*5.3. Response Characteristics of Runoff Changes to Climatic Factors at Different Time Scales* 5.3.1. Response of Runoff Changes to Climatic Factors on Monthly Scale

The cross wavelet power spectra of monthly rainfall and runoff, as shown in Figure 13, illustrate that the interaction between monthly rainfall and runoff is mainly concentrated in the main cycle at 12-month time scales from 1984 to 2015, which indicates a significant correlation between them at 1-year scale. The interaction between monthly rainfall and runoff is also shown in two subcycles at approximately 24-month time scales (1993–1996

and 2007–2013) and 72-month time scales (1993–2008). The energy difference in main cycles and subcycles in the time domain reflects that rainfall in different years has different regulating effects on runoff change.

**Figure 13.** Cross wavelet spectra and wavelet coherence spectra of monthly rainfall, temperature and evaporation with runoff in the Yinjiang River watershed from 1984 to 2015, respectively. The thick black contour designates the 5% significance level against red noise and the cone of influence (COI) where edge effects might distort the picture is shown as a lighter shade. Arrows denote relative phase difference: The arrows from left to right indicate that the influencing factors and runoff are in the same phase, which implies a positive correlation; the arrows from right to left indicate an inverse phase, which implies a negative correlation; the downward arrows indicate that the influence factor is 90◦ ahead of the runoff change and the upward arrows indicate that the influence factor is 90◦ lagging the runoff.

The cross wavelet coherence spectra can compensate for the lack of correlation analysis of the cross wavelet power spectra in the low-energy region. The cross wavelet condensation spectra has a larger time–frequency domain space compared with the cross wavelet power spectra. In addition to the positive correlation at 12-month time scales between 1984 and 2015, a positive correlation is also observed in subperiods between 1991 and 2008 (72-month time scales) and 1984 and 2015 (4–8 months). By contrast, a negative secondary cycle occurs in 1996–2004 (36-month time scales). The contribution of rainfall to runoff in the Yinjiang River watershed can also be determined from the phase relationship between the cross wavelet power spectra and the cross wavelet condensation spectra. The main periods of 4–8-, 12- and 72-month time scales show a significantly positive correlation. This result is mainly because soil moisture is easy to be saturated due to the increase in rainfall on this scale, thus accelerating the formation of slope runoff within the river watershed and forming an effective replenishment for the river. On the contrary, the phase is negatively correlated in the subcycle at 36-month time scales, which indicates that the influence of rainfall fluctuation on runoff changes is negative. On the one hand, runoff will increase with the increase in rainfall on the river recharge. On the other hand, the increasing rainfall will increase vegetation coverage, enhance water conservation, increase evaporation and eventually reduce river recharge. The high-correlation region of the cross wavelet condensation spectra is basically consistent with the high-energy region of the cross wavelet power

spectra. A strong correlation exists between the monthly rainfall and runoff at 64-month time scales from 1995 to 2008, and the phase angle is 30◦, which indicates that the rainfall lags behind the runoff for 1 month.

The monthly evaporation, temperature and runoff have main periods of 12-month time scales in high-energy areas. In low-energy regions, not only a high correlation at 12-month time scales but also a local significant correlation between the effects of temperature and evaporation on runoff changes at 1–8-month time scales are determined. Thus, temperature is mainly regulated indirectly by controlling evaporation. The influence of evaporation on runoff will superimpose the influence of temperature due to the indirect regulation of temperature on runoff because their influence on runoff is generally consistent in different scales but slightly different in different years.

The effects of monthly rainfall, temperature and evaporation on runoff are positively correlated in the main cycle at 12-month time scales, which indicates that their effects on runoff are positive and mainly at the annual scale. The main periodic bandwidth of rainfall on runoff changes is wider than that of temperature and evaporation, which indicates that rainfall is the main factor that affects runoff variation. In other cycles, the phases of rainfall's impacts on runoff changes are the interlacing phenomena of positive and negative, whereas the phases of temperature and evaporation that affect runoff changes are mainly negative. Accordingly, the impact of rainfall fluctuation on runoff changes on this scale is both positive and negative, whereas that of evaporation is always negative. However, temperature and evaporation have negative effects on runoff in each subcycle, which may be because evaporation increases with the increase in temperature, thus reducing runoff recharge.

