cause a net reduction in thermal stress (Broadbent et al 2018). Different human thermal indices have been developed to integrate the effects of some of the four essential climatic variables e.g. Humidex (Masterton and Richardson 1979), or all of them, e.g. Index of Thermal Stress (Givoni 1963) and Universal theraml climate index (UTCI) (Bröde et al 2012). The impacts of irrigation on human thermal stress are best assessed by these thermal indices.

Zoom In Zoom Out Reset image size

Background climate is the average weather conditions of a specific region over multiple decades. Both global (Sacks et al 2009, Thiery et al 2017) and regional (Gao et al 2020) modelling studies have agreed that background T_a and rainfall are important factors that influence irrigation cooling effect. From their global modelling results, Sacks et al (2009) developed a simple linear relationship between irrigation cooling effect and daily irrigation amount separately for areas with a higher rainfall (>2.43 mm d⁻¹) and for those with a lower rainfall (⩽2.43 mm d⁻¹). They predicted that, for a 1 mm d⁻¹ increase in irrigation, the additional cooling effect in the drier areas was −0.7 °C stronger than the wetter areas. Thiery et al (2017) modelled the irrigation cooling effect in seven heavily irrigated regions in the world. Given a similar irrigation amount, they predicted a mean cooling effect of <−1 °C in T_sfc in western North America and central North America in summer, and no cooling effect in Southeast Asia and East Asia. The main reason is that Western North America and central North America have a drier climate than Southeast Asia and East Asia. Irrigation in drier regions will induce a stronger evapotranspirative cooling effect because evapotranspiration is dependent on the availability of soil moisture (Koster et al 2006). A similar conclusion was drawn by a regional modelling study (Gao et al 2020), which predicted an increasing irrigation cooling effect from the coast towards the inland area in metropolitan Sydney, which coincided with the increasing background T_a gradient.

The quantitative relationship between irrigation cooling effect and background climate was established by a systematic review study (Cheung et al 2021). Cheung et al (2021) reviewed 17 studies that have reported the summertime mean irrigation cooling effect. They established a multiple linear regression model to predict irrigation cooling effect by background climate variables, namely T_a, rainfall, specific humidity, wind speed and net radiation. Only T_a and rainfall were the statistically significant variables that remained in the regression model after a stepwise elimination procedure. The model predicted that the irrigation cooling effect can strengthen by approximately −0.1 °C for every 1 °C increase in background mean T_a or 10 mm month⁻¹ reduction in rainfall. In principle, this regression model corroborated with the findings in the literature because it suggested that background T_a and rainfall are the main factors that influence irrigation cooling effect.

Notable seasonal variations in the magnitude of irrigation cooling effect have been reported by two modelling studies which applied a constant daily irrigation amount throughout the year. Lobell et al (2009) modelled the monthly irrigation cooling effect in eight major irrigated regions in the northern hemisphere to be −5 °C in the dry season, while the cooling effect was hardly noticeable in the wet season. Zou et al (2014) modelled the monthly irrigation cooling effect in Haihe River Basin, China. They predicted a strengthening cooling effect from −2 °C in April to −4 °C in July as background T_a increased. They also predicted a warming effect up to 4 °C in the winter months because of the constant irrigation throughout the year.

Day-to-day variations in irrigation cooling effect were also evident in an observational study in two urban parks in Melbourne, Australia (Lam et al 2020). Comparing to the non-heatwave period, the irrigation cooling effect in several lawn areas was −4 °C to −2 °C stronger during heatwaves. Similar to background climate, a warmer weather can increase irrigation cooling effect on a seasonal and daily basis because it provides more energy for evapotranspiration and increases VP deficit.

In an experimental study, the daily mean irrigation cooling effect in maize fields (−0.43 °C) in Nebraska, USA was significantly higher than that in soybean fields (−0.09 °C) (Chen et al 2018). The cooling effect from irrigation correlated with a decrease in sensible heat flux in the maize fields, whereas irrigation induced little change in sensible heat flux in the soybean fields. The difference in the cooling effect between maize and soybean may be attributed to their differences in plant height and leaf area index which affect the transport of heat. There is a paucity of studies that compare irrigation cooling effects among vegetation types, partly because the majority of the current land surface models do not account for different vegetation types (Ozdogan et al 2010). One exception is a modelling study that compared the latent heat flux between a scenario where only one generic crop type was used and a scenario where the crop types where classified in detail (Ozdogan et al 2010). The latent heat flux of the scenario with only one generic crop type was approximately 5 W m⁻² higher, indicating a stronger cooling effect.

