Assessing the Air Humidity Characteristics of Local Climate Zones in Guangzhou, China

Abstract

An urban canopy’s humidity significantly affects thermal comfort, public health, and building energy efficiency; however, it remains insufficiently understood. This study employed 3-year (2020–2022) fixed measurements from Guangzhou to investigate the temporal patterns of relative humidity (RH), vapor pressure (Ea), and vapor pressure deficit (VPD) across eight local climatic zones (LCZs). Clear and distinct patterns in the humidity characteristics among the LCZs were revealed on multiple timescales. The RH and VPD of each zone were higher in summer than in winter, with peak RH observed in LCZ A (83.45%) and peak VPD in LCZ 3 (13.6 hPa). Furthermore, a significant daytime urban dry island (UDI) effect in the summer and a nighttime urban moisture island (UMI) effect in the winter were observed in terms of the Ea difference between urban and rural areas. The strongest UMI occurred during winter nights in LCZ 8, with a peak intensity of 0.8 hPa, while the UDI was more frequent during summer days in LCZ 1, with a maximum value of −1.2 hPa; meanwhile, compact areas had a slightly higher frequency of UDI than open areas. Finally, the effects of the urban heat island (UHI) and wind speed (V) on UMI were analyzed. During the daytime, a weak correlation was observed between the UHI and UMI. Wind enhanced the intensity of the nighttime UMI. This research offers further insights into the canopy humidity characteristics in low-latitude subtropical cities, thereby contributing to the establishment of a universal model to quantify the differences in moisture between urban and rural areas.

1. Introduction

Urbanization transforms the underlying surface of a city, replacing natural landscapes with impervious surfaces, resulting in changes in the heat and humidity environments of urban areas [1,2]. These changes include a decline in the latent heat flux [3], increases in the sensible heat flux and surface temperature, and a decrease in humidity [4]. The urban heat island (UHI) effect and urban moisture island effect (UMI) are urban microclimate modification phenomena [5]. Investigating urban climate issues, particularly in humid environments, is vital for improving living conditions, establishing a scientific evaluation system for the urban physical environment [6], and aiding urban planning and energy-efficient construction [7,8].

Traditionally, the UHI and UMI have been described through air temperature and humidity differences between urban and rural areas. However, this binary classification often fails to capture intricate intra-urban and intra-rural temperature variations. To address this limitation, Stewart and Oke introduced local climate zones (LCZs) as a universal framework for urban climate research, providing a climate-based and objective classification of urban and rural landscapes [9]. This framework categorizes urban spaces into 17 distinct classes based on surface cover and structure (e.g., building height and spacing), facilitating nuanced comparisons and a deeper understanding of urban climate complexities.

Various methods are used to measure the impact of urbanization on changes in humid environments. The most widely used method is based on fixed observations at urban and rural meteorological stations, such as those in Beijing [10], Szeged [11], and Turin [12]. This method utilizes observational data from both urban and rural meteorological stations and can intuitively compare and analyze differences in humidity between urban and rural areas. While this method allows for straightforward comparison of urban and rural humidity, it is limited by environmental factors and the availability of meteorological stations. To better describe the spatial patterns of urban humidity, some researchers have employed mobile measurements to build urban humidity databases, such as by driving on specific routes in cities using automobiles fitted with meteorological sensors [13,14]. This method is cost-effective and provides spatial insights, but it cannot effectively capture the temporal dynamics of humidity [15]. Modeling is an effective way to study urban climate change by quantifying the factors influencing UMI. The urban canopy model (UCM) is often combined with weather research and forecasting (WRF) models to study the influence of the underlying urban surface and canopy on atmospheric motion. The impacts of street and canyon geometry, building emissions, and vegetation on outdoor air temperature (T), humidity, and thermal stress in Hong Kong were studied during a typical heat wave period based on the UCM, and it was found that as the street canyon height/width ratio increased, the daytime UHI phenomenon transformed into an urban cool island (UCI) phenomenon [16].

The observed humidity in urban areas may be influenced by several factors at different scales, such as albedo, the impervious plan area ratio, and average building height (HRE). The difference in humidity parameters in the same local area may be greater than that between regions [11]. In Nanjing, Yang, et al. [17] studied the spatial and temporal characteristics of the UMI in local climate zones (LCZs) based on measured data using the relative humidity (RH) and humidity ratio (q) as variables. The authors found that the difference in RH between urban and rural areas increased during strong nighttime UHI effects. The UDI was observed more frequently during the daytime in warm months. Based on the LCZ theory, Dunjic, et al. [18] studied the temporal characteristics of local-scale humidity (RH, vapor pressure [Ea], absolute humidity [AH], q, and vapor pressure deficit [VPD]) and the influence of T in Novi Sad City with fixed observation data from weather stations and found that the UDI effect frequently occurred in LCZ 2 from February to October. In addition, the UMI intensity reached its maximum during the daytime and at midnight during the heat wave. In Szeged City, Unger, Skarbit and Gál [11] calculated and compared the differences in Ea and temporal characteristics among different LCZs using data obtained from a city meteorological network based on the LCZ. The peak Ea occurred in the summer and the minimum value occurred in the winter. In Hong Kong, Du, et al. [19] used WUDAPT to obtain the urban morphology and land use and land cover (LULC) data to create an LCZ map, which was integrated into the WRF model to reveal the water transportation mechanism within the city. Weakened upward motion, blocked ventilation, and inhibited dew formation were the reasons for the higher moisture content in the cities than those in the suburbs at night. A positive correlation between the nocturnal UHI and UMI has been reported in Guangzhou [20], where the UMI intensity (UMII) decreased at sunrise and increased at sunset. Linear regression analysis showed that the UMII was proportional to the impervious plan area ratio and inversely proportional to the green cover ratio.

