Climate change and tourism demand: Risks for extreme heat?

Abstract

Extremely high temperatures have become a major hurdle to tourists' experiences in various destinations. In this study, we present the first nationwide estimation of the temperature-tourism response function using a fine-grained panel dataset constructed by matching tourism data and daily meteorological datasets for 280 cities from 2005 to 2019 in China. Results show a non-linear relationship between temperature and tourism employing the temperature bins method, which is not sensitive to response equations. In addition, we conduct a rich heterogeneity analysis in the article to further analyse the effects of other external conditions on the temperature-tourism relationship, including the destination's level of tourism economic development, population, infrastructure development, and tourism resource endowment. We also observed that increasing summer temperatures result in decreased tourism arrivals and revenue, while rising autumn temperatures lead to increased tourism arrivals and revenue. This research contributes to a deeper understanding of the complex relationship between temperature dynamics and tourism patterns, offering insights into destination management and adaptation strategies.

1. Introduction

Since the middle of the 21st century, the frequent occurrence of extreme weather caused by global climate change has become more serious due to the influence of human activities. Tourism, as a typical highly weather-sensitive industry, is also deeply affected by climate warming, which alters natural geographic and socio-economic conditions, thus impairing the carrying capacity of tourist destinations, reducing destination comfort, and thus affecting tourist behaviour [[1], [2], [3]]. As a strategic pillar industry of the national economy, tourism is also an important engine for realizing the high-quality development of the local economy in China. Therefore, as tourism is an important sector for enhancing the country's cultural soft power, assessing the impact of temperature on tourism has significant practical implications. Due to the importance of tourism in enhancing China's cultural soft power, evaluating the impact of temperature on the tourism industry holds significant practical significance. This aids in enhancing the tourism sector's identification and management of climate risks, and ensuring the resilient development of tourism. Factors that constrain the development of tourism, including infrastructure [4], economy [5], supportive policies [6], information & communication technology [7], and natural factors such as air pollution [8], are widely discussed. However, the effect of climate change (temperature) on tourism in developing countries has not been quantitatively studied. This study contributes to the enrichment of relevant research at the intersection of climate change and tourism, providing empirical evidence from China.

For destinations with nature-based tourism, the impacts of climate change are significant [5]. In nature-based destinations, such as national nature-protected areas, weather is an important factor for tourists to make visiting decisions. When weather conditions do not conform to tourists' expectations, original support for the destination is withdrawn and another destination choice is made [9]. Thus, tourism strategies are needed to offset the loss of tourism competitiveness [10]. Despite perceiving the negative effect of climate change on destination attractiveness, tourists’ expected experiences can be improved by local destination conservation efforts; however, adaptation measures are inadequate to attract tourists under climate change [3].

Established studies have identified the effects of high temperatures on agricultural output, industrial output, labour productivity, energy demand, economic growth, and other outcomes [[11], [12], [13], [14]]. High temperatures will lead to discomfort, fatigue, and cognitive impairment [[15], [16], [17]]. In this case, people are likely to abandon tourism activities and reduce their need for tourism [5]. Due to China's distinct four seasons, estimating the effects of temperature shocks on tourism using annual average temperatures may ignore seasonal variations, potentially resulting in a downward bias in the results. Therefore, to accurately capture the relationship between temperature and tourism after estimating the linear effect of the former on the latter, a more targeted empirical strategy for the response of tourism to temperature shocks must be developed. According to Chen and Yang [18], the seasonal mean temperature can be used as a variable to investigate whether temperature has a distinct effect on tourism arrivals and revenue. Then, referring to Zhang et al. [13], we first identified the nonlinear relationship between temperature and tourism using the temperature bin method, which is insensitive to the response equation. The results indicate that tourism arrivals and revenue decline when the temperature exceeds 15–20 °C; furthermore, the degree of decline increases with increasing temperature.

The effect of climate change on the natural resource endowments of destinations is continuous [19,20]. Therefore, after proving that temperature shock has a nonlinear effect on tourism, we further investigated the effect of lagged temperature on tourism arrivals and revenue in the current year. The results indicate that the lagged temperature also has a negative effect on tourism arrivals and revenue as the temperature increases. As the lag phase increases, the negative effect of the high temperature gradually decreases.

Our study primarily examines the causal impact of temperature shocks on tourism demand, aiming to explore its implications for tourism planning and management. Our empirical strategy is displayed in Fig. 1. The main objectives and contributions of this paper are summarized as follows:

Fig. 1.

The research framework.

First, given the scarcity of studies examining the nonlinear effect of temperature on tourism demand, this paper fills this gap by constructing a fine-grained dataset for causal estimation. We constructed the dataset by matching the city-level tourism data with daily meteorological data. To ensure the accuracy of the estimation results, precipitation, sunshine hours, atmospheric pressure, relative humidity and wind speed are used as weather control variables.

Furthermore, this paper not only investigates the short-term effects of temperature from both linear and nonlinear perspectives, but also the long-term dynamic effects of temperature through delayed and cumulative effects. The relationship between temperature and tourism is complex and can be influenced by various other factors. Therefore, we indirectly argue the potential mechanism of temperature on tourism demand through heterogeneity analysis. We study the heterogeneity of the impact of temperature on tourism demand from several aspects to explore the differential manifestations of temperature impacts under different conditions, including different economic levels, population densities, tourism resource endowments, and transportation infrastructures, respectively. The heterogeneity analysis provides a reference basis for different regions to develop targeted strategies to cope with climate change. The results reveal that if a tourism destination city has a higher economic level, its ability to resist high temperatures is stronger. Transportation infrastructure construction and a city's tourism resource endowment positively affect the destination city.

Finally, this study provides insights into the impacts of temperature on tourism by combining the economics of climate change with tourism management from an interdisciplinary research perspective. Our findings also have significant policy implications. A comprehensive understanding of the relationship between temperature and tourism demand is essential for policymakers, industry professionals, and researchers to develop effective strategies to mitigate the negative impacts of adverse climate change and ensure the sustainable management of tourism.

2. Literature review

Climate change is now a serious topic worldwide. The concentration of carbon dioxide in the atmosphere exceeded 420 ppm in February 2022, reaching the same level as 3.5 million years ago [21]. Extreme low temperatures in the United States, ‘triple-dip’ La Niña, frequent rainstorms in Brazil, the heaviest rainfall of the last 60 years in South Africa, and severe drought in Kenya prove that the climate crisis is a current issue. Statistically, 85 % of the entire population in 80 % of the world's land area suffers from extreme weather conditions caused by human-induced climate change [22].