Different vegetation types can influence irrigation cooling effect because of their differences in water demand and physical characteristics. For example, cool-season grasses generally have a higher crop factor (∼0.65) than warm-season grasses (∼0.25) (Handreck and Black 2001). Trees and shrubs can have a crop factor >0.7 (Connellan et al 2002). Crop factor is the proportion of water used by the plant in comparison to the water evaporated from a evaporation pan (Doorenbos and Pruitt 1977). Cool-season grasses, trees and shrubs may therefore induce a stronger irrigation cooling effect than warm-season grasses with irrigation because they transpire more water per unit area.

Turfgrasses, shrubs and trees are common urban vegetation types that provide cooling benefits to UGS visitors by reducing air temperature. Their cooling effect is dependent upon their ability to transpire and shade (Rahman et al 2019), as well as their impacts on aerodynamic roughness (Meili et al 2021) and wind speed (Xing et al 2019). In comparison to tall shrubs and trees, turfgrasses do not provide overhead shading and therefore their cooling effect is mainly dependent upon their transpiration rate and albedo, which is further dependent upon their species, root system and plant area index. Short shrubs (<1 m) behave similarly to turfgrasses because they are not tall enough to provide shade for humans or nearby dark impervious surfaces. Appropriate irrigation can support the growth and health of both turfgrasses and shrubs, increasing their plant area index. A vigorously-growing turfgrass can have a crop factor of 0.7, whereas a moderately-growing turfgrass may only have a crop factor of 0.25 (Handreck and Black 2001), meaning that a vigorously-growing grass transpires more water per unit area and induces a stronger cooling effect. Turfgrasses and shrubs with a higher plant area index also have a higher albedo, which further reduces air temperature by reducing the amount of radiation absorbed and later released by the ground surface (Shiflett et al 2017). The impacts of turfgrasses and short shrubs on aerodynamic roughness and wind speed are smaller than those of trees and therefore their cooling benefits may be easily diluted by near-surface turbulent mixing and advection (Spronken-Smith and Oke 1998). Nevertheless, the advected cool air can benefit the urban areas downwind (Sugawara et al 2015).

In the case of tall shrubs and trees, shading may contribute approximately 70% of their cooling effect and transpiration the rest 30% (Tan et al 2018). Appropriate irrigation can support the growth and health of tall shrubs and trees, increasing their plant area index. Tall shrubs and trees with a higher plant area index can induce a stronger cooling effect from increasing overhead shading for humans and dark impervious surfaces, as well as increasing overall transpiration (de Abreu-harbich et al 2015, Sanusi et al 2017). Moreover, the presence of tall shrubs and trees in UGS can reduce wind speed at the pedestrian level (Xing et al 2019) and therefore retain the cool air within the UGS for longer. The impact of irrigation on the cooling effect of tall shrubs and trees are more complex than turfgrasses and short shrubs. Most UGSs have a combination of turfgrass areas with and without trees and shrubs, such that their impacts upon the energy balance and therefore cooling effects are complex. UGSs will often contain vegetation with high and low transpiration rates, high and low leaf area indices, taller vegetation will shade lower vegetation and taller vegetation will change wind speed, aerodynamic roughness and turbulent exchange (Kent et al 2017). More studies are required to understand the complex interactions between irrigation, plant area index and cooling effect of these common urban vegetation types.

Irrigation time refers to the time in 24 h diurnal period when irrigation is applied. Only a limited number of studies have examined the impact of irrigation time. Broadbent et al (2018) modelled a negligible (<0.2 °C) difference in the daily mean cooling effect between daytime (11 am–5 pm) and night-time (11 pm–5 pm) irrigation in Mawson Lake, North Adelaide, Australia; however, the diurnal variation of the cooling effect was not reported. Valmassoi et al (2020) modelled the diurnal variations in the irrigation cooling effect of night-time (5 UTC), midday (12 UTC) and afternoon (15 UTC) irrigation with sprinklers in Po Valley, Italy. The night-time irrigation regime would induce a cooling effect of ⩾−0.4 °C after 2 h from the starting time and it would have almost no impact on T_a in the rest of the day. The midday irrigation regime would induce a cooling effect of ⩾−0.2 °C at 15 UTC and a warming of a similar magnitude at night. The afternoon irrigation would induce a cooling effect of ⩾−0.7 °C which sustained for most of the time at night. Although the daily mean cooling effect was not explicitly reported, night-time and afternoon irrigation seemed to induce a stronger daily mean cooling effect than midday irrigation.

