Severely underestimated rate of wind speed decline jeopardizes China's carbon peak emission target


 The development of wind energy is indispensable in the pursuit of global carbon neutrality. Following decades of climate change, China's annual average wind speed has shown a clear decline, but the rate of this decline and its potential impacts on the need for wind power development in China have not been quantified. Here, we reveal that China's observed wind speed has declined significantly at -0.169 m/s/10 yr, 33.33 times the rate predicted by the Coupled Model Intercomparison Project (CMIP) of World Climate Research Programme, indicating a severe underestimation by those models. We attribute this underestimation to CMIP neglecting the atmospheric boundary layer height and greatly underestimating the Arctic amplification effect on wind speed. Scenario analyses demonstrate that China’s future wind power installed capacity, investment costs, and land area occupied based on the observed trend scenario will increase by approximately 53% compared to any trend of three (high, medium, and low) CMIP emission scenarios in order to meet China’s wind-generated electricity target and carbon peak emission goal in 2030. Hence, formulating its wind energy development plans based on these CMIP scenarios’ trends will prevent China from meeting its low-carbon electricity generation of carbon peak emission target by 2030 and delaying the 2060 goal of carbon neutrality, emphasizing that the CMIP models urgently need to be improved. These findings should serve as a warning to countries throughout the Northern Hemisphere to formulate wind power development plans in consideration of the climate change impacts in the pursuit of global carbon neutrality.


Introduction
The International Renewable Energy Agency (IRENA) predicted that 1 in order to achieve the 1.5°C temperature rise goal of the Paris Agreement, more than one-third of global electricity generation in 2050 will need to derive from wind; as a result, wind energy will become the world's largest source of electricity power. At present, China already has the largest installed capacity and the fastest growth rate of wind power worldwide; at the end of 2020, China's installed wind power capacity reached 37.8% of the global total 2 , and its capacity is set to grow further. According to the plan for achieving China's carbon peak emission target by 2030 3 , 25.9% of all electricity consumed in 2030 will be provided by non-hydro renewable energy 4 . Assuming that half of its electricity will be provided by wind power, China's wind power generation potential in 2030 will be more than twice that in 2020 5 . The development of wind power in China will be decisive for realizing carbon neutrality globally.
Nevertheless, the scale of wind power development that would be required to satisfy China's national strategy depends heavily on future wind speed levels. Yet , a decline of wind speed 6 due to global climate change in the north hemisphere might have a profound impact on wind power development and even global carbon neutrality. At present, Coupled Model Intercomparison Project (CMIP) 7 climate data are widely used to predict future climate trends [8][9][10] , including those of wind speed 6,[11][12][13] . However, past research on the uncertainty and accuracy of CMIP models has typically focused on temperature and precipitation, etc. [14][15][16] , rather than on wind speed. Hence, the reliability of CMIP wind speed predictions and the consequences of errors in those predictions for future wind power development remain unclear.
In this paper, we aim to quantify the rates of wind decline in different speed ranges by observed data in China, and then to identify if CMIP models could predict wind speed correctly in the past, finally to evaluate its potential impacts on the future need for wind power development in China. To this end, by collecting observed wind speed data from 771 ground stations throughout China during 1981-2014, we reveal a longterm decline in the average wind speed that CMIP remarkably underestimates. Our theoretical analysis indicates that this underestimation may be due to CMIP neglecting the atmospheric boundary layer height and greatly underestimating the Arctic amplification effect on wind speed. Finally, comparing our observed wind speed trend and the three CMIP scenarios of high, medium, and low emissions, we estimate the new installed capacity, new investment costs, and new land area occupation required to fulfil China's 2030 wind energy generation goal according to the national carbon emission reduction strategy. The scenario differences demonstrate the potential impacts of underestimating the wind speed rate of decline to China's wind power development goal and how this threat jeopardizes the progress made toward the national carbon peak emission goal. This research is of great significance for facilitating China's carbon emission reduction and carbon neutrality strategies and further provides a scientific basis for formulating wind power development plans under the meteorological influences of global climate change. Our findings should also help other countries at similar latitudes in the Northern Hemisphere formulate wind power development plans more reasonably.