Climate change has prominent effects, and it jeopardises the global economy [23]. The most significantly affected attributes are human health, food, water, infrastructure, economy, and security [24]. Tourism, an essential economic sector, is climate-sensitive and affected by many types of climate hazards. Given the temperature rise, which is three times faster than the average global warming rate, Arctic tourism faces challenges due to climate change. Consequently, Arctic tourism is the last opportunity to promote this disappearing area and reinvented post-Arctic tourism is urgently required [25]. In addition, the vulnerability of coastal tourism partly comes from climate-related drivers such as sea-level rise, ocean acidification, coral bleaching, and increased storm frequency and drought [26]. As a result of the coral bleaching caused by climate change, the Great Barrier Reef is being endangered and in the process of being on UNESCO's ‘in danger’ list, which will further reduce its attractiveness and negatively affect the local tourism economy [27]. For national forest parks in Taiwan, climate change in the form of increased temperature and decreased rainfall has a compound effect and the natural landscape may disappear because of climate change [28]. In the midwestern United States, climate change has a significant effect on the ski industry. Rising temperatures significantly shorten the ski season, resulting in higher skier intensity, which requires a more sophisticated balance between higher operational costs and increased revenue per acre. Furthermore, it is challenging to rigorously oversee and manage water usage related to snowmaking [29].

Climatological tourism is a well-established area in tourism research. Weather and climate are major drivers of travel, shape tourist experiences, and are part of destination images [30]. First, weather affects individuals’ experiences and thus the way they perceive regions and destinations, travel within a place or between regions, and adapt to emotional and physical weather impacts [31]. Consequently, the ways people respond and adapt to various weather circumstances are, to some degree, a function of perceptions of climate and weather, especially, the parts deemed important [32].

Weather attractiveness can be neutral, such as a mist cloud blocking a distant view or clouds, allowing an impressive peak in the valley [33]. To measure weather attractiveness, objective indices of physiologically equivalent temperature can be utilised [34]; subjective research via preference surveys also makes sense [35]. In northern European countries, warm and stable weather significantly contributes to the push and pull factors of local tourism. Furthermore, as an important part of the destination image, satisfactory weather is a significant factor affecting tourist loyalty and thus stimulates the behaviour of revisiting a destination [36].

Behavioural and psychological adaptations are related to choosing destinations, adaptive behaviour, and emotional adjustment. Behavioural adaptation includes the utilisation of weather forecasts, selection of clothing, and adjustment of travel schedules [30,37]. Adaptation to severe weather and even climate hazards also requires further efforts by tourism organisations to spread information, considering the universally insufficient cognition of practical and important weather information [38]. In terms of psychological adaptation, tourists cope with rough weather by having active emotions at a personal or interpersonal level; negative acceptance is another pattern of adaptation. Furthermore, given the differences in tourists’ characteristics and attitudes toward climatic conditions, their concerns about climate change and intentions for adaptive actions may vary; it is unlikely that all tourists have similar opinions [39].

Changing weather conditions are not always negative for tourism and opportunity and threat can simultaneously occur at different spatial locations [40]. For instance, with an increase in temperature resulting from global warming, the tourism industry in Norway faces an opportunity for milder and warmer days in summer, which is related to tourists' improved thermal comfort [41]. Similarly, Austria's tourism industry benefits from warmer and sunnier weather conditions. Both domestic and foreign tourists have stronger intentions to spend their holidays here, but this impact is not as significant in the short run as in the long term, especially for foreign tourist visits [42]. Regarding the hotel sector in Taiwan, higher temperatures and more shiny days interact with tourists' price sensitivity and strengthen the correlation between the number of tourists and price [43]. When the temperature exceeds the inflection point, it can negatively affect tourists' travel motivation. If destinations are already warm, a ‘too warm’ perception of tourists may suddenly change their intention for travelling [3]. Weather and specific climate conditions have important effects on the travel demand and tourist behaviour and they respond to climate change [2].

Among all stakeholders involved in the effect of climate change on tourism, tourists are the most capable of adapting to three resources, that is, money, knowledge, and time, which enables them to evade the negative effects of climate change by changing their travel plans [3]. When weather conditions do not conform to tourists' expectations, original support for the destination is withdrawn and another destination choice is made [9]. For example, half of the tourists interviewed in Florida indicated that they would change their destination or time of arrival because of unacceptably high temperatures and frequent storms [3]. In Spain, the southern coastal provinces are losing their tourism market share to high temperatures to which tourists are sensitive. Thus, tourism strategies are needed to offset the loss of tourism competitiveness [10]. Furthermore, unacceptable conditions are mostly similar but may vary depending on tourists’ nationality, preferences, and destinations.

Panel methods have been adopted in an increasing number of studies to examine the effects of climate change on the economy. These studies are based on changes in the understanding of weather over time within specific spatial areas and show the effect on agricultural output, industrial output, labour productivity, energy demand, economic growth, and other results [[11], [12], [13], [14]]. However, the effect of climate change on tourism remains unclear. This study aimed to fill this gap using the temperature bin method to explore the annual effect of temperature on Chinese cities. This is a new attempt, which has the potential to lead to significant innovations.

3. Data

3.1. Dependent variables

Tourism Data. Considering the data availability, we used domestic tourism arrivals and domestic tourism revenues of 280 cities from 2005 to 2019 as proxy variables for tourism development in Chinese cities. Domestic tourism arrivals represent the number of residents visiting a city for sightseeing. Visitors include both overnight and one-day tourists. Domestic tourism revenue refers to the total money spent by domestic tourism arrivals on sightseeing. Table 1 presents the summary statistics for the samples used from 2005 to 2019. Our data were obtained from the city's Tourism Statistical Yearbook and Tourism Bulletin. Domestic tourism arrivals reflect the scale of tourism development in a destination city and domestic tourism revenue reflects the economic benefits of tourism in a city. Thus, these two datasets provide a comprehensive measure of a city's tourism development.

Table 1.

Descriptive statistics.

Variable	Obs	Mean	Min	Max	Source
Tourism Data
Domestic tourism arrivals (thousand people)	4161	27136.91	110	652966.90	TSY
Domestic tourism revenue (million yuan)	4161	19368.36	22.08	360179.31	TSY
Weather Data
Temperature (°C)	4161	17.25	3.71	31.12	NMIC
Relative humidity (%)	4161	68.85	36.44	88.76	NMIC
Average air pressure (hPa)	4161	967.22	620.75	1017.05	NMIC
Sunshine duration (h)	4161	1985.38	736.34	3389.14	NMIC
Precipitation (mm)	4161	1032.91	52.55	3030.56	NMIC
Average wind speed (mile/s)	4161	2.14	0.997	4.69	NMIC

Notes: The unit of observation is a city-year. Tourism indicators were obtained from the Tourism Statistical Yearbook and City Tourism Statistical Bulletin. The weather data were sourced from the National Meteorological Information Center.