Despite these contributions, there are still deficiencies in understanding the complex dynamics of urban humidity, especially at the finer scales within individual urban zones. Most research has focused on thermal or general moisture patterns, yet few studies have comprehensively examined how multiple humidity parameters, such as RH, Ea, and VPD, interact in specific urban contexts. Additionally, temporal variations in canopy humidity across different LCZs have not been systematically studied. While LCZs offer a valuable framework for analyzing urban climates, understanding urban humidity dynamics requires a broader sample of cities, as the background climate of each city significantly influences its microclimatic patterns.

Establishing a robust database of urban humidity characteristics across various cities is crucial for refining LCZ theory and better understanding the mechanisms driving urban humidity changes. This study focused on Guangzhou, a representative humid subtropical city in southern China. Based on the LCZ theory, we selected eight representative zones and conducted long-term measurements of T and RH in each LCZ. We used RH, Ea, and VPD as variables and analyzed the temporal characteristics under different urban morphologies using 3 y of hourly observation data. This study contributes to establishing a scientific evaluation system for thermal and humid environments, which is of great significance in assisting urban planning and building energy conservation.

2. Methods

2.1. Study Area

Guangzhou, the provincial capital of Guangdong, is located in the south of China (23°08′ N, 113°15′ E) and is situated on the northern edge of the Pearl River Delta, facing the South China Sea. The city terrain is mountainous in the northeast, hilly in the central region, and comprises plains in the south. With an area of approximately 7434 km², Guangzhou has a population of 18.81 million residents, according to the 2021 census. According to the Köppen-Geiger classification system, Guangzhou has a humid subtropical climate (Cfa) characterized by hot and humid summers and cool and dry winters [20]. Based on the dividing region for building thermal design, Guangzhou is a hot summer and warm winter area, where the hottest month is July, with a mean T of 28.7 °C, and the coldest month is January, with a mean T of 13.3 °C. The dominant wind direction is southeast in the summer and north in the winter. Guangzhou has an annual average rainfall of approximately 1600–1900 mm and annual sunshine duration of 1500–2000 h. The total solar radiation during the summer is approximately 7318 w/m² (https://data.cma.cn/, accessed on 23 June 2023).

Guangzhou’s urbanization, marked by changes in land cover and the expansion of built-up areas, has significantly altered local climate conditions. Meteorological data and prior research show that this rapid urbanization has contributed to a notable increase in the city’s average temperatures [21]. Specifically, Guangzhou’s annual average temperature has steadily increased since 1953, with rates accelerating from 0.22 °C per decade (1953–2009) to 0.38 °C per decade (1973–2009), and further to 0.49 °C per decade (1983–2009) [22]. This consistent and speeding up warming pattern indicates both the increasing intensity of the UHI and a clear trend of regional climate warming in Guangzhou. This warming trend, in turn, interacts closely with the city’s diverse urban forms. From densely built-up high-rise areas to more suburban and green spaces, these various urban landscapes result in distinct microclimates that are strongly influenced by urbanization. The LCZ framework can effectively capture these variations, distinguishing zones with differing built environment and land cover characteristics and enabling a detailed analysis of urbanization’s influence on canopy humidity.

2.2. LCZ Sites

To conduct an objective and standardized study of urban climates, Stewart and Oke [9] proposed 17 standard classes of local climates. LCZs are defined as regions with similar surface covers and surface structures (e.g., building height and spacing), with areas ranging from hundreds of square meters to several kilometers. This method helps to identify potential local climate problems and impacts caused by urban development.

We conducted LCZ zoning in Guangzhou using the method proposed previously [9]. First, by referring to the LCZ map of Guangzhou [23], satellite images, remote sensing data, and other information, we screened LCZs with similar surface structures and land cover. Each 500 m radius LCZ was selected to fully consider the influences of buildings, roads, and trees. Finally, on-site inspections and measurements of urban morphology parameters, such as HRE, sky view factors (SVFs), and aspect ratio (H/W), were performed. It was confirmed that the selected LCZ was in accordance with the identified LCZ type. GIS was employed to extract the land use and land cover types for selected regions within Guangzhou, and the specific locations of each zones were subsequently identified, as seen in Figure 1.

The HRE, SVF, H/W, building density (BSF), and impervious surface area ratio (ISF) of the LCZ morphological indicators are listed in

Table 1. Types and morphological indicators of eight local climate zones (LCZs).

