Changes in Vegetation and Assessment of Meteorological Conditions in Ecologically Fragile Karst Areas

+ Author Affiliations + Find other works by these authors
  • Corresponding author: Weihua MO, mwh-0419@163.com
  • Funds:

    Supported by the Guangxi Zhuang Autonomous Region (GZAR) Science and Technology Project (AB20159022 and AB17292051) and GZAR Natural Science Foundation (2018GXNSFAA281338)

  • doi: 10.1007/s13351-021-9170-x

PDF

  • Meteorological conditions have an important impact on changes of vegetation in ecologically fragile karst areas. This study aims to explore a method for quantitative evaluation of these meteorological conditions. We analyzed the changing trend of vegetation during 2000–2018 and the correlations between vegetation changes and various meteorological factors in karst rocky areas of Guangxi Zhuang Autonomous Region, China. Key meteorological factors in vegetation areas with varying degrees of improvement were selected and evaluated at seasonal timescale. A quantitative evaluation model of comprehensive influences of meteorological factors on vegetation was built by using the partial least-square regression (PLS). About 91.45% of the vegetation tended to be improved, while only the rest 8.55% showed a trend of degradation from 2000 to 2018. Areas with evident vegetation improvement were mainly distributed in the middle and northeast, and those with obvious vegetation degradation were scattered. Meteorological factors affecting vegetation were significantly different among the four seasons. Overall, high air humidity, small temperature difference in spring and autumn, and low daily minimum temperature and air pressure were favorable conditions. Low temperature in winter as well as high temperature in summer and autumn were unfavorable conditions. The Climate Vegetation Index (CVI) model was established by PLS using the maximum, minimum, and average temperatures; vapor pressure; rainfall; and air pressure as key meteorological factors. The Enhanced Vegetation Index (EVI) was well fitted by the CVI model, with the average coefficient of determination (r2) and root mean square error (RMSE) of 0.856 and 0.042, respectively. Finally, an assessment model of comprehensive meteorologi-cal conditions was built based on the interannual differences in CVI. The meteorological conditions in the study area in 2014 were successfully evaluated by combining the model and selected seasonal key meteorological factors.
  • 加载中
  • Fig. 1.  Distribution of rocky desertification areas and meteorological stations in the Guangxi Zhuang Autonomous Region (GZAR).

    Fig. 2.  The vegetation change trend in karst rocky areas of GZAR from 2000 to 2018.

    Fig. 3.  Scatter plots of EVI and CVI during 2000–2018.

    Fig. 4.  The evaluation results of meteorological conditions in karst rocky areas of GZAR in 2014.

    Table 1.  The SLOPE grading standard in rocky desertification areas of Guangxi Zhuang Autonomous Region (GZAR)

    GradeChange of vegetationRange of SLOPEProportion (%)
    1Noticeably worse [−1.0, −0.2) 0.04
    2Noticeably worse [−0.2, −0.15) 0.08
    3Noticeably worse [−0.15, −0.1) 0.22
    4Worse [−0.1, −0.05) 0.84
    5Slightly worse [−0.05, 0) 7.37
    6Slightly better [0, 0.05)58.37
    7Better [0.05, 0.1)30.35
    8Noticeably better [0.1, 0.15) 2.55
    9Noticeably better [0.15, 0.2) 0.18
    10Noticeably better [0.2, 1.0) 0.01
    Download: Download as CSV

    Table 2.  The correlation between vegetation EVI and meteorological factors in karst rocky areas of GZAR

    PeriodVapor pressureRainfallRelative humidityAverage temperatureMaximum temperatureMinimum temperatureSunshine durationAir pressureDew point
    Spring+/+/////+
    Summer///////
    Autumn/+/////
    Winter+//+++++
    Note: + indicates the significant positive correlation, − indicates the significant negative correlation, and / indicates that the correlation is not significant.
    Download: Download as CSV

    Table 3.  Values of climatic factors in the five categories of SLOPE

    PeriodCategory (%)Vapor pressure (hPa)Rainfall (mm)Relative humidity
    (%)
    Average temperature (°C)Maximum temperature (°C)Minimum temperature (°C)Sunshine duration
    (h)
    Air
    pressure (hPa)
    Dew
    point
    (°C)
    Spring< 5016.5170.767.421.328.416.4443.8928.614.0
    50–6017.8218.674.920.225.816.3352.6939.715.0
    60–7019.0297.175.521.126.117.6293.5976.316.1
    70–8019.9273.476.521.626.218.4292.9972.716.8
    > 8020.2297.077.221.625.818.5259.5980.917.0
    Summer< 5025.2625.879.325.430.922.0500.0922.721.2
    50–6026.3770.783.625.130.121.8477.8935.321.8
    60–7028.1861.181.126.831.923.7483.6970.222.9
    70–8028.1797.580.726.831.623.7516.7966.222.9
    > 8028.9838.781.127.231.824.2494.9974.623.4
    Autumn< 5018.6173.978.420.326.116.7410.0933.115.9
    50–6018.8199.480.519.825.316.2431.0944.015.9
    60–7019.6204.176.621.327.117.9439.7980.116.5
    70–8020.2211.575.921.927.418.3482.3975.316.9
    > 8020.3222.273.822.527.619.0490.9983.717.0
    Winter< 5010.659.375.612.418.58.5307.2934.97.4
    50–6010.770.477.211.716.88.3275.4946.27.3
    60–7011.1105.475.212.417.29.4243.5984.07.7
    70–8012.099.074.713.718.210.6274.7980.08.7
    > 8011.8114.674.113.517.910.6256.2988.88.4
    Download: Download as CSV

    Table 4.  Characteristics of meteorological elements in areas where vegetation was obviously improved

    PeriodVapor pressureRainfallRelative humidityAverage temperatureMaximum temperatureMinimum temperatureSunshine durationAir pressureDew point
    SpringHHHH/HLHH
    SummerHH/H/H/HH
    AutumnHHLHHHHHH
    WinterHH /L//HLH //
    Note: H indicates that the high-value feature is obvious, L indicates that the low-value feature is obvious, and / indicates that the feature is not obvious.
    Download: Download as CSV

    Table 5.  Selected key meteorological factors that have a strong correlation with vegetation change in the study area

    PeriodVapor pressureRainfallAverage temperatureMaximum temperatureMinimum temperatureAir pressure
    Spring
    Summer
    Autumn
    Winter
    Note: ↑ indicates that a high value is favorable; ↓ indicates that a low value is favorable.
    Download: Download as CSV

    Table 6.  The variation percentage explained by the partial least-square (PLS) factors

    Latent factorX varianceCumulative X varianceY varianceCumulative Y variance
    10.8300.8300.8070.807
    20.0840.9150.0250.832
    30.0570.9720.0180.850
    40.0230.9950.0090.859
    50.0030.9980.0120.871
    60.0010.9990.0070.878
    Download: Download as CSV