To better visualize the distribution of domestic tourism arrivals and domestic tourism revenues across cities, we have plotted Fig. 2, which shows the distribution of average domestic tourism arrivals revenues at the city level, respectively. The figure reveals that eastern seaboard cities and some northern cities have higher annual average domestic tourism arrivals and revenues than other cities.

Fig. 2.

Average annual domestic tourism arrivals and revenues for each city from 2005 to 2019.

Note: Average annual domestic tourism arrivals and revenues were obtained from the city's Tourism Statistical Yearbook and Tourism Bulletin.

3.2. Independent variables

Temperature. We constructed multiple temperature variables to thoroughly examine the effects of temperature shocks on tourism in destination cities. Table 1 reports the summary statistics for the temperature from 2005 to 2019. The National Climatic Data Center of the China Meteorological Administration was the source of the temperature data. First, we used the daily average temperature values of the city from 2005 to 2019 as the core temperature variables for the city on that day. The annual average temperature was then obtained by aggregating the daily average temperatures. Because China has more distinct seasons, the use of annual average temperatures leads to biased estimates. To address this issue, we referred to the approach of Zhang et al. [13] and constructed temperature bins as alternative temperature variables. By using the temperature bin method, we can visualize the nonlinear effects between temperature and tourism development.

3.3. Control variables

Weather. Because other weather variables can affect tourism, we used precipitation, sunshine duration, relative humidity, air pressure, and wind speed as control variables. Summary statistics of the weather variables are presented in Table 1. Except for precipitation and sunshine hours, which were constructed as daily totals, relative humidity, air pressure, and wind speed were constructed as daily averages. Weather control variables were obtained from the National Climatic Data Center. By aggregating the daily data, we obtained the annual average weather variables for the city. Following the construction of daily climate variables, precipitation and sunshine hours were calculated as annual totals.

4. Empirical approach

Three different methods were used to construct the temperature variables to examine the effects of temperature on tourism development. These methods are described in this section.

4.1. Linear regression model

Our common sense and previous studies suggest that weather variables vary randomly. Temperature is a typical exogenous shock during the construction of specifications. Because the effects of weather shocks on the economy have been extensively demonstrated in the literature [18,44,45], our control variables include only exogenous weather variables such as precipitation, sunshine duration, relative humidity, wind speed, and air pressure. This ensures the credibility of our estimation results. We estimated the following econometric model:

$${\mathit{ln}Y}_{it}={\alpha }_0+{\beta }_1{T}_{it}+{\beta }_2{W}_{it}+u_i+v_t+{\epsilon }_{it}\text{,}$$ (1)

where cities are indexed by $i$; years are indexed by $t$; $Y_{it}$ is our outcome variable of interest, which represents domestic tourism arrivals and domestic tourism revenue; ${\mathit{ln}Y}_{it}$ denotes the natural logarithm of the outcome variable; $T_{it}$ is the average annual temperature, which is the core explanatory variable; vector $W_{it}$ includes a set of credible city-level weather control variables; and ${\beta }_1$ is the major coefficient of interest.

We used a two-way fixed effects model to estimate the effect of temperature shock, controlling for characteristics at the time and space levels. The variable $u_i$ represents city fixed effects at the spatial level, which are used to control uncaptured features specific to tourism destination cities. The variable $v_t$ denotes the year fixed effect, which captures the variation that does not vary among individuals for a given year. Our fixed effects are controlled for the tightest dimensions of time and space to ensure the robustness of our results. To address the spatiotemporal correlations of the error term, we clustered it at the city level.

4.2. Nonlinear regression model

The linear regression model estimated the average effect of temperature on tourism, thus providing a relatively intuitive assessment of the relationship between temperature and tourism. However, the use of annual average temperature simplified the complexity of temperature dynamics, failing to capture the seasonal and daily variations in temperature. However, the impact of climate change on tourism often manifests through seasonal variations and extreme temperature events. Therefore, we conjectured a nonlinear relationship between temperature and tourism and used two other temperature variables to estimate the nonlinear effects of temperature shocks.

4.2.1. Seasonable average temperatures as temperature variables

Most of China exhibits distinct seasonal changes. Spring begins in March and ends in May; summer includes June, July, and August; autumn includes September to November; and winter includes December, January, and February. After demonstrating the effect of the annual average temperature on tourism, we constructed a seasonal average temperature to determine whether the temperature has a seasonal effect on tourism. To accurately capture the correlations among temperature, tourism revenue, and arrivals, we constructed temperature variables by season, denoted by $T_{it}^s$. The model can be expressed as follows:

$${\mathit{ln}Y}_{it}={\alpha }_0^s{T}_{it}^s+\sum _{L\ge 1}{\alpha }_L{T}_{i,t-L}^s+\gamma W_{it}+u_i+v_t+{\epsilon }_{it}\text{,}$$ (2)

where $s$ represents the four seasons: spring, summer, autumn, and winter. We also incorporated the lags of the temperature variable in the model to test the effect of temperature on tourism in previous years. Parameter $L$ represents the number of lags in the temperature variable. We controlled for city weather control variables in the specification to exclude the effects of weather variables on the coefficient estimates. We also controlled for city and year fixed effects to ensure the robustness of the estimates. Standard error clustering was performed at the city level. By using a semi-log model, the estimated coefficients of the temperature variable can be interpreted as $\mathrm{\alpha }$* 100 % change in tourism revenue and arrivals for each 1 °C increase in the seasonal average temperature. Our null hypothesis is that the seasonal average temperature does not affect tourism.

4.2.2. Temperature bins as temperature variables

We used the temperature bin method to obtain the specific response function of the tourism industry to temperature shocks. This approach enabled us to overcome the inherent limitations of relying on models. We referred to the method reported by Zhang et al. [13] and defined the temperature variable more specifically as a temperature bin to estimate the effect of the daily average temperature on the annual outcome variable. Based on this method, a set of temperature bins is established that discretises the average daily temperature over an annual distribution. The temperature bin method provides more intuitive evidence for studying the effects of temperature shock.

$${\mathit{ln}Y}_{it}=\sum _m{\beta }^m{T}_{it}^m+\gamma W_{it}+u_i+v_t+{\epsilon }_{it}$$ (3)

The temperature variable $T_{it}^m$ represents the number of days in which the daily average temperature of city $i$ falls within the $m$ th temperature bin in year $t$. We divided the average daily temperature of each city into 10 bins in 5 °C increments. For example, $T_{it}^1$ is the number of days when the daily average temperature of city $i$ is below −5 °C in year $t$. Fig. 3 shows the distribution of daily average temperatures from 2005 to 2019. The height of each bar represents the number of days per year in which the average daily temperature fell within each temperature bin during the study period.

Fig. 3.

Daily temperature distribution (2005–2019).

Note: The daily temperature distribution is averaged across all cities and years.