The response of runoff to rainfall is timely in the high-energy region and the lowenergy significant-correlation region. There has been shown a positive correlation with a smaller phase angle, but there is also a slight lag at 16-month time scales. The phase angles of evaporation, temperature and runoff range from 30◦ to 45◦, which demonstrates that runoff changes have lagged behind temperature and evaporation for 1–2 months. The similarities of the effects of monthly temperature and evaporation on runoff changes have also proved that temperature indirectly affects runoff changes by changing evaporation.

#### 5.3.2. Response of Runoff Changes to Climatic Factors on Seasonal Scale

The effect of rainfall on runoff changes in the four seasons has a significant high-energy region from the XWT power spectra (Figure 14). There are higher power spectrum values, stronger influences and wider time domains in spring and autumn. In spring, the related regions are mainly in the main cycle at 4-year time scales from 1998 to 2012 and the subcycle at 1-year scale from 1991 to 1995. In autumn, the relevant regions are mainly in the main cycle at 7-year time scales from 1992 to 2000 and the subcycle at 2-year time scales from 1995 to 2000 and then are invisible.

The effect of evaporation on runoff changes presents some significant high-energy zones in four seasons. Energy is strongest in autumn and weakest in summer. The evaporation effect is significant at 6–8-year time scales from 1993 to 2005. In spring, summer and autumn, the locations of significant high-energy zones on the same scale are basically the same. The effect of evaporation on runoff in spring and summer differs on different time scales. In autumn and winter, the energy distribution of the effect of evaporation on runoff changes is similar to that of rainfall on runoff changes, but the phase relation is opposite. Accordingly, the regulation of rainfall to runoff changes is positive, whereas that of evaporation is negative. This phenomenon may be caused by drought, less rain and strong evaporation, which can directly reduce runoff.

**Figure 14.** XWT between runoff and rainfall, evaporation and temperature on monthly scale. The thick black contour designates the 5% significance level against red noise and the cone of influence (COI) where edge effects might distort the picture is shown as a lighter shade. Arrows denote relative phase difference: The arrows from left to right indicate that the influencing factors and runoff are in the same phase, which implies a positive correlation; the arrows from right to left indicate an inverse phase, which implies a negative correlation; the downward arrows indicate that the influence factor is 90◦ ahead of the runoff change and the upward arrows indicate that the influence factor is 90◦ lagging the runoff.

The effect of temperature on runoff also has significant high-energy regions in the four seasons. The power spectrum value is high, and the influence is strong in spring and winter. The influence of temperature in spring mainly concentrates on the main cycle at 4-year time scales from 2000 to 2014, and it is consistent with the influences of rainfall and evaporation on runoff at more than 6-year time scales. The effect of temperature on runoff changes mainly concentrates on the main cycles at 1-year time scale in summer and at 2-year time scales in autumn from 2007 to 2010. However, the effect of temperature on runoff in autumn has a subcycle at approximately 8-year time scales with high energy from 1995 to 2005, and it is similar to that of rainfall and evaporation at the same scale in the same time domain. The changes in temperature and evaporation in summer are ahead of runoff change, but the lead time of temperature ahead of runoff is greater than that of evaporation in summer. In addition, the effect of temperature on runoff changes in winter has a significant high-energy region at 4-year scale from 1990 to 1995, and that of rainfall and evaporation on runoff is stronger in the same time–frequency domain. Seasonally, the influence of temperature on runoff is similar to that of rainfall on runoff in the energy distribution in spring, which indicates that the increase in temperature results in increased rainfall and thus increases rainfall supply to runoff. In autumn, the influences of temperature, evaporation and rainfall on runoff have a consistent feature in energy distribution, which also shows that temperature has an important impact on evaporation and rainfall and leads to the same effect on runoff.