Studies that modelled two levels of daily irrigation generally predicted an increase in cooling effect with increased irrigation amount (Kanamaru and Kanamitsu 2008, Sorooshian et al 2011, Zou et al 2014). However, the results were mixed for studies that modelled the cooling effects from more than two levels of daily irrigation (Lobell et al 2009, Broadbent et al 2018, Wang et al 2019). In a global modelling study, Lobell et al (2009) predicted that the cooling effect would be almost the same for keeping soil moistures at 30%, 40% and 90% of saturation, because energy would become the greatest limiting factor upon evapotranspiration and latent heat flux, when soil moisture exceeds 30%. In a local-scale modelling study, however, daily irrigation amounts of 5, 15 and 30 mm d⁻¹ would lead to mean daily cooling effects of −0.5 °C, −1.5 °C to −2.3 °C, respectively (Broadbent et al 2018). This local-scale model predicted a non-linear or 'diminishing return' in cooling effect with increasing daily irrigation amount. A direct comparison between these two studies is impossible because the equivalent daily irrigation amounts of keeping soil moisture at 30%, 40% and 90% are unknown and dependent upon factors such as soil type, regional climate and vegetation type. This concept of diminishing cooling effect with increasing irrigation amount is supported by Wang et al (2019), who modelled the irrigation cooling effects at four irrigation amounts in the contiguous USA. They predicted that the evapotranspiration and latent heat flux can be enhanced with large daily irrigation amounts only in the semi-arid and arid regions because of the large VP deficit in these regions, and only up to an amount of 10 mm d⁻¹. Background regional climate appears to be an important factor in determining how much cooling effect is possible with increasing daily irrigation amounts.

Irrigation cooling effect can persist after irrigation stops as water continues to evaporate from the soil surface and plants continue to transpire. Experimental and observational studies have provided evidence that the cooling effect after irrigation can last for hours (Lam et al 2020) and days (Chen et al 2018), while modelling studies predicted that it would last for months after continuous, daily irrigation in the warm season (Sorooshian et al 2011, Yang et al 2016). The cooling effect of evening and night-time irrigation (8 pm–7 am) in Melbourne Gardens, Melbourne, Australia persisted for several hours into the morning (Lam et al 2020). The cooling effect from irrigating maize fields in an experimental farmland in Nebraska, USA was <−0.5 °C in the first 6 d after irrigation and was still evident (<−0.2 °C) 11 d after irrigation (Chen et al 2018).

The cooling effect of irrigation applied from May to August in the Central Valley, California, USA, was modelled (NCAR/Penn State MM5) to persist into September (−1.1 °C) and October (−0.3 °C) (Sorooshian et al 2011). Yang et al (2016) modelled the climate impacts of springtime (March–April–May) irrigation in the Huang-Huai-Hai Plain, China. The latent heat flux in the irrigated scenario was predicted to reduce significantly in the 3 months (June–July–August) after irrigation stopped, with only a minor reduction in T_a (0.1 °C). Modelling studies reported a much longer duration of cooling after irrigation than experimental and observation studies mainly because they tracked the impacts of irrigation for a much longer period (a few months) after irrigation stopped. In comparison, the experimental study only measured the duration of cooling after irrigation up to 11 d because the irrigation was designed to applied every 4–11 d (Chen et al 2018). The cooling effect might have lasted longer than 11 d, but the experiment did not measure it.