    Table 7.  The analysis of variable projection importance indices

    IndexVTTmaxTminPR
    VIP1.0871.0661.0651.0530.9880.842
    Download: Download as CSV

    Table 8.  Error and variance analyses of the partial-least square regression model

    ModelSum of squaresdfr2Adjusted r2Std. error of the estimateFSig.
    12.33560.8560.8520.042220.3250.000
    0.394431
    2.728439
    Note: degree of freedom—df, coefficient of determination—r2, standard deviation—Std., Fisher test—F, significance—Sig.
    Download: Download as CSV

    Table 9.  The classification standard of meteorological conditions in karst rocky areas of GZAR

    GradeGoodGeneralBad
    Classification standard[0.02, 1][0, 0.02)[−1, 0)
    Download: Download as CSV

    Table 10.  Meteorological elements in rocky desertification areas of GZAR in 2013 and 2014

    Meteorological factor20132014Multi-year
    average
    Vapor pressure (spring; hPa)20.020.719.4
    Minimum temperature (summer; °C)23.824.023.6
    Air pressure (summer; hPa)965.1964.8964.8
    Rainfall (autumn; mm)239.9347.5210.6
    Average temperature (autumn; °C)20.922.121.6
    Minimum temperature (autumn; °C)17.819.118.1
    Average temperature (winter; °C)12.711.513.0
    Maximum temperature (winter; °C)16.416.717.7
    Minimum temperature (winter; °C)10.38.19.9
    Vapour pressure (winter; hPa)11.910.011.5
    Download: Download as CSV
  • [1]

    Ban, M. L., H. Xie, Z. Xie, et al., 2018: Analysis on the current situation and countermeasures of biodiversity conservation in Nanling region of Guangxi. South China Agric., 12, 65–68. doi: 10.19415/j.cnki.1673-890x.2018.16.020. (in Chinese)
    [2]

    Chen, Y. L., B. J. Long, X. B. Pan, et al., 2010: Grassland vegetation change based on MODIS NDVI data and climate information. J. Appl. Meteor. Sci., 21, 229–236. doi: 10.3969/j.issn.1001-7313.2010.02.013. (in Chinese)
    [3]

    Chen, Y. L., B. J. Long, X. B. Pan, et al., 2011: Differences between MODIS NDVI and AVHRR NDVI in monitoring grasslands change. J. Remote Sens., 15, 831–845. (in Chinese)
    [4]

    Chen, Y. L., Y. M. Luo, W. H. Mo, et al., 2014a: Differences between MODIS NDVI and MODIS EVI in response to climatic factors. J. Nat. Resour., 29, 1802–1812. doi: 10.11849/zrzyxb.2014.10.015. (in Chinese)
    [5]

    Chen, Y. L., W. H. Mo, J. F. Mo, et al., 2014b: Comparison between MODIS-NDVI and MODIS-EVI in kasrt rocky desertification area. Remote Sens. Technol. Appl., 29, 943–948. (in Chinese)
    [6]

    Chen, Y. L., W. H. Mo, Y. M. Luo, et al., 2015a: EVI simulation of vegetation in karst rocky area using climatic factors. Trans. Chinese Soc. Agric. Eng., 31, 187–194. doi: 10.11975/j.issn.1002-6819.2015.09.029. (in Chinese)
    [7]

    Chen, Y. L., M. T. Huang, and W. H. Mo, 2015b: Vegetation variation in typical ecological sections of Southwest Guangxi based on MODIS NDVI data. J. Meteor. Sci., 35, 93–99. doi: 10.3969/2014jms.0017. (in Chinese)
    [8]

    Chen, Y. L., J. F. Mo, W. H. Mo, et al., 2018: Temporal and spatial distributions of rocky desertification in Guangxi karst area in the past 30 years. Guangxi Sci., 25, 625–631. doi: 10.13656/j.cnki.gxkx.20181031.001. (in Chinese)
    [9]

    Field, C. B., M. J. Behrenfeld, J. T. Randerson, et al., 1998: Primary production of the biosphere: Integrating terrestrial and oceanic components. Science, 281, 237–240. doi: 10.1126/science.281.5374.237.
    [10]

    Fu, X. F., S. T. Yang, and C. M. Liu, 2006: Spatial–temporal changes of NDVI and their relations with principal climatic factors in Yarlung Tsangpo River basin’s stations. Rese. Soil Water Conserv., 13, 229–232. doi: 10.3969/j.issn.1005-3409.2006.03.075. (in Chinese)
    [11]

    Guo, J., M. C. Li, and D. Y. Liu, 2009: Temporal variation of vegetation NDVI and its relations to climate factors in Tianjin area. Chinese J. Ecol., 28, 1055–1059. (in Chinese)
    [12]

    Han, S. Q., Y. H. Ran, J. X. Liu, et al., 2016: The changing distribution of rocky desertification in the Guangxi Region, 1930s to 2000. Acta Geogr. Sinica, 71, 390–399. doi: 10.11821/dlxb201603003. (in Chinese)
    [13]

    Hou, G. L., D. Y. Liu, Z. X. Zhang, et al., 2012: Response of NDVI in different climatic zone to climate change in Songnen plain. Chinese J. Agrometeorol., 33, 271–277. doi: 10.3969/j.issn.1000-6362.2012.02.019. (in Chinese)
    [14]

    Hu, B. Q., L. Li, and S. F. Jiang, 2005: Spatial pattern analysis on karst rocky desertification of Guangxi based on landscape spatial method. Earth Environ., 33, 581–587. doi: 10.3969/j.issn.1672-9250.2005.z1.122. (in Chinese)
    [15]

    Huang, X. S., H. Lu, X. P. Liao, et al., 2015: The distribution characteristics and defense strategies of droughts and floods of typical stony desertification area in Guangxi. J. Meteor. Res. Appl., 36, 59–61. doi: 10.3969/j.issn.1673-8411.2015.02.013. (in Chinese)
    [16]

    Keeling, C. D., J. F. S. Chin, and T. P. Whorf, 1996: Increased activity of northern vegetation inferred from atmospheric CO2 measurements. Nature, 382, 146–149. doi: 10.1038/382146a0.
    [17]

    Knapp, A. K., and M. D. Smith, 2001: Variation among biomes in temporal dynamics of aboveground primary production. Science, 291, 481–484. doi: 10.1126/science.291.5503.481.
    [18]

    Li, Y. B., S. J. Wang, Q. Tan, et al., 2006: Research development and problems of karst rocky desertification. Earth Environ., 34, 9–14. doi: 10.3969/j.issn.1672-9250.2006.03.002. (in Chinese)
    [19]

