1 Introduction

How to mitigate or avoid the hazards resulting from sand-dust weather has become a hot issue discussed by many scholars(e.g.，Han et al.，2006; Li et al.，2007; Mason et al.，2008; Chase，2009). The pre-cipitation，soil water，and underground water all in-fluence the land surface moisture and vegetation growth，and contribute to dust control and sand protection; they also have a certain relationship with the development and occurrence of sand-dust weather(Lu et al.，2003; Wang，2010; Zhao et al.，2012). The re-search performed by Wang et al.(2003)showed that the annual mean precipitation in high sandstorm inci- dent areas is 198.47 mm，which is lower than the defined value(200 mm)that distinguishes between arid and semi-arid areas. Zhang and Ren(2003)showed that the number of sand-dust storm occurrences is reduced with the increase of precipitation and relative humidity of the soil. Tang and Zhang(2001)showed that changes in the ground surface's hydrological regime and the development of artificial water resources will lead to the continuous decrease of underground water levels in the lower reaches and in the peripheral zones of the middle reaches，and there will be an increase of sandstorm frequency. Zhang et al.(2010)showed that the spatio-temporal changes of sand-dust weather largely depend on the space-time processing relationship between the strong wind and precipitation; in particular，the combined action of strong wind and precipitation has a highly significant correlation with the space(R² = 0.960)and monthly changes(R² = 0.995)of sand-dust weather. Recently，some innovative and frontier achievements have focused on how to quantify the mechanism and influence degree of precipitation，soil water，and underground water on sand-dust weather. However，many of the results still require verification through further testing and investigation. In 2006，the State Forestry Bureau established a monitoring station for desertification in the Heihe middle reaches，for the purposes of long-term local and multi-component monitoring. The Hongshawo test station has obtained nine years of continuous and complete data on weather，hydrology，soil，and vegetation relating to the occurrence and development of land desertification and sandstorms.

2 Data sources and methods 2.1 Overview of the research area

The northern desert areas of the Heihe middle reaches include three major deserts: the Badain Jaran Desert，Tengger Desert，and Taklimakan Desert，with the wind sand line being as long as 1，600 km. The scope of this area covers 37°28'N–39°57'N and 97°20'E–102°12'E，with an average altitude of 1，200~1，700 m a.s.l.. This area has a temperate arid continental desert climate，with an annual average temperature of 7.4~8.5 ℃ and annual average precipitation of 108.3~150.0 mm. The precipitation is mainly concentrated in June to September，accounting for 70%~80% of the annual total precipitation，and the annual total evaporation is 1，340.7~2，388.0 mm. This area has a dense population，and human activities alter the nature of land utilization，influence the physical and chemical weathering processes of the surface materials，and have a direct impact on the sand sources. The area has sparse natural vegetation coverage，mainly including Nitraria tangutorum，Rea-umuria songarica，Salsoal arbuscula，Nitraria sphaerocarpa，and other desert plants. The zonal soil contexts include gray desert soil，sierozem，and gray-brown desert soil，while the azonal soil contexts include aeolian sandy soil，meadow soil，boggy soil，salinized soil，and anthropogenic-alluvial soil. The overall characteristics of this area are depleted soil，lack of organic matter，coarse soil texture，and high salt content. The soil in this area is of the desert soil type and has complicated ecological and plain desert vegetation characteristics. The land in this area has many geographical irregularities and a diverse temporal- spatial distribution of vegetation，and possesses both ancient and modern characteristics. The geomorphologic landscape type includes flowing，semi-flowing，fixed，and semi-fixed sand dunes，as well as inter-dunes depressions.