The background climate of a city, primarily background mean T_a and rainfall, determines whether UGS irrigation is an effective cooling strategy. A higher background T_a and lower rainfall will increase the cooling potential of irrigation, and vice versa. To demonstrate the impact of background climate we use the simple regression model developed by Cheung et al (2021), to estimate the cooling potential of UGS irrigation in three global cities of contrasting climate: Hong Kong, Melbourne and Phoenix. This regression model predicts the cooling potential of UGS irrigation as the difference in daily mean T_a between an irrigated and an unirrigated UGS. The cooling potential of irrigation is presented for four seasonal periods: DJF, March–May (MAM), JJA and September–November (SON) (figure 4).

Zoom In Zoom Out Reset image size

Hong Kong has a dry-winter humid subtropical climate (Köppen–Geiger climate classification: Cwa). It is a city without a warm and dry season (figure 3). The winter (DJF) in Hong Kong is dry (mean rainfall = 30 mm month⁻¹) but not warm (mean T_a = 17.3 °C) (figure 4(a)); the other three seasons are warm (mean T_a > 22.9 °C) but not dry (mean rainfall >160 mm month⁻¹). As a result, the impact of UGS irrigation in the four seasons in Hong Kong is positive (warming) except in DJF (figure 4(a)). This warming is likely due to the increased soil thermal conductivity of wet soil after irrigation (Kanamaru and Kanamitsu 2008). This leads to an increased soil heat storage during the day which releases during the night, causing a substantial night-time warming which outweighs the daytime cooling effect. Other urban cooling strategies such as urban greening and canopy shade (Cheung and Jim 2018), improving urban ventilation (Tan et al 2017) and increasing albedo of impervious surfaces (Akbari et al 2012) should be considered.

Melbourne has an oceanic climate (Köppen–Geiger climate classification: Cfb). It is a city with at least one warm and dry season (figure 3). The summer (JJA) in Melbourne is warm (mean T_a = 17.3 °C) and dry (mean rainfall = 43 mm month⁻¹) (figure 4(b)). The estimated cooling potential of UGS irrigation under these summer climate conditions is −0.6 °C, whereas the irrigation impact in the other three seasons is neutral. Since the cooling potential of irrigation in this simple model is the daily mean difference in T_a the cooling effect during the middle of the day is likely to be <−0.6 °C, suggesting that USG irrigation can be considered an effective cooling strategy to reduce daytime T_a.

Phoenix has an arid, hot desert climate (Köppen–Geiger climate classification: BWh). It is a city with more than one warm and dry season (figure 3). The background mean T_a and rainfall in MAM, JJA and SON are >23.0 °C and <18 mm month⁻¹, respectively (figure 4). Under such warm and dry climate conditions, the estimated cooling potential of UGS irrigation in these three warm, dry seasons are −1.4 °C, −2.3 °C and −1.5 °C, respectively. UGS irrigation is a very effective cooling strategy for Phoenix.

As global climate change progresses, many cities are likely to become warmer with more variable rainfall patterns (Peck et al 2012, Darmanto et al 2019). Urban expansion is also likely to increase the intensity of urban heat islands because of reduced evapotranspiration and increased heat stored in urban structures which releases as sensible heat (Argüeso et al 2014). Thus, this decision framework may be used to consider the projected climate of a city to determine whether or when UGS irrigation will become an effective urban cooling strategy in the future.

After determining the cooling potentials of UGS irrigation in a city based on its background climate, the next decision step is to consider irrigation water supply (figure 3). Although potable water remains the most common source of irrigation water supply in global cities, alternative water sources are being developed in many forward-thinking cities to support UGS irrigation and to reduce potable water consumption (Grant et al 2012). Cities can increase their irrigation water supply by harvesting stormwater and roof water runoff (Hamlyn-Harris et al 2018) and storing this in ponds or above-ground or below-ground (Livesley et al 2021). Stormwater harvesting schemes in new developments in Melbourne, Australia can increase the city's non-potable water supply by seven times (9.8% of municipal water consumption) until 2050 (Environment and Natural Resources Committee 2009). Irrigation water supply can also be enhanced by treating municipal sewage using conventional and advanced techniques (Leverenz et al 2011). The majority of Israel's municipal sewage (73%) is treated and reused for agricultural irrigation (5% of national-wide water consumption) (Tal 2006).

5.2.1. Soil moisture-controlled irrigation

5.2.2. Temperature-controlled irrigation regime

5.3.1. Soil properties

5.3.2. Fauna and flora ecology

5.3.3. Types of amenity use