LCZ Type	Urban Morphological Indicator and Their Range of LCZ Categories
	H/W	SVF	BSF (%)	ISF (%)	HRE (m)
1 Compact high-rise	2.24 (>2)	0.48 (0.2–0.4)	30.55 (40–60)	49.28 (40–60)	35 (>25)
2 Compact mid-rise	1.06 (0.75–2)	0.526 (0.3–0.6)	42.70 (40–70)	54.69 (30–50)	19 (10–25)
3 Compact low-rise	0.84 (0.75–1.5)	0.57 (0.2–0.6)	44.86 (40–70)	51.61 (20–50)	11 (3–10)
4 Open high-rise	1.13 (0.7–1.25)	0.56 (0.5–0.7)	26.14 (20–40)	56.93 (30–40)	28 (>25)
5 Open mid-rise	0.83 (0.3–0.75)	0.51 (0.5–0.8)	28.21 (20–40)	55.27 (30–50)	20 (10–25)
8 Large low-rise	0.18 (0.1–0.3)	0.73 (>0.7)	36.31 (30–50)	35.60 (40–50)	5 (3–10)
A Dense trees	2 (>1)	-	4.4 (<10)	1 (<10)	13 (3–30)
D Low plants	<0.1 (<0.1)	0.9 (>0.9)	4.8 (<10)	2.7 (<10)	<1 (<1)

H/W: aspect ratio; HRE: average building height; SVF: sky view factor; BSF: building density; ISF: impervious surface area ratio.

. Specific methods for calculating the morphological indicators of each LCZ can be found in a previous study [24]. It can be observed that the morphological parameters of the eight zones generally fall within the defined range, and the locations of these districts are also consistent with the classification results obtained by WUDAPT [25].

2.3. Data

The monitoring experiment was conducted continuously from 1 January 2020, until 31 December 2022. Two temperature–humidity data loggers (HOBO MX2301, Onset Computer Corporation, USA), each with accuracies of ±0.2 °C and ±2.5%, were installed in each LCZ plot, and the sampling rate was set to once per hour. To obtain accurate and representative data, both instruments were placed within a 100 m radius of the LCZ center and 100 m apart, and hourly data were collected. The drift of the loggers was found to be less than 0.01 °C and 0.01% for RH, which is considered negligible for the purpose of this study. To further ensure the reliability of the data from each zone, the average value of the readings from these two loggers is then taken as the basis for subsequent analysis. These loggers ensured measurement accuracy in harsh outdoor environments. Each HOBO data logger was equipped with a radiation shield to minimize the effect of radiation during the measurement process. The installation positions of the loggers were well-ventilated, away from heat sources, and avoided being blocked by surrounding obstacles, such as buildings and trees. These instruments were installed on streetlights or utility poles at a height of 2.5 m, as far from vehicles and air conditioners as possible, and at least 3 m from the walls.

LCZ D was used as a reference site for rural areas. It is located far from the city center and is mainly covered by low plants and farmland. Hence, it is representative of typical non-urbanized landscapes [17]. Figure 2b presents one of the measurement points (D2) situated within LCZ D, with the photograph captured during the on-site instrumentation setup.

2.4. Typical Days for UMI Development

The dynamic and magnitude of humidity differences between urban and rural areas exhibit similarities to the UHI. The maximum intensity of both phenomena is observed on windless, clear summer nights, indicating that the UMI develops most favorably under “ideal” weather conditions [24]. It is important to investigate the temporal humidity patterns on such days to study UMI formation and development. Oke proposed weather factors (a calculation formula based on wind speed, cloud cover, and cloud shape) to identify typical meteorological days [24]; this method has been effectively utilized in studies on thermal behavior and humidity characteristics in Nanjing [17,26]. However, acquiring certain meteorological parameters using this method is challenging. Our previous work validated the significant linear correlation between the rural diurnal temperature range (DTR) and the daily maximum urban heat island intensity, with a Pearson correlation coefficient exceeding 0.7 [27]. A threshold for identifying typical meteorological days, DTR ≥ 10 °C, has been established [27]. Specifically, DTR is measured by subtracting the hourly minimum temperature from the maximum hourly temperature in a day (LCZ D). It serves as an indicator of the combined influence of various meteorological factors in rural areas and can be utilized to assess variations in the local thermal and humid environment. Therefore, in the present study, the DTR of the LCZ D was used to identify typical meteorological days.

Specifically, days with accumulated precipitation of >0.1 mm were excluded, and hourly precipitation data during the study period were downloaded from the Guangzhou National Meteorological Station (No. 59287). Days with fog (daily mean RH ≥ 80%) and days with hourly variations in wind speed exceeding 2 m/s were excluded [28]. Finally, 275 days with DTR ≥ 10 °C were selected as typical meteorological days for this study. The numbers of such days in each year and month are presented in

Table 2. Number of typical meteorological days during the study period (2020.1.1–2022.12.31).

	Jan	Feb	Mar	Apr	May	Jun	Jul	Aug	Sep	Oct	Nov	Dec	3 y Total
2020	0	6	2	9	3	3	7	4	1	5	8	11	275
2021	20	16	13	11	5	5	3	3	9	6	16	14
2022	11	0	10	11	5	2	10	8	19	13	1	5

.