Parameter $W_{it}$ represents the weather control variables for city $i$ in year $t$ including the precipitation, relative humidity, wind speed, air pressure, and sunshine duration. We used city fixed effects to control urban characteristics that do not change over time and annual fixed effects to control for regionally common annual shocks such as regional tourism policies. Based on these specifications, we could accurately identify the effects of temperature shocks on tourism arrivals and revenues and exclude other potential disturbances.

To prevent the problem of multiple covariance, we deleted the fifth temperature bin [10 °C, 15 °C) as the reference bin of the specification. We selected the fifth temperature bin as the reference bin because it was in the middle of our temperature range. Our results were independent of this choice. The coefficient of interest is ${\beta }^m$, which represents the marginal effect of adding an additional day to the temperature bin $m$ relative to a day in the reference bin.

5. Empirical results

5.1. Linear regression model

In this section, we present the results of the baseline estimation for Specification (1). In Table 2, Columns (1)–(3) estimate the effect of the annual average temperature on the domestic tourism arrivals, and columns (4)–(6) show the effect of temperature on revenue.

Table 2.

Estimated effects of temperature on tourism arrivals and revenue.

	Domestic tourism arrivals			Domestic tourism revenue
	(1)	(2)	(3)	(4)	(5)	(6)
Temperature	−0.0672a	−0.0809a	−0.0645a	−0.0806a	−0.0604b	−0.0563c
	(0.0142)	(0.0179)	(0.0203)	(0.0223)	(0.0294)	(0.0308)
Weather Controls	N	Y	Y	N	Y	Y
City FE	Y	Y	Y	Y	Y	Y
Year FE	Y	Y	N	Y	Y	N
Region-by-year FE	N	N	Y	N	N	Y
Observations	4161	4161	4161	N	Y	Y
Adj. $R^2$	0.9566	0.9567	0.9658	0.9443	0.9448	0.9564

Notes: Standard errors are clustered at the city level and reported in parentheses.

^ap < 0.01.

^bp < 0.05.

^cp < 0.1.

Our data show that, when other weather variables are excluded, each 1 °C increase in temperature reduces domestic tourism arrivals by 8.09 % and revenue by 6.04 %. The magnitude is large. The baseline regression uses year fixed effects to control annual shocks that do not vary by region at the national level. However, each province has a degree of autonomy over local tourism policies. Therefore, we used region-by-year fixed effects to control common shocks in each year of specification for each province to further ensure the robustness of our results. Each 1 °C increase in temperature leads to a 6.45 % decrease in tourism arrivals and a 5.63 % decrease in tourism revenue.

The linear results highlight that an increase in temperature harms the tourism economy. In the context of global warming, the tourism industry faces declining tourism arrivals and, consequently, shrinking tourism revenues. As temperatures rise, the appeal of many tourism destinations diminishes, directly affecting visitor numbers and economic performance. The negative correlation between temperature increases and tourism metrics underscores the vulnerability of the tourism sector to climate variability. This effect can be particularly pronounced in regions where tourism is a key economic driver, leading to broader economic repercussions such as job losses and reduced local income. However, because the temperature variable was averaged over a full year of daily temperatures, the extremes were not represented. Based on this conclusion, we estimated the nonlinear effect of temperature on tourism to provide more reliable theoretical support for tourism-related sectors and practitioners to cope with global warming.

5.2. Nonlinear regression model

5.2.1. Seasonal average temperatures as temperature variables

Most cities in our sample had widely varying temperatures during all seasons. Therefore, in this section, we used the seasonal average temperature as the core explanatory variable. Specification (2) was used to estimate the utility of temperature without adding lagged terms. By estimating Specification (2), we can determine the effect of temperature on tourism in different seasons. The results are shown in Table 3. Columns (1) to (4) show the coefficients of the estimated temperature shocks on domestic tourism arrivals. Columns (1) and (2) show the baseline regression results using city and year fixed effects. Columns (3) and (4) test the robustness of our baseline results using city and region-by-year fixed effects. The effect of the high temperature is evident in these four columns. Similarly, columns (5)–(8) estimate the effect of temperature on the domestic tourism revenue.

Table 3.

Estimated effects of seasonal temperature on tourism arrivals and revenue.

	Domestic tourism arrivals				Domestic tourism revenue
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
${\text{Temperature}}^{\text{spring}}$	−0.0388a	−0.0474a	−0.0199b	−0.0214b	−0.0416a	−0.0359a	−0.0300a	−0.0196c
	(0.0067)	(0.0073)	(0.0093)	(0.0087)	(0.0090)	(0.0097)	(0.0106)	(0.0101)
${\text{Temperature}}^{\text{summer}}$	−0.0160a	−0.0246a	−0.0262a	−0.0301a	−0.0246a	−0.0251a	−0.0355a	−0.0322a
	(0.0057)	(0.0072)	(0.0078)	(0.0089)	(0.0074)	(0.0084)	(0.0104)	(0.0107)
${\text{Temperature}}^{\text{fall}}$	−0.0000	−0.0045	0.0019	0.0007	−0.0093	−0.0059	−0.0053	0.0014
	(0.0079)	(0.0079)	(0.0086)	(0.0092)	(0.0104)	(0.0110)	(0.0113)	(0.0131)
${\text{Temperature}}^{\text{winter}}$	−0.0077	−0.0047	−0.0110	−0.0100	0.0036	0.0058	0.0128	0.0120
	(0.0080)	(0.0077)	(0.0109)	(0.0107)	(0.0145)	(0.0144)	(0.0193)	(0.0188)
Weather Controls	N	Y	N	Y	N	Y	N	Y
City FE	Y	Y	Y	Y	Y	Y	Y	Y
Year FE	Y	Y	N	N	Y	Y	N	N
Region-by-year FE	N	N	Y	Y	N	N	Y	Y
Observations	4161	4161	4161	4161	4161	4161	4155	4155
Adj. $R^2$	0.9569	0.9570	0.9622	0.9623	0.9445	0.9450	0.9516	0.9522

Notes: Standard errors are clustered at the city level and reported in parentheses.

^ap < 0.01.

^bp < 0.05.

^cp < 0.1.

Table 3 documents that in spring and summer, a 1 °C increase in the average seasonal temperature is associated with a significant decrease in domestic tourism arrivals and revenues. However, in fall and winter, the increase in temperature has no effect on tourism arrivals or revenue. The seasonal analysis reveals the nuanced impact of climate change on the tourism economy, varying significantly by season. During summer, when temperatures often exceed comfortable levels for outdoor activities, further increases in temperature exacerbate the negative impact on tourism. This seasonal sensitivity indicates that summer tourism is particularly vulnerable to rising temperatures, potentially leading to shorter stays, reduced visitor numbers, and lower revenues for businesses reliant on summer tourism. Therefore, when the average summer temperature rises, it potentially causes tourists to choose to cancel their travel plans. In contrast, the lack of significant effects in fall and winter suggests that moderate temperature increases during these cooler seasons do not deter tourists. It implies that popular destinations in fall and winter might not be subject to the same degree of economic stress from rising temperatures as summer hotspots. Spring presents a unique case where the expected positive impact of pleasant weather is countered by post-holiday economic and social factors. Our results show that an increase in spring temperatures leads to a decline in tourism. This may be caused by the Chinese New Year holiday, which just ended, leading to a large number of employees returning to work. Therefore, when the temperature rises in spring, the domestic tourism arrivals tend to decline. Whether we used year fixed effects or region-by-year fixed effects, we obtained similar results, proving that our results are very robust.