From the WTC condensation spectra of rainfall, temperature and evaporation with runoff in the four seasons (Figure 15), the highly correlated area of rainfall impacts runoff changes with an increase in years and scales, and it changes from 1-year scale in 1990–2000 to 4-year scale in 1995–2005. The main period of the impact of rainfall on runoff in summer is concentrated on the high-frequency scales, and the bandwidth tends to widen, which indicates that the period tends to be stable. The effect of rainfall on runoff in autumn is mainly manifested in the main period at approximately 8-year time scales from 1992 to 2005, with a wide bandwidth and an extremely stable period. The significant-correlation area of rainfall on runoff in winter is concentrated at 1-year and 7-year time scales from 1993 to 2003, and the influence is relatively weak. The significant-correlation region of evaporation in each season is consistent with the high-energy region of the XWT power spectrum, but it is more significant in autumn over 4-year time scales. Although the bandwidth in autumn is narrowed from 1990 to 2006, it still has a wide periodic bandwidth, which is similar to the impact of rainfall on runoff in the same season, indicating that the runoff change is mainly affected by rainfall and evaporation in autumn. According to phase characteristics, the phase relation between rainfall and runoff is positive, which implies that the influence is always positive, whereas the potential correlation between evaporation and runoff is negative; hence, the influence is negative, and their impact on runoff has a common main period at 8-year time scales. The influence of temperature in spring, summer and winter on runoff is relatively significant, and it is concentrated at the 4-year scale from 2000 to 2010 in spring, at the 1–4-year time scales from 2004 to 2010 in summer and at the 4-year time scales from 1990 to 2000 in winter. This finding indicates that the influence of temperature on runoff has significant differences in scale and time domain in different seasons, that is, it has significant local characteristics rather than global characteristics. As a whole, the main periodic bandwidth of the significant-correlation region in each season tends to widen, and the time-domain range of the correlation region also increases. The frequency-domain structure of the significant-correlation region of the WTC is basically consistent with that of the XWT high-energy region.

**Figure 15.** WTC between runoff and rainfall, evaporation and temperature on monthly scale. The thick black contour designates the 5% significance level against red noise and the cone of influence (COI) where edge effects might distort the picture is shown as a lighter shade. Arrows denote relative phase difference: The arrows from left to right indicate that the influencing factors and runoff are in the same phase, which implies a positive correlation; the arrows from right to left indicate an inverse phase, which implies a negative correlation; the downward arrows indicate that the influence factor is 90◦ ahead of the runoff change and the upward arrows indicate that the influence factor is 90◦ lagging the runoff.

In summary, runoff changes are mainly affected by rainfall and temperature in spring, mainly by direct rainfall recharge. That temperature increases rainfall and evaporation is the reason why its phase relation presents a positive and negative interlacing phenomenon in spring. In summer, runoff is mainly affected by direct rainfall recharge, the effect of evaporation on runoff changes is negative, and the positive effect is mainly reflected at more than 5-year time scales. In autumn, runoff change is affected by a small amount of rainfall supply and runoff loss is caused by evaporation. In winter, runoff is mainly affected by temperature because the rainfall in karst areas cannot form the effective recharge for runoff due to the drought and minimal rain; however, the temperature can indirectly adjust runoff changes by changing evaporation.

#### 5.3.3. Response of Runoff Changes to Climatic Factors on Annual Scale

For XWT P–Q (Figure 16), a strong-influence period occurs in 2005, which indicates that the intensity of interaction between rainfall and runoff changes after the sudden variation in runoff in 2003. This phenomenon also implies that climatic factor is the main driving factor for the recent runoff increase. The high-intensity effects of P–Q present a significantly positive correlation and occur mainly at approximately 6-year time scales in the period from 2000 to 2010. The effect of P–Q passes the test of the red noise standard spectrum at the 0.05 significance level with a phase angle of 60◦, which indicates that runoff is ahead of rainfall by 2 years on the 6-year time scales. The highly significant correlation after 2010 occurs on the 1–2-year time scales with consistent characteristics.

**Figure 16.** The XWT and WTC for annual rainfall (P) and runoff (Q), evaporation (E) and runoff (Q) and temperature (T) and runoff (Q) in the Yinjiang River watershed from 1984 to 2015. The thick black contour designates the 5% significance level against red noise and the cone of influence (COI) where edge effects might distort the picture is shown as a lighter shade. Arrows denote relative phase difference: The arrows from left to right indicate that the influencing factors and runoff are in the same phase, which implies a positive correlation; the arrows from right to left indicate an inverse phase, which implies a negative correlation; the downward arrows indicate that the influence factor is 90◦ ahead of the runoff change and the upward arrows indicate that the influence factor is 90◦ lagging the runoff.

For WTC P–Q, a significant high-correlation region on 4–7-year time scales exists during the entire period, which indicates that runoff is strongly affected by rainfall. From the phase diagram of P–Q, runoff has shown 2–3-year time scales ahead of rainfall in 1990–2000. Thus, the change in runoff is mainly affected by human activities. After the 2000s, the phase angle decreases gradually, indicating that runoff is gradually aggravated by rainfall. According to the results of the XWT and WTC of P–Q, the high-energy region and high-correlation region of P–Q are mainly concentrated around the middle of 2010, and the main cycle is mainly at 6-year time scales.