To study the differences in the UMI between the day and night, 6 a.m. to 6 p.m. was designated as daytime and 6 p.m. to 6 a.m. on the following day was designated as nighttime.

2.5. Humidity Parameters

Understanding the distribution and variation of urban humidity requires analysis of both the relative moisture content (in relation to the current atmospheric state) and the absolute moisture content (under standard or typical conditions). Although there is no standard for selecting specific indices, most studies use different indicators to capture these two aspects of urban humidity. This study employed RH, Ea, and VPD as key indicators. These indices provide complementary perspectives by addressing both the absolute characteristics of humidity (e.g., UMI or UDI) and the relative characteristics (for comparative analysis), which ensured the comparability and representativeness of the research conclusions.

• RH

RH is the most commonly used parameter for measuring humidity owing to its ease of measurement and comparability. RH significantly impacts urban rainfall, evaporation, and human thermal comfort. It represents the percentage of vapor pressure in the air relative to the saturated vapor pressure (Es) at the same temperature and atmospheric pressure. This information can be obtained directly using a HOBO data logger.

• Ea

The actual water content in air can be assessed using various humidity metrics. Ea directly reflects the vapor content in the air, which does not change with variations in air temperature. However, the difference in temperature can affect evaporation and condensation, and can be reflected by changes in indicators, such as Ea. Ea can be calculated using Es, as follows:

$$E\mathrm{a}=Es\times {\displaystyle \frac{RH}{100}}$$

(1)

• VPD

VPD is an indicator that measures the difference between actual and saturated vapor in air. It represents air dryness and is an important indicator for studying water circulation in urban climates. Similar to the RH, VPD is sensitive to temperature. During the UHI period, higher temperatures can cause an increase in the Es, leading to a water vapor shortage. VPD can be calculated by subtracting Ea from Es, which can be obtained using the Bolton equation [29]. The Equation (3) was adjusted to match Wexler’s results and was extrapolated for temperatures less than 0 °C. It has an accuracy of 0.1% within the temperature range from −30 °C to 35 °C [30].

$$VPD=E\mathrm{s}-Ea$$

(2)

$$Es=6.112\times \exp \left({\displaystyle \frac{17.67\times T}{243.5+T}}\right)$$

(3)

• Definition of UMI

The UMII is usually calculated using Ea, RH, AH, and the humidity ratio. In this study, vapor pressure was selected as the only indicator to measure the UMII, i.e., the difference between the Ea of each LCZ and that of the corresponding LCZ D (ΔEa and hPa). If the value was positive, it was a UMI; otherwise, it was a UDI. Meanwhile, ΔRH and ΔEa are only used as variables to measure humidity characteristics.

$$\Delta E\mathrm{a}=Eurban-Erural$$

(4)

3. Results

3.1. Temporal Pattern of Humidity

Figure 3 depicts the characteristics of RH, Ea, and VPD for each LCZ in terms of seasonal, and diurnal features. The differences in RH among the zones were not significant; however, the seasonal and diurnal variations showed higher humidity in the summer than in the winter, and at night than during the day. During seasonal variations, higher Ea was observed in all zones in summer, while the differences between LCZs within the same season were not significant. Specifically, the average distribution of Ea across all LCZs in summer exhibited a “slight” diurnal variation, with higher values during the day and lower values at night, but the change did not exceed 5 hPa for both daytime and nighttime. It can be observed that the RH, Ea, and VPD in all zones are higher in summer than in winter. This is reasonable and can be explained as follows: (1) Summer is the season with relatively concentrated precipitation in Guangzhou. Although the Es increases with temperature, the frequent precipitation caused by the influence of warm and humid ocean currents is the dominant factor for the increase in RH and Ea. (2) In summer, the high temperature causes intense evaporation of water in surface water bodies and soil. This process is significantly weakened in winter. (3) Due to the higher Es in summer than in winter, even if the water vapor in the air is closer to saturation, the “distance” from saturation can still be higher than in winter. During the summer and summer nights, each LCZ showed a higher average VPD than in the winter and summer days, indicating that the air was drier during these periods. In the winter, the difference in air dryness between the LCZs was not significant.

It can be concluded that the RH in Guangzhou was higher during the summer and at night throughout the year, and that Ea was significantly higher in the summer than in the winter. However, there was minimal difference in the daily variation.

3.2. Temporal UMI Pattern

3.2.1. Monthly Variations

The humidity differences between each urban LCZ and LCZ D were analyzed on a monthly basis (LCZ X—LCZ D). Figure 4 shows the monthly mean ΔRHX-D and ΔEaX-D during the daytime and nighttime. To further explore the statistical significance of these differences, an ANOVA was performed on the ΔRHX-D and ΔEaX-D across the six LCZs.

Table 3. ANOVA Analysis of Humidity Parameters in LCZs.

LCZ Type	ANOVA Results
	Source	Sum_sq	df	F	PR (>F)
Daytime ΔRH	LCZ	9.38774 × 104	5.0	710.2	0.0
	Residual	1.48106 × 106	56,023.0	-	-
Nighttime ΔRH	LCZ	1.07122 × 105	5.0	810.6	0.0
	Residual	1.05777 × 106	40,020	-	-
Daytime ΔEa	LCZ	626.441951	5.0	131.3	8.9519 × 10−139
	Residual	53,471.38311	56,023	-	-
Nighttime ΔEa	LCZ	785.23118	5.0	288.2	5.6970 × 10−304
	Residual	21,809.35418	40,020	-	-

presents the results of this ANOVA analysis, summarizing the statistical differences for each time period and LCZ zone, with F-values, degrees of freedom, and p-values provided for each comparison.