5.2.2. Temperature bins as temperature variables

After studying the effect of temperature on tourism during different seasons, we determined the nonlinear relationship between temperature and tourism development using temperature bins as temperature variables. The results are summarized in Table 4. Our specification includes all temperature bins, except for the reference bin. Because of space constraints, we only show the five bins that exceed the reference bin in the chart. Based on the omission of the fifth temperature bin as the reference bin, the coefficients of the temperature bins above 15 °C are all negative.

Table 4.

Estimated effects of temperature on tourism arrivals and revenue.

	Domestic tourism arrivals		Domestic tourism revenue
	(1)	(2)	(3)	(4)
25–30 °C	−0.0044a	−0.0023b	−0.0044a	−0.0038a
	(0.0012)	(0.0013)	(0.0011)	(0.0013)
30–35 °C	−0.0073a	−0.0049a	−0.0064a	−0.0047a
	(0.0014)	(0.0015)	(0.0014)	(0.0017)
>35 °C	−0.0054a	−0.0034b	−0.0050a	−0.0022
	(0.0018)	(0.0014)	(0.0018)	(0.0022)
p-value of joint test	0.0000	0.0000	0.0004	0.0004
Weather Controls	Y	Y	Y	Y
City FE	Y	Y	Y	Y
Year FE	Y	N	Y	N
Region-by-year FE	N	Y	N	Y
Observations	4161	4161	4161	4155
Adj. $R^2$	0.9578	0.9455	0.9455	0.9182

Notes: The model includes all temperature bins. The 10–15 °C bins were omitted as reference bins to avoid multicollinearity. Standard errors are clustered at the city level and reported in parentheses.

^ap < 0.01, **p < 0.05.

^bp < 0.1.

We visualized the results in Table 4 in Fig. 4, Fig. 5, which show the temperature–tourism response functions. Fig. 4 depicts the correlation between temperature and domestic tourism arrival. Our analysis reveals that replacing a day in the [10 °C, 15 °C) bin by a day with temperatures above 35 °C decreases domestic tourism arrivals by 0.54 %. It is evident that high temperatures negatively affect domestic tourism arrivals. In contrast, the effect of low temperature is relatively small, with a coefficient close to zero. This indicates that extreme heat has a negative effect on tourism, whereas cold weather has a near-zero effect.

Fig. 4.

Estimated effects of the daily temperature on domestic tourism arrivals.

Fig. 5.

Estimated effects of the daily temperature on domestic tourism revenue.

In Fig. 5, we examine the effect of the temperature shock on domestic tourism revenue, finding a similar pattern to that of domestic tourism arrivals. Our results show that a one-day increase in the >35 °C bin leads to a 0.50 % decrease in domestic tourism revenue. We used the national domestic tourism revenue in 2019 to calculate the value loss due to high temperatures. The domestic tourism revenue of China in 2019 was $5725 billion. The calculation shows that an additional day with a temperature exceeding 35 °C results in a loss of $286 million in China's domestic tourism revenue at the aggregate level. This figure implies that the loss of tourism due to high temperatures is enormous and that appropriate countermeasures are needed to mitigate this negative impact. We also report the F-statistic to test the null hypothesis that all temperature bin coefficients are jointly zero.

To eliminate potential bias resulting from common policies, we incorporated city and year fixed effects into Columns (1) and (3) of our analysis. In Columns (2) and (4), we account for heterogeneity in year-specific shocks using a year–region fixed effect instead of a year fixed effect. The results of both specifications demonstrate a consistent negative effect of temperature shock on tourism and the estimated coefficients are very similar. These findings provide robust evidence supporting our findings.

6. Dynamic effect

In previous discussions, we demonstrated that temporary extreme temperatures affect tourism numbers and revenue. One potential explanation for this phenomenon is that extreme temperatures can alter tourists’ travel decisions. The effect of extreme temperatures is cumulative over time, indicating that previous temperatures likely affect current tourism revenue and arrivals. Overlooking this delayed effect may result in underestimating the effects of extreme temperatures on tourism. To determine whether there is a delayed effect of extreme temperatures, we included lagged temperature variables for the last two years to account for dynamic effects. The results are presented and interpreted in this section.

6.1. Delayed temperature effects

The results of the linear model with lagged temperature variables are presented in Table 5. The sign and magnitude of the estimated coefficients of current-period temperature variables are roughly similar regardless of whether the model includes lagged terms. The effect of the current period's temperature on tourism dominates the temperature–tourism relationship. The previous year's temperature negatively affected tourism less than the current temperature.

Table 5.

Estimated delayed and cumulative effects of temperature on domestic tourism arrivals and revenue.

	Domestic tourism arrivals			Domestic tourism revenue
	Model with no lag	Model with one lag	Model with two lags	Model with no lag	Model with one lag	Model with two lags
	(1)	(2)	(3)	(4)	(5)	(6)
Temperature	−0.0809***	−0.0785***	−0.0805***	−0.0604**	−0. 0632***	−0.0813***
	(0.0179)	(0.0144)	(0.0156)	(0.0294)	(0. 0184)	(0.0180)
L1: Temperature		−0.0515***	−0.0537***		−0.0437	−0.0468**
		(0.0151)	(0.0118)		(0.0277)	(0.0199)
L2: Temperature			−0.0440***			−0.0343
			(0.0164)			(0.0237)
Sum of all Temperature coefficients	−0.0809***	−0.1300***	−0.1781***	−0.0604**	−0.1069**	−0.1625***
	(0.0179)	(0.0258)	(0. 0327)	(0.0294)	(0.0431)	(0. 0496)
Weather Controls	Y	Y	Y	Y	Y	Y
City FE	Y	Y	Y	Y	Y	Y
Year FE	Y	Y	Y	Y	Y	Y
Observations	4161	3878	3601	4161	3878	3601
Adj. $R^2$	0.9566	0.9613	0.9594	0.9567	0.9507	0.9473

Notes: Standard errors are clustered at city level and reported in parentheses. ***p<0.01, **p<0.05, *p<0.1.