The long-term trend of observed wind speeds in China
For this study, we collected long-term (1981-2014) observation data from surface meteorological stations throughout China. A total of 771 stations were checked to guarantee that they had never been moved and that the data were continuous.
Considering that the height of a typical modern-day wind turbine is 80 m, the original wind speeds at a height of 10 m (10 m wind speeds) were extrapolated to a height of 80 m (80 m wind speeds) using a power law relation. The annual average wind speed among all 771 stations is plotted in Figure 1. The stations are divided into three wind speed intervals (0-3, 3-6, and 6-9 m/s) according to the annual averages over the 34yr study period. All wind speed averages, and more than 75% of all stations show a declining trend (i.e., have a negative linear fitting coefficient; Table 1). In addition, the rates of decline appear to increase, and are observed at more stations, with increasing wind speed interval. As shown in Table 1, the wind speed rate of decline among all stations during 1981-2014 was -0.169 m/s/10 yr. At wind speeds above 3 m/s and up to 6 m/s, the decline rate was -0.316 m/s/10 yr, and more than 93% of the stations yielded a rate of decline. Furthermore, at wind speeds between 6 and 9 m/s, the decline rate reached -0.488 m/s/10 yr.  Analysing the mechanism responsible for the declining wind speeds Next, we theoretically analysed the mechanism responsible for the declining wind speeds in China by using the classic geostrophic wind relationship in the free atmosphere, the Ekman spiral relationship of the atmospheric boundary layer, and the ideal gas state equation, etc. We derived the following long-term theoretical relationship of the 80 m wind speed 80 : where − 2 is the 2 m height latitudinal temperature ( 2 ) gradient of latitudes ( ), reflecting the Arctic amplification (the Arctic warming pace has been approximately double that of global average in recent decades) effect [17][18][19] , is the height of the local atmospheric boundary layer, is the ideal gas constant, and is the geostrophic parameter. Equation 1 illustrates that the latitudinal temperature gradient − 2 (the horizontal influencing factor) and the atmospheric boundary layer height (the vertical influencing factor) are the two key parameters affecting the wind speed trend.
However, this long-term declining wind speed trend is not restricted to China but is applicable to all other countries at similar latitudes in the Northern Hemisphere.  The theoretically derived declining trend is very close to that of the observation, which verifies the correctness of the theoretical wind speed equation.
To better understand the magnitudes of − 2 and , their long-term trends are plotted in Figure 3. It can be seen that − 2 presents a long-term declining trend caused by the Arctic amplification effect, which leads to a decrease in wind speed according to Equation 1. This north-south temperature difference is the driving force of atmospheric circulation, and thus, a decrease in circulation corresponds to a decline in wind speed.
In contrast, presents a long-term increasing trend, which also leads to a decrease in wind speed. The long-term change in is affected mainly by the local surface temperature, surface roughness, etc. In particular, surface warming raises .
As increases, the upper free atmosphere (associated with high wind speeds with strong momentum) above the atmospheric boundary layer moves away from the surface, increasing the distance over which momentum is transmitted (between high altitudes and the surface) and thus decreasing the surface wind speed. although the influence of is slightly greater.

Evaluating the CMIP-predicted wind speeds in China
Currently, wind speed trends are predicted mainly based on CMIP climate models.
For example, by using CMIP data, researchers have evaluated the future trends of wind speed and wind energy for the Northern Hemisphere 6 , including Europe 11 , China [12][13] and India 13 .
To evaluate the CMIP wind speed predictions, we selected the latest release, CMIP Phase 6 (CMIP6) 21 for an analysis of three future scenarios from among the CMIP shared socioeconomic pathways (SSPs) 22 : the low-emission SSP126 scenario, the medium-emission SSP245 scenario, and the high-emission SSP585 scenario.    Table 2. USD, and the new area of land occupation is anticipated to span more than 28,000 km 2 .
The above results show that the CMIP models' underestimation of the wind speed rate of decline will jeopardize China's wind power development strategy that is being implemented to achieve the carbon peak emission target by 2030; accordingly, this discrepancy will certainly delay the goal of reaching national carbon neutrality by 2060.