    Liu, C. L., R. H. Fan, J. J. Wu, et al., 2009: Temporal lag of grassland vegetation growth response to precipitation in Xilinguolemeng. Arid Land Geogr., 32, 512–518. (in Chinese)
    [20]

    Melillo, J. M., A. D. McGuire, D. W. Kicklighter, et al., 1993: Global climate change and terrestrial net primary production. Nature, 363, 234–240. doi: 10.1038/363234a0.
    [21]

    Meng, J. J., and J. Wang, 2007: The response of vegetation dynamics to climate change in the southwestern karst region of China since the early 1980s. Geogr. Res., 26, 857–865. doi: 10.3321/j.issn:1000-0585.2007.05.001. (in Chinese)
    [22]

    Nan, Y., Z. F. Liu, Y. H. Dong, et al., 2010: The responses of vegetation cover to climate change in the Changbai Mountain Area from 2000 to 2008. Sci. Geogr. Sinica, 30, 921–928. (in Chinese)
    [23]

    Nemani, R. R., C. D. Keeling, H. Hashimoto, et al., 2003: Climate-driven increases in global terrestrial net primary production from 1982 to 1999. Science, 300, 1560–1563. doi: 10.1126/science.1082750.
    [24]

    Su, W. C., H. Yang, Q. Li, et al., 2006: Rocky land desertification and its controlling measurements in the karst mountainous region, Southwest of China. Chinese J. Soil Sci., 37, 447–451. doi: 10.3321/j.issn:0564-3945.2006.03.008. (in Chinese)
    [25]

    Wang, J. H., W. H. Mo, Y. L. Chen, et al., 2014: Distribution characteristics and evolution of rocky and desertified land in Pingguo County of Guangxi based on “3S” techniques. Sci. Soil Water Conserv., 12, 66–70. doi: 10.3969/j.issn.1672-3007.2014.03.010. (in Chinese)
    [26]

    Wang, S. J., Y. B. Li, and R. L. Li, 2003: Karst rocky desertification: Formation background, evolution and comprehensive taming. Quat. Sci., 23, 657–666. doi: 10.3321/j.issn:1001-7410.2003.06.009. (in Chinese)
    [27]

    Weltzin, J. F., M. E. Loik, S. Schwinning, et al., 2003: Assessing the response of terrestrial ecosystems to potential changes in precipitation. BioScience, 53, 941–952. doi: 10.1641/0006-3568(2003)053[0941:ATROTE]2.0.CO;2.
    [28]

    Wu, L. L., S. W. Zhou, and S. K. Zheng, 2012: Study on spatial relations between rocky desertification and mountain climate in Northwest Guangxi Province based on GIS technology. J. Anhui Agric. Sci., 40, 11,787–11,790. doi: 10.3969/j.issn.0517-6611.2012.23.090. (in Chinese)
    [29]

    Xu, Y. F., M. Huang, W. S. Pan, et al., 2018: Variation characteristics of rocky desertification pattern in Qiannan prefecture of Guizhou. Ecol. Sci., 37, 105–113. doi: 10.14108/j.cnki.1008-8873.2018.04.013.
    [30]

    Yu, H. D., X. C. Yang, B. Xu, et al., 2013: Changes of grassland vegetation growth in Xilin Gol League over 10 years and analysis on the influence factors. J. Geo-Inf. Sci., 15, 270–279. (in Chinese)
    [31]

    Yuan, D. X., 2008: Global view on karst rock desertification and integrating control measures and experiences of China. Pratac. Sci., 25, 19–25. (in Chinese)
    [32]

    Zhang, J., X. F. Gu, X. J. Zhou, et al., 2009: Application of removing MODIS stripe noise using cubic spline interpolation. Microcomput. Inf., 25, 189–191. doi: 10.3969/j.issn.1008-0570.2009.31.081. (in Chinese)
    [33]

    Zhang, Q. Y., S. H. Wu, D. S. Zhao, et al., 2013: Responses of growing season vegetation changes to climatic factors in Inner Mongolia grassland. J. Nat. Resour., 28, 754–764. (in Chinese)
    [34]

    Zhang, S. J., T. Wang, T. M. Wang, et al., 2009: The variations in NDVI of different vegetation types in Xinjiang and its relation to climate factors. Pratac. Sci., 26, 26–31. (in Chinese)
    [35]

    Zhao, W. W., Y. H. Lv, W. W. Guo, et al., 2006: The correlations between climatic factors and NDVI in hilly and gully area of Loess Plateau in northern Shaanxi. Res. Soil Water Conserv., 13, 112–114. doi: 10.3969/j.issn.1005-3409.2006.02.036. (in Chinese)
    [36]

    Zheng, Y. F., L. Y. Niu, R. J. Wu, et al., 2009: Vegetation cover change in Guizhou of Southwest China in 1982–2003 in response to climate change. Chinese J. Ecol., 28, 1773–1778. (in Chinese)
    [37]

    Zhou, G. S., and X. S. Zhang, 1996: Study on NPP of natural vegetation in China under global climate change. Acta Phyto-ecol. Sinica, 20, 11–19. (in Chinese)
  • 13. Yanli CHEN.pdf

  • 加载中
通讯作者: 陈斌, bchen63@163.com
  • 1. 

    沈阳化工大学材料科学与工程学院 沈阳 110142

  1. 本站搜索
  2. 百度学术搜索
  3. 万方数据库搜索
  4. CNKI搜索

Changes in Vegetation and Assessment of Meteorological Conditions in Ecologically Fragile Karst Areas

    Corresponding author: Weihua MO, mwh-0419@163.com
  • 1. Institute of Meteorological Sciences, Guangxi Zhuang Autonomous Region (GZAR) Meteorological Bureau, Nanning 530022
  • 2. Hechi Meteorological Service, GZAR Meteorological Bureau, Hechi 547000
  • 3. College of Geography and Planning, Nanning Normal University, Nanning 530001
Funds: Supported by the Guangxi Zhuang Autonomous Region (GZAR) Science and Technology Project (AB20159022 and AB17292051) and GZAR Natural Science Foundation (2018GXNSFAA281338)

Abstract: Meteorological conditions have an important impact on changes of vegetation in ecologically fragile karst areas. This study aims to explore a method for quantitative evaluation of these meteorological conditions. We analyzed the changing trend of vegetation during 2000–2018 and the correlations between vegetation changes and various meteorological factors in karst rocky areas of Guangxi Zhuang Autonomous Region, China. Key meteorological factors in vegetation areas with varying degrees of improvement were selected and evaluated at seasonal timescale. A quantitative evaluation model of comprehensive influences of meteorological factors on vegetation was built by using the partial least-square regression (PLS). About 91.45% of the vegetation tended to be improved, while only the rest 8.55% showed a trend of degradation from 2000 to 2018. Areas with evident vegetation improvement were mainly distributed in the middle and northeast, and those with obvious vegetation degradation were scattered. Meteorological factors affecting vegetation were significantly different among the four seasons. Overall, high air humidity, small temperature difference in spring and autumn, and low daily minimum temperature and air pressure were favorable conditions. Low temperature in winter as well as high temperature in summer and autumn were unfavorable conditions. The Climate Vegetation Index (CVI) model was established by PLS using the maximum, minimum, and average temperatures; vapor pressure; rainfall; and air pressure as key meteorological factors. The Enhanced Vegetation Index (EVI) was well fitted by the CVI model, with the average coefficient of determination (r2) and root mean square error (RMSE) of 0.856 and 0.042, respectively. Finally, an assessment model of comprehensive meteorologi-cal conditions was built based on the interannual differences in CVI. The meteorological conditions in the study area in 2014 were successfully evaluated by combining the model and selected seasonal key meteorological factors.