The results showed that the ΔRHX-D was negative for most of the year (except for LCZ 8), which indicated that the urban environment was perceived as drier than that in rural areas. This dryness was more pronounced in the summer and at night. The smallest ΔRHX-D occurred at nighttime in September (−8.7% in LCZ 3). In the comparison between urban and rural areas, for each zone except LCZ 8, the RH is lower in every period throughout the year. This trend reaches its peak in summer. The high RH in LCZ 8 throughout the year may have been due to its proximity to the water body, as the river was located within the radius of the two stations. Continuous evaporation would not only reduce the local temperature, but also increase the Ea. Although the perceived humidity in urban areas is lower than that in rural areas, in terms of the actual water vapor content, this is not the case.

The ΔEa pattern exhibited a more remarkable difference throughout the year than ΔRH. Most LCZs exhibited a stronger UMI on winter nights than on summer nights. The UMI gradually weakened with increasing temperature and eventually transformed into a UDI in August and September. Daytime data throughout the year showed that the UDI dominated during warm months. From the perspective of monthly averages, most urban areas exhibit a UMI throughout the day from January to April. In the hotter month of August, a UDI is presented throughout the day. For most months of the year, a daytime UDI and nighttime UMI are manifested.

The impervious surface area in urban environments, such as roads, buildings, and pavements, plays a significant role in the UDI effect. These surfaces absorb and store heat during the day, leading to higher temperatures and increased evaporation. This results in lower vapor pressure in urban areas compared to rural zones, contributing to the UDI effect, especially during the day. This is particularly evident in LCZ 4, where due to its largest ISF among the six zones (Table 1), a UDI occurs throughout the day throughout the year.

3.2.2. Hourly Variation

To further investigate the local-scale differences in UMI, Figure 5 presents the all-day variations in ΔRHX-D and ΔEaX-D for each LCZ during the summer (Figure 5a,c) and winter (Figure 5b,d). ΔRH and ΔEa were defined as the differences in relative humidity and vapor pressure, respectively, between urban and rural areas. The diurnal variations in ΔRH were similar across all LCZs, with the exception of LCZ 8; thus, the humidity levels of urban areas can be represented by the characteristics of most LCZs. In winter, the humidity characteristics of urban areas demonstrate the following pattern: RH is lower than that of rural areas from 00:00 to 12:00, higher from 12:00 to 18:00, and returns to a drier state after 18:00. This suggests that urban areas experience relatively higher humidity during specific daytime periods, although this condition is not consistently maintained. Throughout the day, ΔRH decreases gradually in the initial 2 h following sunrise (6:00 in the summer and 8:00 in the winter) and increases during the pre-sunset period (15:00 in the summer and the winter). Except for the significant temperature fluctuations during sunrise and sunset, the disparities in RH between the urban and rural areas remained stable and were mostly maintained at levels within a range of 0% to −10% during the night (excluding LCZ 8). This trend was observed for both summer and winter, with no distinguishable differences.

Based on the ΔEa curve, all LCZs in the summer exhibited nighttime UMI and daytime UDI. Compared with winter, a larger diurnal variation in UMII (up to −2.4 hPa to 1.28 hPa) was observed for each LCZ in the summer. During the first two hours after sunrise, a trend of initial decrease followed by an increase was observed in ΔEa across all zones. In simpler terms, this indicates a weakening of the UMI and a strengthening of the UDI. The decrease in UMII observed during sunrise could be attributed to evapotranspiration from residual dew and vast vegetation in rural areas. The evaporation process continued for 2 h after sunrise; however, the increase in the mixing layer height and strengthened upward motion resulted in a decrease in the observed Ea. The higher SVF in rural areas and increased surface cooling led to greater condensation and a continuous increase in ΔEa after sunset. Overall, whether in summer or winter, the state of higher RH in urban areas is transient. However, from the perspective of Ea, urban areas consistently exhibit greater moisture levels than rural areas throughout the winter.

3.3. UMI Frequency

According to the definition of Kuttler et al., 0 hPa < ΔEa < 0.5 hPa is defined as a weak UMI, and ΔEa > 0.5 hPa is defined as a strong UMI [31]. The frequency characteristics of three types of ΔEa (weak UMI, strong UMI, and UDI) and two types of ΔRH (ΔRH < 0 and ΔRH > 0) in the four seasons were analyzed (Figure 6). Here, a state with UMII > 0 is defined as urban moisture excess (UME).