For XWT E–Q, since 2010, it has a highly significant correlation on the 1–2-year time scales, which demonstrates that evaporation has a significant impact on runoff at this time scale. For WTC E–Q, the E–Q cycle mainly concentrates at 1–2-year time scales during the period from 1990 to 2000 and from 2005 to 2015. The E–Q cycle mainly concentrates on 6–8-year time scales from 1995 to 2005, which indicates that the effect of evaporation on runoff is small. The E–Q phase correlation also shows that the E–Q phase correlation is an inverse phase with a phase angle of 45◦ from 1990 to the end of 2010 at 1–2-year time scales, which implies that evaporation is 1–2 years ahead of runoff. After 2005, evaporation remains in an inverse phase with runoff, with an initial phase angle of approximately 30 (at 1-year time scale), and then decreases and finally increases. The relationship between evaporation and runoff changes from lag to consistency to advance, which indicates that evaporation pays an important role on runoff changes. The results of XWT and WTC show that the phase angle on 7-year time scales is approximately 45◦ from 1995 to 2005, indicating that E is approximately 1.5 years ahead of Q.

For XWT T–Q, there is a high-energy region existing at 1–2-year time scales after 2005, which has passed the test of the standard spectrum of red noise at the 0.05 significant level. Therefore, the influence of temperature on runoff suddenly strengthens around 2005. However, the influence is relatively weak on the 4-year time scale and 6–8-year time scales, and it has not passed the test of red noise standard spectrum at the 0.05 significance level. The WTC results show two low-energy areas with the greatest impact. One is at the time of 3–4-year time scales in the period from 1990 to 2002, in phase with the phase angle between 60◦ and 70◦, showing that temperature is ahead of runoff for more than 2 years. The other is at 1–2 year time scales after 2005, in which T–Q shows a negative correlation in the opposite phase with the initial phase angle of 45◦ and then gradually reduces to 0◦. The above results show that the effect of temperature on runoff gradually changes from lag to consistency, indicating that the effect of temperature on runoff changes is increasingly obvious.

Overall, the interaction of rainfall with runoff changes at 6-year time scales across the entire period. However, the effects of temperature and evaporation on runoff changes are locally significant. The effect of evaporation on runoff changes is similar to that of temperature and has obvious local characteristics, mainly on small cycles.

#### **6. Discussion**

#### *6.1. Multi-Scale Effects of Rainfall on Runoff Changes*

Although the interaction between rainfall and runoff is positively correlated on the whole, the temporal effects are inconsistent in different time domains and scales. The effect of rainfall on runoff is ahead, lagging and consistent in time, ahead in high-frequency and low-frequency scales, lagging at medium-frequency scales (approximately 4 years) and consistent at 1-year scale and significant main periodic scales. For the leading effect, runoff may be mainly affected by early rainfall, which mainly occurs in rainy weather in spring and autumn. The surface and underground areas of karst are filled with soluble rocks with the main type of carbonate, which is vulnerable to erodible water that contains CO2, thereby forming a large number of karst pipelines and fissures [51–53] over a long period of time and two sets of surface and underground hydrological systems [46]. Rainfall requires first to saturate soil water due to low soil moisture in karst areas [54–56]. Surface runoff is difficult to form with small rainfall due to the fragmented surface, steep and rugged