1.   Introduction
  • Karst, loess, desert, and alpine regions are four ecologically fragile areas in China. In the karst areas of fragile ecological environment, the gradual exposure of rocks similar to desert landscape on the earth’s surface due to vegetation destruction, soil erosion, and decline or loss of land production capacity is called rocky desertification (Yuan, 2008). The low disaster-tolerance capacity of these areas is a major hidden danger of ecological security and has attracted great attention from all walks of life (Li et al., 2006).

    The karst rocky areas in southwestern China are located between the tropics and subtropics. The hot and rainy climate in the subtropical southwestern China provides a powerful driving force for karst development, water and soil loss, and rocky desertification (Wang et al., 2003; Su et al., 2006). The correlation (Melillo et al., 1993; Keeling et al., 1996; Zhou and Zhang, 1996; Field et al., 1998; Knapp and Smith, 2001; Nemani et al., 2003; Weltzin et al., 2003; Chen et al., 2010; Zhang et al., 2013) and hysteresis (Guo et al., 2009; Liu et al., 2009; Chen et al., 2010; Hou et al., 2012; Yu et al., 2013) between the climate and vegetation changes have been proposed and verified at global and regional scales. Researchers have also studied the vegetation and climate changes in karst rocky areas.

    By using the Global Inventory Modeling and Mapping Studies (GIMMS) Normalized Difference Vegetation Index (NDVI) and Net Primary Productivity (NPP) datasets, Meng and Wang (2007) studied the response of vegetation change to climate change in southwestern karst rocky areas of China and found the greater effect of temperature change than that of precipitation change. By extracting and analyzing pixel-by-pixel information through regression and correlation analyses, Zheng et al. (2009) investigated the change in vegetation cover and its relationship to major climatic factors in Guizhou, Southwest China based on GIMMS NDVI data and corresponding climatological data during 1982–2003. The interannual variation trends of the NDVI and temperature are synchronous, and a certain lag exists between NDVI and precipitation. Wu et al (2012) analyzed the correlation between the rocky desertification and spatial distributions of climatic factors in karst mountainous areas of northwestern Guangxi Zhuang Autonomous Region (GZAR) of China in 2008. The incidence of rocky desertification increased steadily with the increasing annual average temperature and increased rapidly with the increasing annual average precipitation.

    In the existing studies on the correlation between vegetation and climate in karst rocky areas, NDVI is often used to characterize vegetation, and climatic factors are only represented by temperature and precipitation. However, the influence of climate on vegetation is strongly integrated. In most studies, NDVI is also used to characterize the vegetation change (Fu et al., 2006; Zhao et al., 2006; Zhang S. J. et al., 2009; Chen et al., 2011). The Enhanced Vegetation Index (EVI) has the advantages of the widely used NDVI, but with NDVI’s shortcomings being improved. In recent years, studies on the correlation between the vegetation and climate in karst rocky areas of China based on EVI sequences have made progress (Nan et al., 2010; Chen et al., 2015b).

    Located in southern China, GZAR is the birthplace of many rivers, such as Xiang and Xi rivers, and has an irreplaceable preponderance in ecological resources (Ban et al., 2018). The karst rocky areas in southwestern GZAR are one of the most important areas for biodiversity conservation. However, the widespread distribution of the karst landform, prominent rocky desertification, and serious natural disasters (Huang et al., 2015) have posed increasing threats to the regional ecological security and have seriously restricted local economic and social development. Therefore, the ecology of karst rocky areas in GZAR is always a focus and hot topic of the local government and researchers in this region.

    The authors of this paper previously carried out considerable research on karst rocky areas of GZAR. Based on multi-source remote sensing data, they established an identification model of rocky desertification, drew spatial distribution maps of rocky desertification in different periods, and studied the distribution characteristics of rocky desertification in karst rocky areas of GZAR (Hu et al., 2005; Wang et al., 2014; Han et al., 2016; Chen et al., 2018). The sensitivity of NDVI and EVI to vegetation monitoring in some rocky desertification counties in the karst rocky areas was compared. The EVI considers influences of soil background and is more objective in reflecting the vegetation characteristics in the areas (Chen et al., 2014b). The correlation and hysteresis of EVI and meteorological factors were analyzed. The effects of various meteorological factors on EVI were studied. The response of EVI to meteorological factors is sensitive, and their correlation is very high (Chen et al., 2014a, 2015a). According to the response mode of the two indices, the climate fitting model of EVI was established by using the stepwise regression method for fitting and predicting EVI accurately (Chen et al., 2015a).

    Many results have been obtained in studies of the karst rocky areas in GZAR. Although the multi-tempospatial-scale distributions of rocky desertification has been interpreted, its interannual vegetation evolution remains unclear. Meteorological factors have an important influence on vegetation in karst rocky areas, but no models and factors are established to quantitatively evaluate the meteorological conditions. In this study, a mathematical model was used to quantitatively analyze the structural characteristics of the rocky desertification grade change. According to the characteristics of meteorological conditions in vegetation change areas, a quantitative evaluation model of meteorological conditions affecting vegetation EVI in the karst rocky areas of GZAR was established by using the idea of the climate fitting model of vegetation EVI. The results can provide reference for evaluation of the meteorological conditions and rocky desertification pattern change in karst rocky areas.

2.   Data and methods
  • The GZAR (20°54′–26°24′N, 104°26′–112°04′E) is located in southwestern China, with the Tropic of Cancer crossing in the middle. The region is bordered by the tropical ocean in the south, Nanling Mountains in the north, and Yunnan–Guizhou Plateau in the west. The region is higher than the surrounding areas but low in the middle and looks like a basin. The region has more mountains but fewer plains, and the area of karst landform accounts for 37.8%. The poor land, harsh ecologi-cal environment, prominent rocky desertification, and serious natural disasters seriously restrict local economic and social development, so it is a poverty-stricken area. The distribution of karst rocky areas in Guangxi is shown in Fig. 1.

