The SDSM was employed for downscaling of daily mean temperature of 32 meteorological stations (1954–2014) and future scenarios were generated up to 2100. The data were daily NCEP/NCAR reanalysis data and the daily mean climate model outputs for the RCP2.6, RCP4.5, and RCP8.5 scenarios from the MRI of Japan. Periodic features were obtained by wavelet analysis. The results showed the following. (
The Earth’s temperature increased by 0.85
In order to assess the potential future impact of climate, research has been directed toward the development of GCMs with spatial resolutions of hundreds of kilometers, and important global or continental-scale atmospheric processes, as well as future climate prediction, under different emission scenarios can be reproduced by state-of-the-art GCMs. However, hydrologists are interested in scales of a few kilometers, and, furthermore, hydroclimatic information at finer scales is essential for climate impact studies [
Based on the way of using large-scale GCM information on local scale, downscaling methods can be broadly grouped into two categories: dynamical downscaling (DD) and statistical (or empirical) downscaling (SD) [
According to the techniques involved in its application, the SD approach can be divided into three categories, namely, weather generator, weather typing, and regression/transform function [
Simultaneously, Chinese researchers have focused increased attention on the SDSM. Most previous work was concentrated on maximum and minimum air temperatures [
Good knowledge of future scenarios of daily mean temperature in Yunnan is critical for grain yield, for planning and management of water resources, and for design of systems for monitoring climate change and will provide valuable information in policy making for local government. Therefore, this study aims to offer a detailed projection of daily mean temperature for Yunnan in the twenty-first century using the SDSM. It strives to (
Yunnan province is located in southwestern China and covers a total area of about 39.4 × 104 km2 between 21.9–29.25°N and 97.32–106.12°E [
Basin profile and distribution of meteorological stations in Yunnan province.
Three kinds of data were used in this study: observed daily mean air temperature, NCEP reanalysis data, and GCM output data. Observed daily mean air temperatures from 32 meteorological stations (Figure
The terrestrial carbon cycle process is an essential component for the accurate estimation of climate change. In CGCM3, the terrestrial biosphere absorbs CO2 from the atmosphere by producing vegetation and releases CO2 into the atmosphere by decomposition of soil organic carbon. The terrestrial carbon cycle model is based on models of the biochemical processes of photosynthesis on the organism leaf level and on a dynamic global vegetation model on the ecosystem-biogeochemical level. What is more, the model contains the interaction mechanism of C3 and C4 photosynthesis plants. The leaf-level photosynthesis model is calculated with a time interval of 30 minutes to 1 hour. On the ecosystem level, the exchange of CO2 between the atmosphere and the ecosystem is evaluated by the difference between net primary production and soil respiration with a time interval of 1 day to 1 month (see [
The SDSM, developed by Wilby et al. (2002) [
In general, five steps are involved when using the SDSM [
The SDSM has been applied to produce high-resolution climate change scenarios in a range of geographical contexts [
Wavelet analysis has many advantages over the fast Fourier transform (FFT) [
The most commonly used wavelet functions are the Morlet wavelet [
The wavelet coefficient
The variance of the wavelet coefficient, the wavelet variance
The study carried out the downscaling process for each station, which produced 20 ensembles (default), and the study then took the mean of these ensembles. The model was calibrated and validated using observations of 40 years of data (1954–1993, termed the base period) and the remaining 21 years of data (1994–2014, termed the late period), respectively. During the calibration process, a monthly submodel was developed and the downscaling process was selected as an unconditional process for daily mean temperature. The predictand was daily mean temperature from the observed surface variables, and the predictor variables were from the NCEP reanalysis data, including near surface specific humidity, mean sea level pressure, near surface relative humidity, geostrophic airflow velocity, mean temperature, zonal velocity component, and meridional velocity component.
The mean air temperature (
Figure
Modeled and observed mean values of air temperature during calibration and validation periods.
Calibration
Validation
Scatter plots of daily mean temperature between modeled and observed series during calibration and validation periods.
Calibration
Validation
On the whole,
Simulated versus observed daily mean temperatures in both calibration and validation periods are shown in Figure
Mean monthly changes in projected temperature with respect to the base period under three scenarios.
RCP2.6
RCP4.5
RCP8.5
On the contrary, substantial decreases in temperature are predicted in March and April under both the RCP2.6 and RCP4.5 scenarios for future periods, with the greatest decrease (−0.79°C) being seen in March under RCP8.5 in the 2020s.
Changes in mean seasonal temperature (compared with the base period 1954–1993) in Yunnan province under the three scenarios are shown in Figure
Mean seasonal changes in projected temperature with respect to the base period under three scenarios.
Areal average of future daily temperature (°C) compared with the base period.
Observation | Projected air temperature | ||||||||
---|---|---|---|---|---|---|---|---|---|
RCP2.6 | RCP4.5 | RCP8.5 | |||||||
Base period | 2020s | 2050s | 2080s | 2020s | 2050s | 2080s | 2020s | 2050s | 2080s |
16.15 | 16.31 | 16.55 | 16.68 | 16.43 | 16.78 | 17.05 | 16.41 | 17.15 | 17.99 |
Increase | 0.16 | 0.40 | 0.53 | | 0.63 | 0.89 | 0.26 | | |
Base period = 1954–1993; 2020s = 2011–2040; 2050s = 2041–2070; 2080s = 2071–2100.
More details of changes in air temperature can be seen in Figure
Annual mean temperature series in Yunnan province under three scenarios.
Results of wavelet analysis under the RCP4.5 scenario.
The real part of the wavelet coefficient contour map
Wavelet variance diagram
Figure
Spatial distribution of the increase in annual mean temperature in Yunnan province under three scenarios.