2.2 Data sources and monitoring methods

The purpose of the Hongshawo test station is the comprehensive prevention and control of desertification. The station is located 15 km north of the Zhangye city center，Gansu Province，and has an altitude of 1，450 m a.s.l.. Table 1 shows the location of each of its test points. The precipitation data are from the meteorological station(No. Y₁)，where one data monitoring sequence is performed each day and night，once per hour during 00:00 to 23:00. The data of underground water depth is from wells No. 1 and No. 2(No. Y₃ and Y₄)，and one assessment is made on the 15th day of each month to take the average value of the two wells. The data of soil mass water content are from soil samples taken from the digging profile in three representative sample sites randomly selected from the surrounding area(No. Y₅ to Y₂₁ in the desert area)，and one assessment is made on the 15th day of each month. The data of biomass and vegetative cover degree are also from the sample sites No. Y₅ to Y₂₁ in the desert area; the sample site size is 2m×2m(as shown in Table 1)，and one assessment is made on the 15th day of each month.

The data for the sand-dust weather are from the dust and weather monitoring tower(No. Y₂). The equipment and apparatuses include four sets of vertical- fall dust samplers used in the case of sandstorm，one KC-1000 type large-flow sandstorm sampler，one SC-1 type sandstorm sampler，one KC-16A single- channel particulate sampler，one KC-120H intelligent mid-flow TSP sampler，and one LS-2009 portable meteorological parameter comprehensive tester. The monitoring data mainly include the sand-dust weather conditions(type，visibility，and start and ending time of the sand-dust weather)，atmospheric dust concentration，and dust fall quantity(Pm^2.5，PM₁₀，total dust and sand concentration，and dust fall quan- tity)，as well as sandstorm disaster effects(e.g.，agriculture，forestry，traffic，communication，electricity power，personal injury，economic loss). The research described in this paper only extracts the data for the occurrence frequency and duration of precipitation，soil water content，underground water depth，amount of precipitation，and dust and sand occurrence frequency from 2006 to 2014.

The data of soil mass water content were obtained by regularly taking samples from the fixed sample sites in the test station area by means of a large cutting ring，with a sampling depth of 80 cm. This depth was divided into four layers，0~20，20~40，40~60 and 60~80 cm，and each layer was sampled three times. The mass water content of the soil samples was determined in the laboratory using the oven-drying method(105 ℃). Multi-profile repetitive measurements were used in each site sample. Beginning from the growth of plants in March，the sampling was performed on the 15th day of each month(the observation time was delayed by 1 h if there was precipitation at the normal measuring time，and was performed the next day if there was precipitation throughout the whole day)，until the plants stopped growing in November.

2.3 Analysis methods 2.3.1  Eigenvalue parameter algorithm The soil mass water content，underground water burial depth，precipitation，frequency of sand-dust weather，duration of sand-dust weather，visibility in sand-dust weather，annual and inter-annual variation of the minimum visibility in sand-dust weather，and other eigenvalue parameters were calculated by using standard statistical methods including average value μ，standard deviation σ，and variation coefficient C_v. 2.3.2  Correlation analysis method The monthly and inter-annual scales of precipitation，soil mass water content，underground water burial depth，occurrence frequency of sand-dust weather，duration of sand-dust weather were analyzed by the correlation coefficient r. 2.3.3 Stepwise linear multi-variation regression analysis We established the sand-dust weather regression model by the fitting calculation and variance analysis on the sand-dust weather and environmental moisture coefficient，as well as the R test，F test，and t test of the regression model's optimized degree. 3 Results and analysis 3.1 Analysis of the inter-annual variation characteristics of the sand-dust weather and environmental moisture

Table 2 shows the μ，σ，and C_v values for the precipitation，soil mass water content，groundwater depth，frequency of sand-dust weather，duration，visibility，and average value of the minimum visibility. The precipitation usually fluctuated within 104.33 to 174.89 mm，and the average precipitation was 139.61 mm. Because the soil distribution in the various layers had no obvious change in the vertical aspect，the average values of mass water content，standard difference，variation coefficient，and fluctuation scope of each soil layer could be averaged again to obtain 3.97%，1.23%，0.33，and 2.69%~5.52%，respectively. During the period of 2006–2014 there were a total of 86 sand-dust weather events，ranging from 4.77 times/a to 14.34 times/a; the average occurrence frequency was 9.56 times/a. The duration of sand-dust weather usually fluctuated from 8.31 h/a to 181.79 h/a，and the average duration was 95.05 h. Sand-dust weather had the largest variation coefficient，with an average value of 0.79，while the precipitation and soil moisture had the smaller variation coefficient，with an average value of 0.31. The groundwater depth had the smallest variation coefficient of 0.07(Table 2).