    Figure 1.  Distribution of rocky desertification areas and meteorological stations in the Guangxi Zhuang Autonomous Region (GZAR).

  • EVI data, in format of V005, are the Moderate-resolution Imaging Spectroradiometer (MODIS) vegetation index product MOD13Q1 developed by the NASA MO-DIS terrestrial product group according to statistical algorithm; namely the global synthetic vegetation index product that was synthesized in 16 days and has a resolution of 250 m. The MODIS vegetation index product is improved and designed based on the existing vegetation indices, including two vegetation index products (NDVI and EVI). MODIS NDVI is a continuation of NOAA NDVI series accumulated for 20 yr and can provide long-term data for business monitoring and research. EVI takes advantage of MODIS radiometer to correct surface reflectance so as to increase sensitivity to high biomass areas and to improve vegetation monitoring accuracy by coupling the canopy background signal and reducing the atmospheric effects.

    The EVI data used were obtained by using the internationally accepted maximum value composite (MVC) method, which can further eliminate the interference of clouds, atmosphere, and solar elevation angle.

    $${\rm{EV}}{{\rm{I}}_{{\rm{m}}i}} = \max \left({{\rm{EV}}{{\rm{I}}_{ij}}} \right),$$ (1)

    where EVImi is the maximized composite value of EVI in the ith 16-day, and EVIij is the jth-day value during the period.

    The obtained MOD13Q1 remote sensing dataset was preprocessed by the subset extraction, image mounting, data format conversion, projection conversion, and quality inspection to obtain a reliable EVI dataset. Many cloud pollution pixels still existed in the GZAR EVI dataset because of the influence of cloud and rainy weather. In this study, the spline interpolation method was used to deal with cloud pollution pixels and to reconstruct the high-quality EVI data series (Zhang J. et al., 2009).

  • The daily water vapor pressure, precipitation, maxi-mum temperature, minimum temperature, average temperature, dew point temperature, sunshine duration, and air pressure at 25 meteorological stations from 2000 to 2018 were provided by the Meteorological Information Center of GZAR Meteorological Bureau. The 16-day statistical values of various climatic factors and EVI in the corresponding periods (including the same and previous periods) were calculated. Precipitation and sunshine duration are cumulative values, and the remaining clima-tic factors are averages in the periods. The meteorologi-cal data and remote sensing EVI were matched by the near-distance matching method; that is, the vegetation EVI of karst areas in the county is matched with the data of meteorological observation stations in that county.

  • Geographic information data include the administrative boundary of counties, vector boundary of karst rocky areas, and latitude and longitude information of meteorological stations in GZAR.

  • In the trend line method, the regression analysis is performed on a set of variables that change over time to predict their changing trends. This method can be used to simulate the interannual changing trend of EVI. The calculation formula is as below:

    $${\rm{SLOPE}} = \frac{{n \times \displaystyle\sum\limits_{k = 1}^n {k \times {\rm{EV}}{{\rm{I}}_{{{k}}}}} - \displaystyle\sum\limits_{k = 1}^n {k\sum\limits_{k = 1}^n {{\rm{EV}}{{\rm{I}}_{{{k}}}}} } }}{{n \times \displaystyle\sum\limits_{k = 1}^n {{k^2}} - {{\Bigg(\displaystyle\sum\limits_{k = 1}^n k\Bigg)}^2}}},$$ (2)

    where k is the number of years from 1 to n; EVIk is the average of EVI in the kth year; and SLOPE is the changing trend. If SLOPE > 0, EVI increases in n years, that is, the vegetation in the area is improved; otherwise it degrades.

  • The EVI value of a certain area was calculated by using the averaging method, namely, the EVI average of all pixels in the statistical area. The formula is as follows:

    $${\rm{EV}}{{\rm{I}}_{ap}} = \sum {{\rm{EV}}{{\rm{I}}_{x,y}}} /n,\hspace{84pt} $$ (3)

    where EVIap is the EVI average of a certain area (p is the code of the area), x is the line number of pixels, y is the column number of pixels, and n is the total number of pixels in the statistical area.

  • PLS is a multivariate statistical analysis method that combines the multiple linear regression analysis, principal component analysis of variables, and canonical correlation analysis between variables to make full use of sample information. Regression modeling can be performed under conditions of the small sample size, multiple independent variables, and severe multiple correlations.

    Taking the PLS of a single dependent variable as an example, the principle is as follows. Let the known dependent variable be Y, and k independent variables be x1, x2, ···, xk, and the number of samples be n, which constitute a data table X = [x1, x2,···, xk]n × k and Y = [y]n × 1. Extracted from X, component t1 is a linear combination of x1, x2, ···, xk, which should carry the information of variation in X as much as possible and the degree of correlation with Y is the largest. After the first principal component t1 is extracted, the regression of Y and X is performed. If the regression equation has reached the satisfactory accuracy at this time, the algorithm is stopped; otherwise, the second principal component t2 is extracted by using the participation information after X is explained by t1 and Y is explained by t2, and the regression of Y and X to t1 and t2 is continued until the satisfactory accuracy is achieved. If a total of m components t1, t2, ···, tm (mn) are finally extracted from X, the PLS will be used for the regression of Y to t1, t2, ···, tm. Since t1, t2, ···, tm are linear combinations of x1, x2, ···, xk, they can be expressed as a regression equation of dependent variable Y and independent variable X.

    In the PLS analysis, the variable projection importance index (VIP) is used to measure the explanatory power of the independent variable to the dependent variable. This index is defined as follows:

    $$\begin{array}{*{20}{l}} {{\rm{VI}}{{\rm{P}}_j} = \sqrt {\frac{{k \displaystyle\sum\limits_{h = 1}^m {{\rm{Rd}}} (y,{t_h})W_{h\!j}^2}}{{\displaystyle\sum\limits_{h = 1}^m {{\rm{Rd}}} (y,{t_h})}}},\;\;\;\;\;\;\;\;(j = 1,\; 2, \cdots, \; k)} \end{array},$$ (4)
    $$\hspace{4pt} {\rm{Rd}}(y,{t_h}) = {r^2}(y,{t_h}),$$ (5)

    where Whj is the jth component of the axis Wh, which is used to measure the contribution of xj to the structure component th; r(y, th) is the correlation coefficient between the dependent variable y and component th; k is the number of independent variables; and m is the number of components. For xj with a large VIPj, xj plays a more important role in explaining y.

  • The coefficient of determination (r2) and root mean square error (RMSE) were used to test the model’s accuracy. The larger the r2 is, the better the model is. The smaller the RMSE is, the better the fitting effect is.

3.   Results
  • The univariate regression trend line method was used to study the changing trend of vegetation in karst rocky areas of GZAR from 2000 to 2018. Given that the SLOPE values were mostly concentrated within −0.2 to 0.2, the classification standards of SLOPE were formulated according to Table 1, and corresponding vegetation change categories and percentages were derived.