slopes, low runoff coefficients in slope surfaces and small river network density [57–61]. Rainfall is the main factor for runoff formation, and its intensity, duration and areas have great influences on runoff change. When raining heavier, rainwater may hardly infiltrate and leak and thus then increase runoff. If the rainfall intensity is smaller, most of the rainwater infiltrates into the soil and leaks through enormous karst fissure pipelines, which can reduce the runoff. The longer the duration of rainfall and the larger the area of rainfall, the easier the soil moisture will be saturated, and the runoff generated on the slope will inevitably be larger. Runoff monitoring studies on karst slopes show that light rainfall intensities (15~30 mm/h) generate subsurface lateral flow and underground fissure flow, whereas great rainfall intensities generate surface runoff in addition to subsurface and underground flows [62]. However, only a single rainfall of more than 60 mm on a karst slope can produce stable runoff because once the atmosphere rains [63], it immediately runs off into the ground through a broken surface and underground fissure, with distribution ratios of 27.8–78.0%, dominating the total flow yield. Therefore, the loss of rainfall and the formation of runoff in the slope surfaces of karst areas are much more difficult than those in non-karst areas. Only the last rainfall may form slope surface runoff under repeated rainfall because of the recharge of soil moisture first and the loss through fragmented surface leakage. Pre-rainfall mainly supplements soil moisture or leaks down through the broken surface to the pipeline and fissure. For all that, it has been found that all climatic factors exhibit a main cycle at 12-month time scales with runoff changes, which may show that the hydrometeorological processes in karst watersheds represent the same characteristics as those in non-karst watersheds at 1-year time scale (12-month time scale) periodic variations. This may be mainly because the impact of karst characteristics on hydrometeorological processes is mainly manifested on the slope scale, and all flows in the watershed will eventually converge to the outlet of watershed [45,46], which leads to the same annual periodic characteristics as those of non-karst watersheds. If the interval of multiple rainfall is long and the cumulative rainfall is less than 60 mm, it may lead to rainfall changes ahead of runoff on a monthly scale. If the cumulative rainfall is large under the condition of multiple short-term rainfall, the last few small rainfall events will produce obvious runoff on the slope after the saturation of soil water, and then the rainfall before the saturation of soil water will produce a leading effect on runoff. The effect of rainfall after soil water saturation on runoff changes will be synchronous because surface runoff would only occur when both soil and carbonate fissures and fractures are fully saturated with water [36,39]. Most of the rainfall is transported to the groundwater system through carbonate fractures and fractures, while the rainfall that can form surface runoff is very small [53,64].

In addtion, some studies have shown that antecedent rainfall and rainfall intensity are the major factors that control rainfall–runoff and soil erosion processes [65]. Rainfall intensity, slope angle and groundwater porosity [57] are the influencing factors of runoff changes mainly because the runoff mechanism caused by rainfall is different in years with different soil water contents. Different soil moisture contents are present in the early stage and the runoff generated by rainfall is also different in the year of rainfall approaching due to the different soil moisture contents in the early stage. In this case, the annual runoff depth is related to the rainfall year. For some places where recharging soil moisture by rainfall is difficult, the annual runoff depth is even related to the last rainfall year or even the previous years.

The process from rainfall to runoff will undergo seepage storage, slope overflow and channel flow collection. In karst watersheds, each process will be accompanied by underground leakage and the broken surface will affect the time, which greatly lengthens the lag time of runoff change. The influence of the changes in underlying surface conditions on runoff is a gradual process, but the influence of human activities on runoff is a catastrophic process. Therefore, the main reason why rainfall lags behind runoff is that human activities lead to catastrophic changes in runoff, especially land use changes, which destroy the original runoff production and confluence conditions. Such human activities as pumping and storing or introducing water into farmland can also lead to catastrophic changes in

runoff. Therefore, the effect of rainfall on runoff will be delayed. Due to the large amount of runoff that will be produced when a heavy rainfall falls, the runoff series will show a great jump. At this period, obvious runoff will be produced directly on karst slopes because soil moisture is absorbed and saturated in a short time due to the large amount of rainfall, and its response to rainfall is timely with the consistent variation relationship. On the contrary, the runoff series may jump negatively due to the lack of rainfall when the watershed suffers from years of rare drought, but the positional correlation between the two is positive. The effect of rainfall on runoff can be influenced by human activities, such as soil and water conservation, which may play an important role in reducing runoff. However, the role of soil and water conservation will become small or ineffective when encountering heavy rain or rainstorm. The effect of rainfall on runoff changes will change the relationship between rainfall and runoff because of different patterns, intensity or frequency when raining. Temporary water intake by human activities can also alter runoff, thereby resulting in different time effects of advance, synchronization and lag.