Conclusion
In this study, we analysed wind speed observation data during 1981-2014 from 771 meteorological stations and simulation data from numerous CMIP models, and revealed that these models severely underestimate the wind speed rate of decline in China. The theoretical derivation of an equation for computing the 80 m wind speed suggests two main reasons for the declining wind speed trend, that is, a decreasing latitudinal temperature gradient and an increasing atmospheric boundary layer height.
Specifically, the CMIP models neglect to account for the atmospheric boundary layer height and greatly underestimate the Arctic amplification effect on wind speed. Overall, these insights should be of great significance for achieving global carbon neutrality.

Calculation of the annual average wind speed at a height of 80 m
The wind speed at each observation station is recorded at a height of 10 m, but the hub height of a wind turbine is generally 80 m. Therefore, the wind speed at a height of 10 m is extrapolated to a height of 80 m using a power law relation (a power law coefficient of 1/7 is chosen, but the power law coefficient has been shown to only slightly impact the wind energy potential 23 ): The observations were acquired every 6 hours, but after taking the annual average of the data series, the data adopt a yearly resolution.
The original CMIP6 wind speed data are also represented at a 10 m height and are also extrapolated to 80 m using the power law wind profile above (Equation 2). The original temporal resolution is monthly, but after annual averaging, the resolution also becomes yearly. The CMIP6 wind speed data are spatially gridded. Thus, to obtain CMIP6 data spatially coinciding with the 771 observation stations, we perform bilinear interpolation for the CMIP data on the grids adjacent to each observation station. Finally, the CMIP data corresponding to the 771 stations are averaged to compute the linear declining trend for each CMIP model and each SSP scenario (see Supplementary Tables   2-4).

Theoretical analysis of the long-term trend of wind speed at a height of 80 m
For the high-altitude free atmosphere (that is, the atmosphere above the atmospheric boundary layer), the geostrophic Coriolis force is at equilibrium with the pressure gradient force, yielding the following geostrophic wind relationships 24-25 : where , , , , and are the zonal geostrophic wind, meridional geostrophic wind, air pressure, geostrophic parameter, and air density, respectively.
For the atmosphere under the atmospheric boundary layer, the geostrophic Coriolis force, turbulent friction force, and pressure gradient force are balanced, and the relationship is as follows 24-25 : where ��� and � are the average zonal wind and average meridional wind at a height (under the atmospheric boundary layer, such as at the 80 m height of a wind turbine hub), respectively, and is the turbulent friction coefficient.
Based on the above relationship between the free atmosphere and boundary layer, the ideal gas state equation, the vertical declining temperature gradient, the boundary layer height , and the reasonable simplification of secondary items (using a Boussinesq approximation 26 , etc.), the theoretical relationship of the long-term change in the wind speed at a height of 80 m 80 is derived as follows: The detailed derivation of Equation 7 and the method for calculating the wind speed based on ERA5 data are shown in the Supplementary Information.

Method for estimating the average wind speed in 2014 employed to generate wind power in China
Here, the Rayleigh distribution, which is commonly used for the initial estimation of wind energy resources, is employed as the universal statistical wind speed distribution 27 . In the Rayleigh distribution, the distribution function ( ， ) is related only to the average wind speed and the wind speed , namely, A typical Rayleigh distribution under equals 5 / is shown in Figure 6 a.
For the power curve, we refer to the 2 MW Goldwind GW-115 wind turbine, which is commonly used in Chinese wind farms. Its theoretical power curve ( ) is plotted in Figure 6 b, and its rated power is recorded as 0 . Determining the annual power generated by a wind turbine is equivalent to integrating ( ) with all wind speed distribution intervals over all times of the year.
Consequently, the power generated by a wind turbine 0 is