RCP2.6_2020s
RCP2.6_2050s
RCP2.6_2080s
RCP4.5_2020s
RCP4.5_2050s
RCP4.5_2080s
RCP8.5_2020s
RCP8.5_2050s
RCP8.5_2080s
The air temperature under different scenarios shows slightly diverse dominant periods (Table
Main periods (
Main period | RCP2.6 | RCP4.5 | RCP8.5 |
---|---|---|---|
| 3 | 4 | 4 |
| 6 | 6 | 6 |
| 10 | 12 | 13 |
| 22 | 21 | — |
| 39 | — | 35 |
| 47 | 50 | 53 |
| 79 | 83 | 83 |
| 125 | 125 | 125 |
The spatial distributions of future changes in annual mean temperature in Yunnan province (compared with the base period) under the RCP2.6, RCP4.5, and RCP8.5 scenarios were obtained by a radial basis function (RBF) interpolation technique using the ArcGIS10.1 software package. The results for the annual changes are shown in Figure
On the whole, most of the province may experience an obviously increasing trend in the three future periods under all three scenarios. However, a few areas, including some patches from the center of Yunnan, such as Kunming and Yuxi, are expected to show a negative trend. With the passage of the decades (2020s, 2050s, and 2080s), the increase will be greater, and the area of higher temperature will grow in size.
In the 2020s (Figures
For the 2050s (Figures
In the 2080s (Figures
Under the different scenarios, in general, the increase in air temperature under RCP8.5 will be highest, followed by that under RCP4.5 and by that under RCP2.6.
Statistical downscaling methods, as effective measures, are generally used to construct the bridge between large-scale atmospheric variables from GCM output and local-scale climate response; among these methods, the SDSM is widely used owing to its simplicity and superior capabilities, as well as the free availability of the required software [
The SDSM generally performs better in reproducing temperature than rainfall and evaporation [
In further studies, the results obtained here need to be subjected to a comprehensive comparison with the outputs of other GCM models to achieve greater reliability in terms of future projections. In particular, analyses of the uncertainties related to the model and the GCM data are needed for a more profound understanding of the future changes in predictands.
This paper has presented future projections of air temperature in the Yunnan Plateau over the twenty-first century under the RCP2.6, RCP4.5, and RCP8.5 emission scenarios from the MRI using the SDSM. The study has investigated the applicability of the SDSM by downscaling the mean air temperature, which is important for assessing the impact of climate change on evapotranspiration and the hydrologic cycle. Changes in daily mean temperature have been addressed to provide better understanding of future changes in climate extremes in this region. Furthermore, the study has analyzed the patterns of change for the 2020s, 2050s, and 2080s, which should help government make policy on extreme climate events and agriculture and also pave the way for the study of hydrological impacts under future climate change in the middle and lower reaches of the Lancang River basin. The major conclusions can be summarized as follows: The SDSM showed good applicability to the simulation of air temperature in Yunnan both for individual stations and for the whole region. During the calibration period, the maximum relative errors (MREs) were 0.65% and 0.77% for mean and median air temperature over 12 months, and 0.28% and 0.37% for mean and median air temperature over all seasons. For the validation period, the MREs were −6.16% and −6.23% for mean and median air temperature over 12 months and −6.09% and −5.89% for mean and median air temperature over all seasons. The monthly submodel of the SDSM was found to be effective for downscaling of air temperature, with high correlations ( The projected results of future air temperature showed that the monthly mean temperature of three future periods would all increase with different magnitudes under the three different scenarios compared with the base period, except for March and April. The increase under RCP8.5 was the highest for all months. On average, June showed the most obvious increase among the 12 months. Seasonally, air temperature would decline slightly in spring in the 2020s and clearly increase in the other seasons. Moreover, the greatest change would occur in the summer and the least change in spring. Annually, on average, the most remarkable changes in the 2020s, 2050s, and 2080s were 0.27, 1.00, and 1.84°C, respectively. The annual air temperature from 1954 to 2100 in Yunnan under the different scenarios showed similar periodicity, with seven periods, but under the RCP2.6 scenario, there were eight periods. Typical 22- and 39-year periods appeared under RCP2.6 and a typical 21-year period under RCP4.5, while an alternative period of 35 years appeared under RCP8.5. On the whole, the annual mean temperature in most parts of Yunnan would be dominated by increasing trends during all of the future periods under all three scenarios, with the magnitude and percentage of increase being greatest in the 2080s under RCP8.5. Furthermore, there would be a more distinct increase in northwest and southeast Yunnan in most future periods, whereas there would be a significant decrease in the center of Yunnan (e.g., Yuxi and Kunming). This may be partly because of the abnormal conditions caused by special terrain and, on the other hand, a result of the influence of the distribution of nine plateau lakes. Five of these lakes (accounting for approximately 59.87% of the total area of the nine), namely, Dianchi, Fuxian, Qilu, Xingyun, and Yangzong, are located in Yuxi and Kunming, and their surface water will absorb heat from the air by evaporation, thereby limiting or retarding rises in temperature.
The authors declare that they have no competing interests.
This research was supported by the National Natural Science Foundation of China (no. 91547114) and the National Water Pollution Control and Treatment Science and Technology Major Project (no. 2013ZX07102-006-04). The authors would like to express their gratitude to the following projects and data centers: the National Meteorological Information Center in Beijing for providing valuable station climate datasets and the Intergovernmental Panel on Climate Change for providing the MRI/CGCM3 model output data. Furthermore, the authors would like to thank the following individuals: Professor R. L. Wilby from Loughborough University, UK, for providing SDSM software and NCEP reanalysis datasets, and Sun Zhen and Yin Yuan-yuan from the Institute of Geographic Sciences and Natural Resources Research and Liang Li-qiao from the Institute of Tibetan Plateau Research, CAS, for providing comments.