### *6.2. Multi-Scale Effects of Evaporation on Runoff Changes*

In the process of rainfall, evaporation exerts a minimal effect on runoff but has a great impact on the water storage capacity of the basin before rainfall. The greater the evaporation intensity, the smaller the soil water content before the rain, which increases the infiltration loss of rainfall and reduces the small-diameter flow. This study has supported the previous conclusion in annual scale that the effect of evaporation on runoff change was significantly enhanced, showing a great contribution of 10–90% [32], but there were some new discoveries during the year. The effect of evaporation on runoff was only in the high-frequency scale in summer and the 6-year scale in winter. In other seasons or scales, most hysteresis effects with a few synchronous relationships in the time domain have shown at different time scales, which indicates that runoff changes are affected by evaporation. The evaporation is larger in summer; hence, short-term evaporation has a significant impact on runoff changes, which results in the changes in runoff lagging behind evaporation. The effect of runoff ahead of evaporation has been virtually masked by rainfall and human activities. The runoff changes are greatly influenced by abrupt rainfall and human activities, whilst evaporation shows a continuous stable process. Rainfall burst or human activities will contribute to the changes in the underlying surface of the watershed, which directly alters the evaporation conditions and volume that will cause the time dislocation in different time domains. The strong disturbance of human activities on runoff will directly lead to relatively stable and persistent evaporation lagging behind the change in runoff.

The essence of changing runoff by evaporation is to reduce the recharge of runoff and increase the evaporation of the river surface. In addition, evaporation shows a high impact on runoff change also because of the influence of the subtropical monsoon climate, abundant light and heat resources in Southwest China. The influence of evaporation on runoff varies obviously in different periods, which is mainly affected by the light, temperature, heat, climate and water content of underlying surface. However, runoff changes are affected not only by evaporation but also by rainfall and human activities, which makes it impossible for the evolution of runoff and evaporation to be completely consistent.

#### *6.3. Multi-Scale Effects of Temperature on Runoff Changes*

The influence of temperature on runoff is consistent with that of evaporation in both time and frequency domains and has the same multi-time-scale characteristics and time– frequency relationship. However, a negative correlation exists between temperature and runoff because an increase in temperature leads to the intensification of evaporation on slopes and rivers of the watershed and decreases air humidity, thus changing the runoff.

On monthly and annual scales, as well as in summer, the effect of temperature on runoff is mostly ahead of schedule, whereas it is mainly lagging behind in spring, autumn and winter. Thus, the regulation of temperature on runoff is mainly reflected on the season scale. In summer, the change in runoff is mainly caused by changing evaporation and increasing rainfall to recharge soil moisture, and hence its impact on runoff shows a longer lead time than that of rainfall and evaporation. In other seasons, human activities change runoff intensely because of the relatively minimal rainfall, which leads to the relative lag of temperature change. On annual scale, the temperature regulation effect before 2000 is relatively stable, which mainly changes the runoff by changing the roles of evaporation and rainfall, resulting in a leading effect. Overall, the inter-annual temperature regulation is gradually lost, and the temperature regulation during the year is relatively prominent, but this regulation remains affected by human activities.

#### **7. Conclusions**

In this study, the multi-scale influences of climate factors on runoff changes in the Yinjiang River watershed are identified by using wavelet analysis, and the evolution relationship of time and frequency between runoff changes and climatic factors is further revealed at different time scales. The main conclusions are as follows:

(1) All climatic factors exhibit a main cycle at 12-month time scales with runoff changes, but the main periodic bandwidth of rainfall on runoff changes is much wider than that of temperature and evaporation, indicating that rainfall is the main factor affecting runoff changes.

(2) In other cycles, the impact of rainfall on runoff changes is the interlacing phenomena with positive and negative, but the impact of temperature and evaporation on runoff change is mainly negative.

(3) The response of runoff to rainfall is timely in the high-energy region and the lowenergy significant-correlation region and shows a positive correlation with a smaller phase angle, but it is slightly lagged at 16-month time scales, in which the runoff changes lag behind temperature and evaporation for 1–2 months.

(4) It has been found that there is a strong effect of rainfall over runoff but a lesser effect of temperature and evaporation over runoff.

(5) The interaction of rainfall with runoff changes at 6-year time scales across the entire period. The effect of evaporation on runoff changes is similar to that of temperature and exhibited obvious local characteristics, mainly at small cycles.

The study has revealed the evolution process of river runoff in typical karst basins and the interaction mechanism between river runoff and climatic factors on multiple time scales, providing theoretical inspiration for fully solving the regional water shortage and engineering water shortage problems in the karst areas of Guizhou Province.