The lagged temperature term shows a decreasing trend for the tourism effect as time progresses. The effect of temperature on tourism revenue exhibits a similar pattern. The delayed effect of temperature on tourism may have been due to people's expectations. If high temperatures at a tourism destination lead to an unsatisfactory visitor experience, the probability that visitors will forego revisiting the destination is high. In addition, adverse word-of-mouth publicity may discourage potential visitors from selecting destinations for their travel.

Therefore, the delayed effect of seasonal temperature must be considered. The findings obtained for Specification (2) including the lagged terms for the last two years are shown in Fig. 6, Fig. 7. The results demonstrate that lagged spring and summer temperatures significantly negatively affect tourism. In contrast, the lagged fall in temperature has a positive impact on tourism in the current year. Autumn temperature, which is conducive to outdoor activities, has a positive effect on prospective travellers when selecting a destination city. However, the estimated coefficients of the average seasonal temperature in the preceding years do not demonstrate a discernible temporal trend.

Fig. 6.

Estimated seasonal temperature effects on domestic tourism arrivals.

Fig. 7.

Estimated seasonal temperature effects on domestic tourism revenue.

To account for intertemporal dynamic impacts, we included lagged temperature bins in $t-1$ and $t-2$ year. However, due to space constraints, only two temperature bins are shown in Table 6. After exceeding 30 °C, the lagged temperature bin has a negative effect on tourism. This is consistent with our baseline regression results, in which the impact of extreme temperatures on tourism was severe. An additional day with temperatures falling into the 35 °C range is associated with a loss of 5.2%–6.6 % of tourism arrivals. In contrast, the revenue loss ranges from 5.0% to 5.2 %.

Table 6.

Estimated delayed and cumulative effects of temperature.

	Domestic tourism arrivals			Domestic tourism revenue
	Model with no lag	Model with one lag	Model with two lags	Model with no lag	Model with one lag	Model with two lags
	(1)	(2)	(3)	(4)	(5)	(6)
30–35 °C	−0.0073a	−0.0074a	−0.0079a	−0.0064a	−0.0068a	−0.0065a
	(0.0014)	(0.0014)	(0.0014)	(0.0014)	(0.0012)	(0.0012)
>35 °C	−0.0054a	−0.0052a	−0.0066a	−0.0050a	−0.0050a	−0.0052a
	(0.0018)	(0.0017)	(0.0018)	(0.0018)	(0.0014)	(0.0014)
L1: 30–35 °C		−0.0037a	−0.0037a		−0.0039b	−0.0033b
		(0.0009)	(0.0009)		(0.0015)	(0.0016)
L1: >35 °C		−0.0020c	−0.0010		−0.0020	−0.0005
		(0.0011)	(0.0011)		(0.0019)	(0.0019)
L2: 30–35 °C			−0.0026a			−0.0039a
			(0.0009)			(0.0012)
L2: >35 °C			−0.0004			−0.0009
			(0.0012)			(0.0018)
Sum of all 30–35 °C coefficients	−0.0073a	−0.0111a	−0.0142a	−0.0064a	−0.0107a	−0.0137a
	(0.0014)	(0.0021)	(0.0241)	(0.0014)	(0.0023)	(0.0028)
Sum of all >35 °C coefficients	−0.0054a	−0.0072a	−0.0080b	−0.0050a	−0.0070b	−0.0066c
	(0.0018)	(0.0025)	(0.0024)	(0.0018)	(0.0027)	(0.0039)
Weather Controls	Y	Y	Y	Y	Y	Y
City FE	Y	Y	Y	Y	Y	Y
Year FE	Y	Y	Y	Y	Y	Y
Observations	4161	3878	3601	4161	3878	3601
Adj. $R^2$	0.9578	0.9591	0.9608	0.9455	0.9474	0.9484

Notes: Standard errors are clustered at city level and reported in parentheses.

^ap<0.01.

^bp<0.05.

^cp<0.1.

6.2. Cumulative temperature effects

In this section, an analysis of the cumulative effect of temperature shocks on tourism is presented, which considers the delayed effects of extreme temperatures. Specifically, we calculated the cumulative effect by summing the coefficients estimated for the current year and those of the two previous years.

The linear model presented in Table 5 exhibits a persistent negative cumulative effect of temperature. An increasing magnitude is observed when additional lags are included. The results show that the current effect of temperature reduces tourism arrivals by 8.09 %. However, including two lagged terms, the cumulative effect of temperature reduces tourism arrivals by 17.81 %. These results underscore the significant underestimation of the effect of temperature on tourism when the delayed effect of temperature is not considered.

Based on the use of two lags, rising temperatures in spring, summer, and winter are negatively correlated with domestic tourism arrivals. However, the average temperature in autumn increases tourism arrivals by 6.50 %. The average winter temperature has a small negative effect on tourism numbers, with a cumulative reduction in tourism arrivals of 3.02 % over three years. The results for domestic tourism revenue and arrivals are similar. The temperature bins method also supports linear results.

7. Heterogeneity analysis

Tourism is a multifaceted socioeconomic phenomenon influenced by, but not limited to, various factors such as natural factors, the economy, infrastructure, supporting policies, and social culture. Therefore, we performed a series of heterogeneity tests to examine the response of city tourism to temperature increases under different influencing factors.

7.1. Economic development level

The level of economic advancement and improvement of the material base provides a basis for the development of tourism. Furthermore, it is a fundamental driving force of tourism development. Economic development can improve a city's local infrastructure and attract more tourists. A city's economic growth has the potential to elicit augmented investments from local businesses and external investors. Such investments provide financial support for the continued development of the city's tourism industry. The extent of financial investment in cities has played a significant role in the practical implementation of various tourism development policies proposed in recent years.

Table 7 illustrates the effects of temperature on tourism across various levels of economic development. We used the GRP as a proxy variable for the level of economic development in the city. To classify cities into low and high economic development categories, we first calculated the mean GRP for 2005–2019. Cities with a GRP below the mean were considered to have low economic development, whereas others were considered to have high economic development. Subsequently, we performed a subsample regression based on this classification. Columns (1a) and (2a) use domestic tourism arrivals as dependent variables. Columns (1b) and (2b) show domestic tourism revenue as the dependent variable.

Table 7.

Estimated effects of temperature on tourism arrivals and revenue in cities with different GDP levels.

	High economic level		Low economic level
	(1a)	(1b)	(2a)	(2b)
30–35 °C	−0.0054b	−0.0039c	−0.0058a	−0.0051a
	(0.0023)	(0.0023)	(0.0017)	(0.0018)
>35 °C	−0.0046c	−0.0032	−0.0041b	−0.0041c
	(0.0027)	(0.0023)	(0.0020)	(0.0023)
Weather Controls	Y	Y	Y	Y
City FE	Y	Y	Y	Y
Year FE	Y	Y	Y	Y
Observations	1110	1110	3051	3051
Adj. $R^2$	0.9667	0.9687	0.9505	0.9267

Notes: Standard errors are clustered at the city level and reported in parentheses.