Figure 6 shows the statistics of ΔRH and ΔEa (positive or negative) in the four seasons. The types of ΔRH did not differ significantly among the four seasons, with ΔRH > 0 accounting for an average of 26% and ΔRH < 0 accounting for an average of 74% in one season. As the ΔRH trend in a day was basically consistent (Figure 5), the seasonal pattern of RH differences between urban and rural areas was mainly reflected in the magnitude of the value, rather than the type (positive or negative). Notably, LCZ 8 consistently exhibited ΔRH values greater than zero throughout winter (Figure 5). Despite the fact that the presence of such high ΔRH values may elevate the overall average or alter the distribution of results, the impact of larger building surface areas and smaller vertical heights on the distribution of urban moisture should not be overlooked. Low-rise buildings typically exert a significant influence on wind, restricting vertical air exchange and promoting the accumulation of moisture near the ground. The ΔEa curve showed obvious differences among the four seasons. The UMEs were dominant throughout the four seasons. The frequency of strong UMIs was the highest from March to May (67%). A relatively dry period occurred from October to November, with a UDI frequency of 37% (Figure 6b).

Figure 7 shows the proportion of ΔEa types for each LCZ in the summer (Figure 7a) and winter (Figure 7b). It can be clearly observed that the UDI frequency in most LCZs (except for LCZ 4) was higher in the summer, and the frequency of weak UMI was higher in the winter. The UDI frequency decreased gradually with a decrease in the HRE of the LCZ, and compact areas had a slightly higher UDI frequency than open areas. This trend was more pronounced during the summer. The dense high-rises blocked airflow, causing heat and moisture to diffuse slowly, resulting in the formation of a high-local-RH environment, an increase in the latent heat, and a decrease in Es. This decreased the amount of moisture contained in the air. However, as the HRE decreased (LCZ 2 and LCZ 3), the effect gradually weakened.

3.4. Influencing Factors of UMI

3.4.1. UHI

To investigate the relationship between the UHII and UMII, data from LCZ 1 and LCZ D were used for analysis. Due to partial data loss during the measurement process of LCZ 1, analysis was conducted utilizing the remaining 120 typical days. Seasonal and diurnal conditions were investigated, including those in the summer daytime (Figure 8a), summer nighttime (Figure 8b), winter daytime (Figure 8c), and winter nighttime (Figure 8d).

During the daytime in the summer (Figure 8a), the majority of the scatter was distributed in the lower-left quadrant of the plot (ΔEa < 0 and UHII < 0). A weak correlation between UHI and UMI can be found in daytime in summer (Figure 8a) and winter (Figure 8c), which was weaker during nighttime. This correlation was not evident in the comparison of different seasons. During the daytime, a decrease in the intensity of the UHI implied that evapotranspiration in the suburbs was stronger than that in urban areas, which directly led to an decrease in the UMII.

3.4.2. Wind Speed

Figure 9 provides statistics on V and the corresponding ΔEa of all LCZs on typical days, with the aim of investigating the influence of wind speed magnitude on the UMI. To enhance representativeness, wind speeds from 0 to 6 m/s were selected and divided into datasets with 1 m/s increments. The ΔEa corresponding to the V at each moment is categorized and averaged under different wind speed classes. Daytime (Figure 9a) and nighttime (Figure 9b) conditions are discussed.

During the daytime, the relationship between ΔEa and V was not clearly defined. In some LCZs (e.g., LCZ 2, 4, 5), ΔEa decreased with increasing V, while in others (e.g., LCZ 1, 8), it increased. This randomness can be attributed to the varying influence of V on local humidity in different LCZs, which may be affected by the surrounding areas. While the air flow does influence temperature and evapotranspiration, thereby altering humidity levels, the underlying mechanisms remain unclear and require further data to support a more robust understanding.

At night, a more consistent positive correlation between ΔEa and V was observed in LCZs 1, 2, 3, 5, and 8, where ΔEa increased with wind speed. However, this trend exhibited changes at a “threshold”, such as a decrease in LCZs 2 and 5, and fluctuation in LCZ 3. The emergence of this threshold effect can be attributed to the fact that excessively high wind speeds result in a more pronounced dissipation to evapotranspiration.

LCZ 1 typically exhibits a higher threshold compared to LCZs 2, 3, and 5 due to its increased surface roughness and more localized airflow dynamics. This leads to the creation of more turbulence and enhanced moisture exchange at higher wind speeds, thereby allowing for the UMI to persist for a longer duration. In contrast, an explanation for the lower threshold in LCZs 2, 3, and 5 could be related to the role of evaporative cooling and the influence of UHI. In LCZs with more open or mixed building structures, there is often greater exposure to solar radiation and urban heat island effects, which leads to higher surface temperatures and greater evapotranspiration during the day. As a consequence, when wind speeds increase, they can disrupt the localized moisture accumulation faster, causing a more immediate response in humidity changes. In contrast, in denser LCZs like LCZ 1, where buildings are closer together and shaded areas are more common, there is more localized heat retention, and the wind has to be stronger before it can disrupt the more stable moisture conditions in place.

4. Discussion

4.1. Reference Rural Site

Rural areas often exhibit lower T and higher RH than urban areas due to differences in multiple surface characteristics, such as radiation (albedo and radiance), LULC, and aerodynamics (roughness) in rural areas [32,33]. It should be noted that the peak Ea usually appears several hours before and after sunrise and sunset in rural areas, as reported in many studies [11,12,17]. The appearance of these “two peaks” is related to the contraction and expansion of the urban boundary layer during sunset and sunrise, respectively. In this study, the same phenomenon was observed in the diurnal cycle of the LCZ D from June to September; however, the second peak was not pronounced.