**Author Contributions:** Conceptualization, methodology and validation, X.B. and S.W.; Software and data curation, L.W. and C.R.; Formal analysis, L.W. and C.L.; Investigation, L.W. and F.C.; Writing original draft preparation, L.W., C.R. and S.Z. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research work was supported jointly by the Western Light Cross-team Program of Chinese Academy of Sciences (No. xbzg-zdsys-202101), National Natural Science Foundation of China (No. 42077455 & No.42167032), Strategic Priority Research Program of the Chinese Academy of Sciences (No. XDB40000000 & No. XDA23060100), Guizhou Provincial Science and Technology Projects (No. 2022-198), High-level innovative talents in Guizhou Province (No. GCC[2022]015-1 & No. 2016-5648), Guizhou Provincial 2020 Science and Technology Subsidies (No. GZ2020SIG), Opening Fund of the State Key Laboratory of Environmental Geochemistry (No. SKLEG2021072001 & No. SKLEG2022206 & No. SKLEG2022208) and Doctoral Research Startup Fund Project of Tongren University (No. trxyDH2103).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data analyzed in this study are subject to the following licenses/ restrictions: The dataset can only be accessed from China Meteorological Data Sharing Service System, Karst Scientific Data Center and Guizhou Provincial Hydrology and Water Resources Bureau. Requests to access these datasets should be directed to jgywlh@gztrc.edu.cn.

**Acknowledgments:** We would like to thank all the authors and reviewers for their great guidance and help in writing this manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Xuan Li 1, Li Rong 1,\*, Mengmeng Zhang 1, Wensong Yang 1, Zhen Zeng 1, Chengjun Yuan <sup>2</sup> and Qi Wang 1,2**

<sup>1</sup> School of Geography and Environmental Sciences, Guizhou Normal University, Guiyang 550000, China

<sup>2</sup> School of Karst Science, Guizhou Normal University, Guiyang 550001, China

**\*** Correspondence: ronglit@gznu.edu.cn

**Abstract:** In recent decades, vegetation coverage and land use/land cover (LULC) have constantly changed, especially in southwest China. Therefore, it is necessary to conduct in-depth research into the temporal–spatial variation patterns of vegetation greening, LULC, and gross primary productivity (GPP). Here, we used remote sensing to analyze the spatial and temporal variation in the normalized difference vegetation index (NDVI) and GPP in the growing season under different LULCs in southwest China. Results showed: (1) From 2000–2019, the forest area in southwest China had increased by 2.1%, while the area of cropland and grassland had decreased by 3.2% and 5.5%, respectively. Furthermore, there are significant differences in spatial variation patterns. (2) NDVI and GPP in the growing season showed a general increasing trend (*p* < 0.01); vegetation coverage is dominated by high coverage to highest coverage and medium coverage to high coverage transfer. (3) Under different LULCs, the migration directions of NDVI and GPP were different. The center of gravity migration of highest and medium coverage shifted to the southeast by 1.69◦ and to the northwest by 1.81◦, respectively. The results showed the ecosystem evolution and will help to guide the maintenance measure of ecosystem balance and sustainable development.

**Keywords:** southwest China; normalized difference vegetation index (NDVI); gross primary productivity (GPP); land use/land cover (LULC); center of gravity shift model

#### **1. Introduction**

The ecological environment of karst landforms in southwest China is fragile and has been significantly affected by climate and human activities in recent decades [1–3]. Changes in vegetation, LULC, and GPP affect biogeochemical cycles, and social effects in this region impact the area range of influence [4–7]. However, the spatial and temporal characteristics of LULC, vegetation, and carbon storage are not clear. This has a significant impact on ecological evolution and regional social development [8,9]. Therefore, there is a need to clarify the temporal change characteristics of LULC, vegetation, and carbon storage in southwest China.

Southwest China has a large number of karst ecosystems, which are hypersensitive and fragile. First, this area is one of the largest exposed areas of carbonate rock salts in the world [10], and in these environments, the soil formation rate is low, and the permeability is high due to the presence of interstitial fractures. Furthermore, it has unique and fragile geomorphological and hydrogeological features [11]. In recent decades, long-term and severe climate change and human activities have brought enormous pressure to the ecosystem in this area [12–14]. Rocky desertification has become one of the most serious environmental problems in karst areas [15,16]. Terrestrial vegetation types and compositions have changed due to climatic conditions, carbon dioxide fertilization effects, and LULC [17,18]. Second, under the background of population pressure and urbanization, the intensity of human activities has increased rapidly, and the land cover has undergone drastic changes [19]. Third, since the end of the 20th century, China has implemented a large number of ecological