    GradeChange of vegetationRange of SLOPEProportion (%)
    1Noticeably worse [−1.0, −0.2) 0.04
    2Noticeably worse [−0.2, −0.15) 0.08
    3Noticeably worse [−0.15, −0.1) 0.22
    4Worse [−0.1, −0.05) 0.84
    5Slightly worse [−0.05, 0) 7.37
    6Slightly better [0, 0.05)58.37
    7Better [0.05, 0.1)30.35
    8Noticeably better [0.1, 0.15) 2.55
    9Noticeably better [0.15, 0.2) 0.18
    10Noticeably better [0.2, 1.0) 0.01

    Table 1.  The SLOPE grading standard in rocky desertification areas of Guangxi Zhuang Autonomous Region (GZAR)

    According to Table 1 and Fig. 2, 91.45% of vegetation in karst rocky areas of GZAR tended to improve from 2000 to 2018, in which 58.37% of the vegetation was improved slightly, 30.35% improved, and 2.73% improved significantly. Only 8.55% of the vegetation trended to deteriorate, in which 7.37% deteriorated slightly, 0.84% deteriorated, and 0.34% deteriorated significantly. Areas with obvious vegetation improvement were mainly distributed in Laibin and Liuzhou cities in the middle of karst rocky areas. Areas with obvious vegetation degradation were scattered, of which the degradation in Chongzuo, Guilin, and Nanning cities was more obvious than that in the other areas.

    Figure 2.  The vegetation change trend in karst rocky areas of GZAR from 2000 to 2018.

  • Meteorological conditions are very important for vegetation growth in karst rocky areas, and influences of meteorological factors on vegetation vary in different periods. When a business service product report is established, the impact of meteorological factors needs to be specified, such as the specific impact of various meteorological factors in different seasons. Therefore, the correlation between EVI and meteorological factors in the four seasons including spring (March–May), summer (June–August), autumn (September–November), and winter (December–next February) was discussed. The correlation between the same meteorological factors and vegetation in different areas are different because of the difference in topography and vegetation—Not only the size difference but also the positive and negative correlations may exist at the same time. In our study, the correlation between meteorological elements and EVI was valid only if 80% of the stations passed through the significance test. If 80% of the sites are positively (or negatively) correlated with a weather factor, we assume that the correlation is positive (or negative).

    The influence of meteorological factors on vegetation is quite different in different seasons (Table 2). In spring, the three meteorological factors (water vapor pressure, relative humidity, and dew point) related to humidity were positively correlated with EVI, that is, the more humid the air was, the better the vegetation grew. In summer, the minimum temperature that generally occurred in the early morning was negatively correlated with EVI, so the low temperature at night was not conducive to vegetation growth. In autumn, precipitation was positively correlated with EVI, while the average temperature and minimum temperature were negatively correlated with EVI. The high temperature and less rainfall in autumn in GZAR could easily lead to droughts. Sufficient precipitation and mild temperature in autumn were conducive to vegetation growth. In winter, vapor pressure and dew point were positively correlated with EVI, that is, humid air was conducive to vegetation growth. The average temperature, maximum temperature, minimum temperature, and sunshine duration were positively correlated with EVI. That is, warm winter was conducive to vegetation growth. If temperature was too low, soil in karst rocky areas was not conducive to heat storage, and vegetation roots were likely to be frozen, which was not advantageous to vegetation growth.

    PeriodVapor pressureRainfallRelative humidityAverage temperatureMaximum temperatureMinimum temperatureSunshine durationAir pressureDew point
    Spring+/+/////+
    Summer///////
    Autumn/+/////
    Winter+//+++++
    Note: + indicates the significant positive correlation, − indicates the significant negative correlation, and / indicates that the correlation is not significant.

    Table 2.  The correlation between vegetation EVI and meteorological factors in karst rocky areas of GZAR

    We further analyze the characteristics of meteorological factors in different vegetation improvement/degradation areas. The vegetation improvement was obvious when SLOPE ≥ 0.05 according to analysis of the vegetation change trend. The vegetation change in these areas had less man-made disturbance, but was mainly due to influences of meteorological factors. The sites included in the study area were divided into five categories, and the characteristics of meteorological factors of each category were examined. The five categories were defined according to the proportion of pixels with SLOPE ≥ 0.05 in each county, and the proportion intervals were > 80%, 70%–80%, 60%–70%, 50%–60%, and < 50%. Results were listed in four seasons (Table 3).

    PeriodCategory (%)Vapor pressure (hPa)Rainfall (mm)Relative humidity
    (%)
    Average temperature (°C)Maximum temperature (°C)Minimum temperature (°C)Sunshine duration
    (h)
    Air
    pressure (hPa)
    Dew
    point
    (°C)
    Spring< 5016.5170.767.421.328.416.4443.8928.614.0
    50–6017.8218.674.920.225.816.3352.6939.715.0
    60–7019.0297.175.521.126.117.6293.5976.316.1
    70–8019.9273.476.521.626.218.4292.9972.716.8
    > 8020.2297.077.221.625.818.5259.5980.917.0
    Summer< 5025.2625.879.325.430.922.0500.0922.721.2
    50–6026.3770.783.625.130.121.8477.8935.321.8
    60–7028.1861.181.126.831.923.7483.6970.222.9
    70–8028.1797.580.726.831.623.7516.7966.222.9
    > 8028.9838.781.127.231.824.2494.9974.623.4
    Autumn< 5018.6173.978.420.326.116.7410.0933.115.9
    50–6018.8199.480.519.825.316.2431.0944.015.9
    60–7019.6204.176.621.327.117.9439.7980.116.5
    70–8020.2211.575.921.927.418.3482.3975.316.9
    > 8020.3222.273.822.527.619.0490.9983.717.0
    Winter< 5010.659.375.612.418.58.5307.2934.97.4
    50–6010.770.477.211.716.88.3275.4946.27.3
    60–7011.1105.475.212.417.29.4243.5984.07.7
    70–8012.099.074.713.718.210.6274.7980.08.7
    > 8011.8114.674.113.517.910.6256.2988.88.4

    Table 3.  Values of climatic factors in the five categories of SLOPE

    In karst rocky areas, as the proportion of pixels with SLOPE ≥ 0.05 increased, water vapor pressure, rainfall, average temperature, minimum temperature, air pressure, and dew point mostly showed increasing change trends. That is, the higher average temperature, minimum temperature, air pressure, dew point, and sufficient water vapor and rainfall are better for vegetation improvement (Table 4).