The selection of rural reference sites directly affected the results and characteristics of the UMI. This also explains the diversity in UMIs in different cities, as the selection of reference rural sites varies across studies. In Nanjing [17], when using LCZ D as the reference rural site, a diurnal UDI and a nocturnal UMI were present in January, whereas the UDI was present throughout the day in July. In Hong Kong [34], two rural meteorological stations (one in a rural residential area and the other in a forested area) were used as reference sites. The nocturnal UMI performance was similar between the two reference sites; however, the daytime UMI was larger when the rural residential area was used as the reference site. In Novi Sad [18], LCZ A was used as the reference rural site because of its high evapotranspiration rate. The largest UDI intensity was observed in July, whereas a slight UME was observed in colder months, similar to Guangzhou.

In selecting a reference rural site, the most important aspect is to ensure that all sites are located under the same climatic conditions and terrain [35], and to preferably locate rural stations outside the range of urban climate impact [36]. If these conditions are met, the differences in atmospheric variables between urban and rural areas can be used to quantify urban climate effects [24].

4.2. Humidity-Based Indices

Studying the amount of air humidity and moisture transfer in urban environments must involve considerations of the saturation level and specific quantity of water vapor. This is because the urban vapor circulation process involves various physical processes, such as vegetation evapotranspiration [37], dew condensation [38], and precipitation [39], which cannot be completely measured by a single variable. For example, the rate of evapotranspiration is related to both RH and Ea.

Generally, the lower the RH, the faster the surface evaporation rate [40]. This is because a lower RH causes a more rapid diffusion rate of water molecules into the air and, correspondingly, a faster surface evaporation rate. However, the RH cannot directly reflect the surface evaporation rate on urban surfaces, because it only describes the content and degree of saturation of water molecules in the air. The surface evaporation rate is affected not only by RH, but also by meteorological factors, such as temperature, wind speed, and solar radiation [41]. Additionally, Ea is an important factor that affects the evaporation process. If Ea in the air is lower than Es on the water surface, the evaporation rate is relatively fast until Ea is equal to Es on the water surface, after which the evaporation rate slows down or stops. Overall, both RH and Ea affect the evaporation rate, and their combination can describe the evaporation process more accurately.

VPD describes the difference between Ea and Es, which better reflects the movement and transmission of water molecules [42]. In urban environments, dry and hot conditions can cause an increase in VPD, making it easier for water molecules to evaporate from vegetation and enter the atmosphere [43]. This leads to an increase in the evaporation rate of urban vegetation, which is particularly significant in the summer. However, for the same VPD, there may be two different Ea and Es values at different temperatures (i.e., VPD values may be the same when Ea and Es are both large or small). Therefore, by integrating RH and VPD, we can better describe the moisture content, thermal comfort, and other aspects, thus providing a more comprehensive assessment of humidity.

4.3. Humidity Characteristic

In extensive research on UMI and humidity variations, the role of urban morphology, green space coverage, and climatic conditions in influencing local humidity has been thoroughly examined. For instance, studies conducted in Beijing have shown that urban expansion leads to a decline in RH, Ea, and specific humidity within urban areas, particularly during the spring and autumn when dry conditions become more pronounced [44]. This trend is consistent with the seasonal variations in this study. Similar findings from Guangzhou and Novi Sad highlight the more pronounced UMI at night [18,20], a similar pattern also observed in this study, where the UMII is most prominent during the nighttime, particularly in high-density built-up zones such as LCZ 1 and LCZ 2.

Urban morphology features such as building density, green cover, and surface albedo have been shown to significantly influence the development of both UHI and UMI. For example, increasing green cover in LCZ 2 and LCZ 3 can notably enhance cooling and humidification effects [45]. Additionally, research on sky conditions has demonstrated that factors like solar radiation and cloud cover substantially impact local climate patterns [20]. On clear days, the UHI tends to intensify during the daytime, while overcast or rainy conditions promote moisture retention and latent heat release.

Table 4 summarizes the key findings related to UMI and humidity characteristics from the relevant studies. These findings underscore the complex interactions between urban morphology, climatic conditions, and the UMI. Building upon previous work, this research expands the understanding of canopy humidity dynamics in humid subtropical regions by exploring UMI patterns across various temporal scales and identifying key influencing factors. Based on these findings, future urban planning strategies should incorporate a holistic approach to mitigating both UHIs and UMIs. Specifically, urban designs should consider elements such as wind speed, urban morphology, and seasonal fluctuations to effectively reduce both heat and moisture imbalances and improve thermal comfort in urban areas.

Table 4. Summary of urban humidity researches.