**Citation:** Li, X.; Rong, L.; Zhang, M.; Yang, W.; Zeng, Z.; Yuan, C.; Wang, Q. Temporal Changes in Land Use, Vegetation, and Productivity in Southwest China. *Land* **2022**, *11*, 1331. https://doi.org/10.3390/land11081331

Academic Editor: Xiaoyong Bai

Received: 19 July 2022 Accepted: 15 August 2022 Published: 17 August 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

engineering constructions, which have achieved an increase in vegetation coverage and carbon storage through ecological restoration and improved ecosystem services [20–22]. In addition, the southwest region is an important ecological barrier and ecologically fragile, with extensive potential for various ecosystem services, such as soil and water conservation, climate regulation, and carbon balance [23–25], providing a huge contribution to social development, ecosystem balance and carbon sequestration [24,26,27].

At present, the spatial and temporal changes in LULC, vegetation cover, and gross primary productivity (GPP) in southwest China are not clear. Vegetation is an important factor affecting the ecological balance and is usually considered as a direct and obvious indicator to analyze the impact of natural seasonal changes and human activities on the ecological environment [28,29]. Gross primary productivity (GPP) is an important indicator reflecting vegetation status, ecosystem structure, and function [30] and plays a key role in carbon cycling in terrestrial ecosystems [31], and is an important factor in measuring the regional ecological value [32,33]. Therefore, clarifying temporal and spatial evolution processes is of great significance for understanding the value and sustainable development of ecosystems. Studies have found that China's vegetation has shown an overall greening trend in the past 30 years [24]. However, due to the vast heterogeneity of climate, topography, and human activities in the southwest, the spatial and temporal distributions of LULC, vegetation dynamics, and gross primary productivity (GPP) are significantly different [2,14,34]. Since 2000, the LULC change in southwest China has been mainly manifested in the expansion of forest land and the reduction in cropland [4,35]. The study found that NDVI increased significantly in low- to mid-altitude areas < 3400 m due to improved afforestation and agricultural productivity [36]. In the afforestation and grassland restoration areas, the direct contribution of forest land to the annual growth rate of GPP is 24.64% [37]. In addition, according to long-term remote sensing vegetation data, it is found that short-term extreme climate events respond differently to different land-use types, resulting in differences in regional ecological effects [38,39]. Therefore, it is of great significance to understand the temporal and spatial pattern characteristics and change processes of different LULC types, vegetation dynamics, and gross primary productivity in the region for correctly understanding the temporal dynamic changes and spatial changes in regional vegetation dynamics and gross primary productivity.

The changes in vegetation and productivity center can reflect the evolution of ecosystems influenced by human activities and climate change. Human activities affect vegetation and productivity changes, such as ecological engineering, which increases vegetation growth and carbon storage in southwest China [40,41], and positively contributes to vegetation productivity [42]. However, the expansion of arable land and the surge in population has also led to the degradation of vegetation [15]. Deforestation reduced the GPP and leaf area index in China between 1982 and 2011, and their centers of gravity shifted [43]. The spatial and temporal changes in vegetation cover and productivity in different regions have obvious uncertainties [6,44,45]. Natural evolution is also an important factor leading to the migration of its center of gravity; for example, the northward shift of the climatic zone makes the ecological center of gravity move northward [46,47]. In summary, combined with different LULC types, studying the temporal and spatial variation patterns of different levels of vegetation cover (NDVI) and its gross primary productivity (GPP) in southwest China can deepen the understanding of vegetation and productivity changes in southwest China. It has very important ecological value and practical significance for the balance and sustainable development of the ecosystem.

The purpose of this study is to clarify the temporal and spatial dynamic of LULC, vegetation cover, and GPP in southwest China and the migration pattern of the center of gravity. Combined with MODIS remote sensing, we analyzed the temporal and spatial changes of vegetation cover (NDVI) and gross primary productivity (GPP) under different LULC types in southwest China. Therefore, our aims in the study are: (1) to clarify the migratory direction of LULC in southwest China and the spatial and temporal change patterns of NDVI and GPP in the growing season; (2) to explore the change characteristics of

GPP under different land use types and different vegetation coverage levels; (3) to analyze the migration law of vegetation cover and GPP center of gravity.