    PeriodVapor pressureRainfallRelative humidityAverage temperatureMaximum temperatureMinimum temperatureSunshine durationAir pressureDew point
    SpringHHHH/HLHH
    SummerHH/H/H/HH
    AutumnHHLHHHHHH
    WinterHH /L//HLH //
    Note: H indicates that the high-value feature is obvious, L indicates that the low-value feature is obvious, and / indicates that the feature is not obvious.

    Table 4.  Characteristics of meteorological elements in areas where vegetation was obviously improved

    According to the correlation between EVI and climatic factors (Table 2) as well as climate characteristics of different rocky desertification improvement areas in different seasons (Table 4), the number of rainy days, water vapor pressure, rainfall, average temperature, maximum temperature, minimum temperature, and air pressure were selected as key factors for evaluating the meteorological conditions in karst rocky desertification areas of GZAR (Table 5). The correlation between water vapor pressure and vegetation EVI was the highest, and this was more obvious in various vegetation improvement areas. The relative humidity, dew point, and sunshine hours were not included, considering that water vapor pressure was enough to reflect air humidity and sunshine duration varied irregularly in different grades of vegetation improvement . In specific cases, the multi-year average values of meteorological factors are used.

    PeriodVapor pressureRainfallAverage temperatureMaximum temperatureMinimum temperatureAir pressure
    Spring
    Summer
    Autumn
    Winter
    Note: ↑ indicates that a high value is favorable; ↓ indicates that a low value is favorable.

    Table 5.  Selected key meteorological factors that have a strong correlation with vegetation change in the study area

  • According to the above analysis, meteorological factors had significant effects on EVI in karst rocky areas of GZAR. In a previous study, it was found that interactions exist among the meteorological factors (Chen et al., 2014a). The EVI characterizing vegetation growth is in fact the combined result of the comprehensive meteorological conditions. The stepwise regression method was used to establish the fitting model of EVI and meteorological factors in the previous study (Chen et al., 2015a). Establishment of the model requires to understand the effects and interactions of meteorological factors, which involves complex data operations. In this work, the PLS method was used to establish such a model according to the multi-collinearity of statistical samples. The independent variables include water vapor pressure, precipitation, average temperature, maximum temperature, minimum temperature, and air pressure; and the dependent variable is EVI.

    In the principal component analysis of PLS modeling, the variation percentages explained by the components Tk (k = 1, 2,···, 6) extracted from the independent variable group are 83.0%, 8.4%, 5.7%, 2.3%, 0.3%, and 0.1%, respectively. The variation percentages of the dependent variable group explained by Tk (k = 1, 2,···, 6) are 80.7%, 2.5%, 1.8%, 0.9%, 1.2%, and 0.7%, respectively. The cumulative contribution rate of x variance of the first three eigenvectors has reached 97.2%; that is, the basic characteristics of the independent variable X can be captured. In the meantime, the cumulative variance contribution of Y is 87.8%. That is, the regression equation constructed by the first three principal components can achieve a satisfactory accuracy (Table 6). The final regression equation is as below:

    Latent factorX varianceCumulative X varianceY varianceCumulative Y variance
    10.8300.8300.8070.807
    20.0840.9150.0250.832
    30.0570.9720.0180.850
    40.0230.9950.0090.859
    50.0030.9980.0120.871
    60.0010.9990.0070.878

    Table 6.  The variation percentage explained by the partial least-square (PLS) factors

    $$\begin{split} {\rm{CVI}} = & - 6.838 + 0.015V + 0.0001R - 0.067T \\ &+ 0.038{T_{\max }} + 0.032{T_{\min }} + 0.007P, \end{split}$$ (6)

    where the Climate Vegetation Index (CVI) is the meteorological fitting value of vegetation EVI; V is vapor pressure; R is rainfall; T is average temperature; Tmax is maximum temperature; Tmin is minimum temperature; and P is air pressure.

    Each variable of the regression equation had an important effect on vegetation in karst rocky areas in the analysis of VIP (Table 7). The factors that dominated vegetation in karst rocky areas were ranked as follows: vapor pressure (V) > average temperature (T) > maximum temperature (Tmax) > minimum temperature (Tmin) > air pressure (P) > rainfall (R).

    IndexVTTmaxTminPR
    VIP1.0871.0661.0651.0530.9880.842

    Table 7.  The analysis of variable projection importance indices

    The regression equation was tested, with r2 of 0.856 and adjusted r2 of 0.852, which passed the significance test. EVI observation values and their fitting values (CVI) are basically distributed near the 1:1 line in the scatter plots of the two (Fig. 3). The CVI model has high fitting precision and good fitting effect (Table 8).

    Figure 3.  Scatter plots of EVI and CVI during 2000–2018.

    ModelSum of squaresdfr2Adjusted r2Std. error of the estimateFSig.
    12.33560.8560.8520.042220.3250.000
    0.394431
    2.728439
    Note: degree of freedom—df, coefficient of determination—r2, standard deviation—Std., Fisher test—F, significance—Sig.

    Table 8.  Error and variance analyses of the partial-least square regression model

    CVI actually characterizes the meteorological potential value of EVI, namely, the vegetation EVI under the combined influence of meteorological conditions and without the influence of other factors. If the weather conditions are good, EVI is high; if weather conditions are poor, EVI is low. Based on this assumption, the difference in meteorological conditions can be determined by the difference in CVI, and the evaluation model of vegetation meteorological conditions in karst rocky areas can be obtained as below:

    $${\rm{DCVI = CV}}{{\rm{I}}_i} - {\rm{CV}}{{\rm{I}}_j},$$ (7)

    where CVIi and CVIj are the CVI values of vegetation in karst rocky areas in years i and j. If DCVI is greater than zero, the overall meteorological conditions in karst rocky areas in the evaluated year is superior to that of the comparison year.

    In ecological service materials released to public, a simple and intuitive evaluation of meteorological conditions is required. For example, images with visually good, medium, and poor ratings are more easily understood by readers. From 2000 to 2018, 91.45% of vegetation in karst rocky areas of GZAR tended to be improved, of which 58.37% of the vegetation was slightly improved, and 33.08% was improved or noticeably improved (Table 1). The meteorological condition was judged to be moderate in areas with slight vegetation improvement, and that in areas with improvement or noticeable improvement was judged to be good. The DCVI values from 2000 to 2018 was calculated in these areas, and the classification standard of meteorological conditions was formulated (Table 9).