^ap < 0.01.

^bp < 0.05.

^cp < 0.1.

Our study demonstrates that cities with low economic levels suffer more statistically significant declines in domestic tourism and revenue under the influence of high temperatures. Conversely, cities with higher economic levels were less susceptible to the negative effects of high temperatures. Our findings indicate that cities with a robust economic foundation are better equipped to deal with the detrimental effects of high temperatures on tourism. In other words, a strong economic base in a destination city enhances the city's ability to respond to the challenges posed by high temperatures and promotes the growth of sustainable tourism.

7.2. Population

With the acceleration of urbanization, the population density of cities is increasing. A gradual increase in the population density is conducive to the development of urban public services. However, coordination is necessary between tourism construction in a city and the population density of the destination city. Cities with different population densities may experience this effect differently because the availability of resources and infrastructure vary widely.

We conducted a subsample test to examine the effect of temperature on tourism in cities with different population densities. The results are shown in Table 8, which classifies the destination cities into two groups based on population density, similar to the division of cities with high and low economic development. Columns (1a) and (1b) show the effect of temperature on tourism in cities with high population densities, whereas columns (2a) and (2b) show the effect of temperature on tourism in cities with low population densities.

Table 8.

Estimating the effect of temperature on tourism arrivals and revenues in cities with different population densities.

	High population density		Low population density
	(1a)	(1b)	(2a)	(2b)
30–35 °C	−0.0026	−0.0047*	−0.0092a	−0.0062a
	(0.0026)	(0.0027)	(0.0017)	(0.0018)
>35 °C	−0.0008	−0.0038	−0.0075a	−0.0044b
	(0.0031)	(0.0030)	(0.0021)	(0.0022)
Weather Controls	Y	Y	Y	Y
City FE	Y	Y	Y	Y
Year FE	Y	Y	Y	Y
Observations	1536	1536	2625	2625
Adj. $R^2$	0.9644	0.9404	0.9523	0.9485

Notes: Standard errors are clustered at the city level and reported in parentheses.

^ap < 0.01, **p < 0.05.

^bp < 0.1.

In cities with high population densities, such as large metropolitan areas, the effect of temperature may be less noticeable because these cities typically have well-developed tourism industries and a wider range of attractions and activities that are independent of weather conditions. In contrast, tourism in destination cities with low population densities is severely affected by high temperatures. Cities with lower population densities may be more dependent on weather for tourism. Inclement weather or extremely hot temperatures can discourage tourists from visiting an area, leading to a decline in the local tourism industry. Therefore, the effect of temperature on tourism significantly varies among cities with different population densities. Although large cities may have more resilient and diverse tourism industries, smaller cities may be more vulnerable to changes in weather conditions.

7.3. Highspeed railway

The enhancement of transportation infrastructure is widely perceived as a crucial strategy for fostering tourism growth [46]. The high-speed railway (HSR), the most ubiquitous mode of transportation in contemporary society, appeals to travellers for comfort, speed, and reliability. HSR implementation facilitates the growth of the tourism industry by improving accessibility and mobility [47].

To examine the effect of the HSR on tourism destination cities coping with temperature shocks, we employed a dummy variable approach to categorize cities into those with and without HSR. The results of this analysis are presented in Table 9. Our results indicate that cities equipped with an HSR do not experience a significant decrease in tourism arrivals when faced with high temperatures. Conversely, cities lacking HSR systems experience a significant decline in tourism arrivals because of high temperatures. Tourism revenue shows the opposite results to tourism arrivals.

Table 9.

Estimate effects of temperature on tourism arrivals and revenue in cities with high-speed rail.

	With high-speed rail		Without high-speed rail
	(1a)	(1b)	(2a)	(2b)
30–35 °C	−0.0044b	−0.0042a	−0.0046a	−0.0026
	(0.0018)	(0.0015)	(0.0017)	(0.0019)
>35 °C	−0.0029	−0.0042a	−0.0048b	−0.0023
	(0.0021)	(0.0016)	(0.0021)	(0.0023)
Weather Controls	Y	Y	Y	Y
City FE	Y	Y	Y	Y
Year FE	Y	Y	Y	Y
Observations	1653	1653	2487	2487
Adj. $R^2$	0.9600	0.9507	0.9515	0.9347

Notes: Standard errors are clustered at the city level and reported in parentheses.

^ap < 0.01.

^bp < 0.05, *p < 0.1.

The integration of cities and destinations through the HSR shortens travel times, allowing tourists to visit multiple locations within a limited timeframe. This enhances the overall travel experience and attracts a greater number of tourists because remote areas become more accessible through the HSR. However, HSR provision alters tourist spending patterns. Tourists may allocate less time and money to a single destination if they can travel quickly and easily to multiple destinations [47]. The HSR may attract a different type of tourist who prioritizes convenience and cost efficiency over a leisurely, in-depth experience. An increase in day trippers and a reduction in overnight tourists have led to a decrease in tourism revenue.

7.4. 5A level tourist attractions as proxies for tourism resource endowment

The level and type of tourism resource endowment determines the tourism development potential of an area. These resources play critical roles in tourism development because they provide the basis for the tourism experience. Tourism resource endowment is a significant factor affecting tourist behaviour. A rich endowment increases the number of visitors to a destination and enhances its economic benefits [48]. In contrast, limited resource endowment restricts the types of tourists and tourism activities, which may reduce the economic benefits for the destination.

We used 5A level tourist attractions as a proxy variable for a city's tourism resource endowment. A city with 5A level tourist attractions is a rich tourism resource, whereas a lack of these attractions implies weaker tourism resources. The number of 5A level tourist attractions was obtained from the Ministry of Culture and Tourism website of the People's Republic of China. The results shown in Table 10 reveal that cities with limited tourism resources are more susceptible to the negative effects of high temperatures because tourists have fewer options. When the temperature rises, tourists may be reluctant to participate in outdoor activities or visit certain tourist attractions, resulting in lower tourism revenue for these cities. In contrast, cities rich in tourism resources are more resistant to the effects of high temperatures. Tourists visiting these cities have more options and are less likely to change their travel plans because of the high temperatures. Tourism resource endowment contributes to tourism development by providing attractions and facilities needed to attract tourists and bring economic benefits to the destination city.

Table 10.

Estimate effects of temperature on tourism arrivals and revenue in cities with 5A level tourist attractions.

	With 5A level tourist attractions		Without 5A level tourist attractions
	(1a)	(1b)	(2a)	(2b)
30–35 °C	−0.0061a	−0.0065a	−0.0076a	−0.0059b
	(0.0017)	(0.0016)	(0.0023)	(0.0023)
>35 °C	−0.0039c	−0.0043b	−0.0061b	−0.0057b
	(0.0022)	(0.0022)	(0.0025)	(0.0027)
Weather Controls	Y	Y	Y	Y
City FE	Y	Y	Y	Y
Year FE	Y	Y	Y	Y
Observations	2220	2220	1941	1941
Adj. $R^2$	0.9635	0.9644	0.9493	0.9136

Notes: Standard errors are clustered at the city level and reported in parentheses.