Ref.	Location	Theme	Method
[12]	Turin	Temporal—spatial characteristics of UDIs and the impact of wind	Fixed measurement
[16]	Hong Kong	Mechanism of synergistic UHIs and UMIs	Numerical modeling
[18]	Novi Sad	Relationship between T and RH, AH, Ea, q, VPD, and the humidity characteristics under special weather conditions	Fixed measurement
[20]	Guangzhou	Characteristics and mechanisms of UMIs under different urban morphology and sky conditions	Fixed measurement
[34]	Hong Kong	Effects of urbanization (ventilation and land use) on humidity environments and temporal—spatial characteristics of UMIs	Numerical modeling
[44]	BTHUA	Long-term trend of atmospheric humidity and the urban expansion effect	Fixed measurement
[45]	Beijing	Evaluating the cooling and humidifying effects of urban vegetation coverage across different LCZs	Fixed measurement
[46]	China	Impact of urbanization on the humidity environment and mechanisms of climate change that affect UDIs and UMIs	Fixed measurement
[47]	Beijing	Seasonal and spatial characteristics of UMIs and UDIs, and their causes	Numerical modeling
[48]	15 cities	Differences in ΔT and Δq between the urban and rural sites (LCZ X − LCZ D) were analyzed on multiple time scales (diurnal, seasonal, and annual)	Fixed measurement
[49]	Barcelona	The impact of model resolution and urban land use information (LCZs) on humidity predictions	Numerical modeling

Note: T: temperature (°C); RH: relative humidity (%); AH: absolute humidity (g/m³); Ea: vapor pressure (hPa); q: humidity ratio (g/kg); VPD: vapor pressure deficit (hPa).

Table 4. Summary of urban humidity researches.

Ref.	Location	Theme	Method
[12]	Turin	Temporal—spatial characteristics of UDIs and the impact of wind	Fixed measurement
[16]	Hong Kong	Mechanism of synergistic UHIs and UMIs	Numerical modeling
[18]	Novi Sad	Relationship between T and RH, AH, Ea, q, VPD, and the humidity characteristics under special weather conditions	Fixed measurement
[20]	Guangzhou	Characteristics and mechanisms of UMIs under different urban morphology and sky conditions	Fixed measurement
[34]	Hong Kong	Effects of urbanization (ventilation and land use) on humidity environments and temporal—spatial characteristics of UMIs	Numerical modeling
[44]	BTHUA	Long-term trend of atmospheric humidity and the urban expansion effect	Fixed measurement
[45]	Beijing	Evaluating the cooling and humidifying effects of urban vegetation coverage across different LCZs	Fixed measurement
[46]	China	Impact of urbanization on the humidity environment and mechanisms of climate change that affect UDIs and UMIs	Fixed measurement
[47]	Beijing	Seasonal and spatial characteristics of UMIs and UDIs, and their causes	Numerical modeling
[48]	15 cities	Differences in ΔT and Δq between the urban and rural sites (LCZ X − LCZ D) were analyzed on multiple time scales (diurnal, seasonal, and annual)	Fixed measurement
[49]	Barcelona	The impact of model resolution and urban land use information (LCZs) on humidity predictions	Numerical modeling

Note: T: temperature (°C); RH: relative humidity (%); AH: absolute humidity (g/m³); Ea: vapor pressure (hPa); q: humidity ratio (g/kg); VPD: vapor pressure deficit (hPa).

5. Conclusions

This research utilized 3-year fixed measurements to study the characteristics of RH, Ea, and VPD across eight LCZs over varying time spans, with a particular focus on the temporal patterns of urban–rural differences, as well as the factors that influence UMIs. In most LCZs, RH, Ea, and VPD were higher in summer, with RH being higher during the nighttime, while the differences in Ea and VPD between the daytime and nighttime were not significant. From the perspective of urban–rural differences, the RH at urban stations was lower than that at rural stations every month, with this trend being most pronounced during summer nights. The UMI was more pronounced during winter nights and then transformed into a UDI during the summer. Regarding daily variation, ΔRHX-D increased after sunrise and decreased before sunset, with no notable seasonal features. In terms of the factors influencing the UMI, it was discovered that (a) a weak correlation between the UHI and UMI can be observed during the day in both winter and summer, and (b) nocturnal UMII increased with an increase in wind speed. The frequency analysis revealed that the highest frequency of strong UMIs occurred from March to May, whereas the highest frequency of UDIs occurred from October to November. Finally, the frequency of UDIs decreased with a decrease in the HRE, and compact areas had a higher frequency of UDIs than open areas. The identified characteristics of UMIs and UDIs cannot be universally classified as “good” or “bad”. Instead, they should be considered in the context of urban planning aimed at creating comfortable spaces. Therefore, managing key factors influencing UMIs, such as the UHI and wind dynamics, is essential. For instance, by optimizing building layouts, urban planners can regulate wind distribution across seasons, ensuring that wind speed remains within optimal thresholds for each LCZ to control temperature and humidity, thereby enhancing thermal comfort.

Although this study analyzed the canopy humidity for each type of climate zone, the conclusions drawn may have certain limitations in terms of applicability. For example, the observed data obtained from different sites may vary, and the impact of natural and human factors on humidity can lead to errors and biases in the data. Based on the different humidity characteristics of the various LCZs in Guangzhou examined in this study, a more accurate humidity model could be established for further research. Building on the LCZ, more intricate and detailed models can be developed by further incorporating factors such as topography, vegetation, and land use, leading to improved accuracy in prediction and diagnosis.