    GradeGoodGeneralBad
    Classification standard[0.02, 1][0, 0.02)[−1, 0)

    Table 9.  The classification standard of meteorological conditions in karst rocky areas of GZAR

  • The model of vegetation meteorological conditions was used to evaluate the meteorological conditions in karst rocky areas of GZAR in 2014. The results of the 2014 assessment of the meteorological conditions were compared with those of 2013. In 2014, the meteorological conditions were good mainly in the west and middle of the study area, including Tianlin, Lingyun, Bama, Donglan, Napo, and Jingxi counties, northern Longlin county, western and eastern Xilin county, and the area outside the central part of Fengshan county. The meteorological conditions were medium mainly in the middle and south, including Leye, Tianyang, Debao, Tiandeng, Daxin, Long’an, Longzhou, Dahua, and Xincheng counties, western and southern Du’an county, and central Xilin county, central and southern Longlin county, eastern Luocheng county, and southern Rong’an county. The overall meteorological conditions in the above areas were conducive to vegetation growth. However, the meteorological conditions in Huanjiang, Rongshui, Shanglin, and Ningming counties, western and northern Luocheng county, central and northern Rong’an county, northern Du’an county, and southeastern Long’an county were slightly worse, which was not conducive to vegetation growth (Fig. 4).

    Figure 4.  The evaluation results of meteorological conditions in karst rocky areas of GZAR in 2014.

    According to the meteorological evaluation factors in the rocky desertification areas, the advantages and disadvantages of the meteorological conditions were analyzed, and the ecological meteorological evaluation model was used to analyze the overall impact of meteorological conditions on vegetation improvement in rocky desertification areas. Favorable conditions were the high air humidity; small temperature difference in spring, autumn, and dry season; and low daily minimum temperature and air pressure. Unfavorable conditions included the quite frequent rainy days, low temperature in winter, and high temperature in summer and autumn.

    The meteorological conditions in spring and summer were quite similar between 2013 and 2014. In 2014, the meteorological conditions were slightly worse in the early stage but slightly better in the later period. Specifically, from December 2013 to February 2014, the weather was cold and persistently cloudy, which was not conducive to the safe wintering of vegetation in the study area; however, the precipitation in the autumn of 2014 increased by 107.6 mm, compared to 2013, which was beneficial to vegetation growth (Table 10).

    Meteorological factor20132014Multi-year
    average
    Vapor pressure (spring; hPa)20.020.719.4
    Minimum temperature (summer; °C)23.824.023.6
    Air pressure (summer; hPa)965.1964.8964.8
    Rainfall (autumn; mm)239.9347.5210.6
    Average temperature (autumn; °C)20.922.121.6
    Minimum temperature (autumn; °C)17.819.118.1
    Average temperature (winter; °C)12.711.513.0
    Maximum temperature (winter; °C)16.416.717.7
    Minimum temperature (winter; °C)10.38.19.9
    Vapour pressure (winter; hPa)11.910.011.5

    Table 10.  Meteorological elements in rocky desertification areas of GZAR in 2013 and 2014

4.   Discussion
  • From 2000 to 2018, most of the karst rocky areas of GZAR had improved vegetation, while areas where vegetation degraded obviously were scattered. In the future research, the reasons for vegetation degradation should be elucidated to provide additional scientific reference for government decision making. The obvious improvement trend of vegetation fully demonstrates that a series of measures, such as closing hillsides to facilitate afforestation and returning farmland to forests implemented by the GZAR government, have achieved remarkable results in recent years. But significant vegetation degradation still occurred in some areas. Xu et al. (2018) studied the changes in the rocky desertification pattern in Qiannan Prefecture, Guizhou Province and found that potential rocky desertification could also be transformed into severe rocky desertification due to vegetation degradation. This finding indicated that the vegetation ecology in karst rocky areas was highly vulnerable.

    The response of vegetation to the climate in karst rocky areas has special characteristics. Rock desertification in these areas occurs at different degrees, and the rate of exposed bedrock is relatively large. Hence, the soil layer is shallow and contains mostly stone gravel soil, and its vegetation community has poor ability to improve the microclimate. In the absence of buffering of the upper-layer vegetation, the temperature increase in the daytime is large, and the heat dissipation is fast at nighttime due to the small specific heat capacity of rock, resulting in a rapid change in temperature, humidity, and surface temperature, as well as a great change in temperature difference in the areas. Therefore, meteorological conditions had complicated impact on vegetation, and the main meteorological factors affecting vegetation in the areas in the four seasons were quite different (for details on seasonal differences, see Section 3.2.1).

    Precipitation has a significant effect on vegetation only in autumn probably because of the large spatial and temporal variability of rainfall in GZAR and the big difference in the vegetation canopy density due to different proportions of shrubs as well as shrubs and trees growing in different regions. For areas with sparse vegetation, the water retention capacity is poor, and the excessive concentration of precipitation is not conducive to vegetation growth and may even aggravate water and soil losses. However, for karst rocky areas with high canopy density, the soil layer is thicker and has strong water retention capacity, so precipitation will promote vegetation growth.

    In this paper, the fitting model of EVI and meteorological factors was established by using the PLS method. In a previous study, the climate fitting model of vegetation EVI was established by using the stepwise regression method (Chen et al., 2015a). In comparison, the fitting model of EVI established by using the PLS method has higher simulation accuracy possibly due to the different factors of the models. The response of vegetation growth to meteorological factors has a lag effect (Zheng et al., 2009; Chen et al., 2014a). The method of using early meteorological factors for modeling may be an effective way to improve the accuracy of model simulation. When the meteorological fitting model of the vegetation EVI was established by the stepwise regression method, the mixed model based on early meteorological factors has higher simulation accuracy (Chen et al., 2015a). In addition, vegetation growth is affected by many factors, such as the soil type and lithology, which needs to be further studied in model optimization.

5.   Conclusions
  • During 2000–2018, 91.45% of the vegetation in karst rocky areas of Guangxi Zhuang Autonomous Region (GZAR) was improved, while only 8.55% tended to degrade. Areas where vegetation was obviously improved were mainly distributed in the middle and northeast of karst rocky areas. Areas with obvious vegetation degradation were scattered, and the vegetation degradation in the southwest was more obvious than that in the other areas.

    The meteorological factors affecting vegetation in karst rocky areas were significantly different in the four seasons. Favorable conditions included the high air humidity, small temperature difference in spring and autumn, and low daily minimum temperature and air pressure. Unfavorable conditions included low temperature in winter as well as high temperature in summer and autumn.

    The Climate Vegetation Index (CVI) model in karst rocky areas was established by the PLS method using the maximum, minimum, and average temperatures; water vapor pressure; rainfall; and air pressure as key meteorological factors. EVI was well fitted by the CVI model, with average r2 and RMSE of 0.856 and 0.042, respectively. Finally, the assessment model of the comprehensive meteorological conditions in karst rocky areas was built based on interannual differences in CVI. The meteorological conditions in the study area in 2014 were successfully evaluated by combining the model and the selected seasonal key meteorological factors.

    Acknowledgments. The authors thank the China Center for Resource Satellite Data and Application for providing Huan Jian (HJ)-1 Charge Coupled Device (CCD) remote sensing data and technical support.

Reference (37)

Catalog

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return