^ap < 0.01.

^bp < 0.05.

^cp < 0.1.

8. Conclusions

8.1. Conclusions and policy suggestion

Global warming has substantially affected the tourism sector. Climate change has resulted in many destinations facing increasingly frequent and severe weather extremes, with disruptions in tourism activity and declines in visitor arrivals. In the context of global warming, an accurate estimation of temperature-related tourism losses is an important prerequisite for an objective understanding of the impacts of climate change and the targeted formulation of policy responses. To achieve this objective, we employed city-level tourism data spanning 2005–2019 and paired them with detailed daily meteorological data. Multiple approaches were employed to precisely estimate the temperature–tourism correlation.

Firstly, the results of the linear regression model show that each 1 °C increase in the average temperature reduces domestic tourism arrivals by 8.09 % and revenue by 6.04 %. We also used the temperature bin method, which is not sensitive to the construction of the response equation, and found results that the tourism industry is susceptible to extreme high temperatures but not to cold temperatures. This conclusion is further supported by our seasonal analysis, which reveals that rising temperatures in spring and summer have a negative impact on the tourism demand, while rising temperatures in autumn and winter do not have a significant effect. These outcomes imply that seasonal discrepancies and extreme value variations in weather variables must be accounted for rather than solely relying on the average temperature when investigating the effects of temperature on tourism in a country such as China, which experiences significant temperature fluctuations throughout four seasons. We also considered the delayed effects and cumulative effects, and the results demonstrated that disregarding delayed effects would significantly underestimate the negative impact of extreme high temperatures.

We conducted a comprehensive heterogeneity test and made several noteworthy observations. The tourism industry of a city is affected by several important factors such as its economic level, population density, transportation infrastructure, and regional tourism resources. Our research indicates that cities with strong economic foundations can enhance their urban tourism facilities, enabling them to effectively cope with the adverse effects of high temperatures. Furthermore, cities with high population densities may experience less severe effects of temperature because of the presence of a well-established tourism industry. Improvements in the transportation infrastructure also play vital roles in enhancing the tourism industry by increasing accessibility and mobility, thus enabling destination cities to better withstand the effects of extreme heat. Another critical factor in promoting tourism is the abundance of tourism resources, which provides visitors with a wide range of options. Therefore, it is less likely that they will abandon their travel plans because of the heat.

Our findings provide an important reference for the development of tourism policies to cope with high temperatures, as well as new ideas and methods for future related research. A comprehensive understanding of the effect of climate change on tourism can help policymakers and industry leaders to develop effective mitigation and adaptation approaches. These strategies can help policymakers to develop long-term plans and maintain sustainable tourism. Governments should take pre-emptive actions to mitigate the effect of climate change on the tourism sector. Such actions may encompass investments in mitigation and adaptation measures such as the establishment of early warning systems for natural disasters, improvement of infrastructure in susceptible regions, and promotion of sustainable tourism development. Increased investment in destination infrastructure can improve the resilience of the tourism industry, which means improving the infrastructure conditions of transportation, accommodation, food, and beverage in tourism destinations. For example, the Government could increase the number of drinking water sources, build indoor resting places, and construct more shade facilities. The government can also optimize the public transportation system to improve accessibility and comfort, making it easier for tourists to reach their destinations. Meanwhile, given the significant negative impact of high temperatures on domestic tourism, the government can take various measures to strengthen the promotion of the international tourism market. Enhancing international tourism marketing efforts also contributes to the diversified development of the tourism industry, thereby increasing its status and contribution to the national economy.

8.2. Limitations and future study

The limitations of previous studies of the effects of rising temperatures on tourism are as follows.

Limited data are available on the effect of temperature on the tourism sector, particularly in developing countries. Therefore, it is difficult to accurately evaluate the effect of the temperature increase on this industry. Various methods, data sources, and measurement standards have been used in different studies, leading to inconsistent and incomparable results. The correlation between temperature and tourism is complex and depends on many factors including economic conditions, cultural and social factors, and infrastructure development. Separating the effect of temperature from that of other factors and determining their relative degrees of importance can be challenging. Furthermore, the limitations of previous studies in terms of the effects of rising temperatures on the tourism industry highlight the need for a more comprehensive and interdisciplinary approach to understand the complex relationships among temperature, tourism, and other related factors.

China is a country with diverse geographical features and rich tourism resources. A few cities located in tropical regions, such as Haikou and Sanya, experience high temperatures year-round. Such climatic conditions have become an important tourist resource for these cities, which may have taken several measures to adapt to the high temperatures. Therefore, the increase in temperature may not have a serious impact. Therefore, the impact of temperature on tourism may differ in cities with different tourism resources. However, due to data limitations, this paper fails to fully take into account the heterogeneity of tourism resources in different places when considering the impact of temperature on tourism. Nevertheless, most Chinese cities are located in the northern temperate zone. Therefore, our research findings are still of critical relevance. It can provide some insights into understanding the impact of temperature change on most Chinese tourist cities.

Tourist behaviour varies according to factors such as age, income, cultural background, and previous travel experiences. Firstly, tourists of different age groups may have varying sensitivities to temperature. In addition, the income level of tourists may also affect their response to weather. Finally, tourists' cultural backgrounds and previous travel experiences may also have an impact on their responses to weather. Tourists from different cultural backgrounds may have different adaptations to hot weather. Thus, while we have considered the effect of temperature on tourism demand, the lack of individual-level data makes it difficult for the study to gain insight into the differences between groups.

In future research, based on the methodology of our study, the understanding of the impact of temperature on tourism can be further deepened in three main areas. First, by targeting different types of tourist cities, such as those with natural landscapes or cultural heritage as their main tourism resources, researchers can study these cities' adaptability to temperature changes and their tourism industry's resilience adjustment mechanisms. Second, future research could analyse tourists' perceptions and preferences of temperature changes in relation to factors such as age, income, cultural background, and travel experience, and further explore the trade-off considerations of weather conditions in travel decisions by different groups. Finally, future research needs to accumulate long-term weather data to build more accurate models to comprehensively assess the long-term impacts of climate change on tourism.

Data availability statement

Data will be made available on request.

CRediT authorship contribution statement

Dingyi Chang: Writing – original draft, Visualization, Investigation, Formal analysis, Data curation. Naipeng Bu: Writing – review & editing, Validation, Project administration, Conceptualization. Ning Zhang: Software, Methodology, Funding acquisition. Honggen Xiao: Supervision, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.