It can be seen from Figure 1 that the frequency and duration of sand-dust weather had an obvious trend of fluctuant increase，while the precipitation，soil water，and underground water level(the underground water burial depth as opposed to the water level)had smaller magnitudes of variation. The response relationship of the sand-dust weather to the change of environmental moisture was closely related with the time scale. On the scale of inter-annual variation，the sand-dust weather still had a large magnitude of variation when the variation of environmental moisture became stable，which signified that the sand-dust weather was less correlated with the environmental moisture on the inter-annual scale.

3.2 Analysis of the annual variation characteristics of the sand-dust weather，and the environmental moisture

As shown in Figure 2，the variation of precipitation was basically consistent with the underground water depth in March–November，especially during the vegetation growing season of each year. The precipitation in March was 4.14 mm and the groundwater depth was 83.56 cm. Then the figures increased gradually: the precipitation in July was 36.10 mm and the underground water depth was 578.11 cm，both of which was the maximum yearly values. Thereafter they decreased gradually: in October the underground water depth was 191.89 cm. In November the underground water depth increased to 300.20 cm，while the precipitation continued to decrease to 0.70 mm in that month.

The variation of water content in each soil layer was basically consistent. The average mass water content of each soil layer in March was 3.48%，which then increased gradually to the maximum value of 4.69% in November. It could be seen from Table 3 that the vertical change of soil water content was relatively small. The variations of frequency and duration of sand-dust weather were basically consistent. The occurrence of sand-dust weather gradually increased from November onwards(from an average frequency of one time and average duration of 14.5 h). Then these two values increased gradually and reached their maximum values in March−April，with a monthly average frequency of three times and average duration of 38.82 h. From May onwards the average frequency was two times and the duration was 12.07 h. The sand-dust weather gradually decreased to reach the minimum value in June−August，with a monthly average frequency of one time and duration of 3.15 h. There was no sand-dust weather in September–October.

In Table 3 the variation coefficient of the mass water content in each soil layer is 0.21. Thus，in combination with Figure 2，the annual magnitude of variation(variation coefficients)in descending order is as follows: duration of sand-dust weather ＞ visibility of sand-dust weather > frequency of sand-dust weather ＞ precipitation ＞ minimum visibility of sand-dust weather > underground water depth ＞ soil mass water content. Overall，the sand-dust weather had the largest magnitude of variation，followed by the precipitation and underground water，while the variation of soil water content was most stable.

3.3 Correlation analysis of the sand-dust weather and environmental moisture

After performing correlation analysis on the monthly scale，the critical value F_0.05(1，9)at the level of a = 0.05(p＜0.05)was 5.117，and we substituted this into the critical value of the correlation coefficient ${{r}_{0.05}}=\sqrt{{{F}_{0.05}}\left( 1,9 \right)/\left( {{F}_{0.05}}\left( 1,9 \right)+9 \right)}$ to obtain an r_0.05 value of 0.60.(A correlation is significant when the correlation coefficient is larger than the critical value; otherwise it would be non-significant. As shown in Table 4，the frequency of sand-dust weather had a weak correlation with the soil mass water content，and an even weaker correlation with the precipitation. However，the duration of sand-dust weather had a significant correlation with the underground water depth and precipitation.

3.4 Regression model analysis 3.4.1  Analysis of the fitting and variance of regression model After verification through R fitting inspection，F variance inspection，and t regression coefficient inspection，we established a stepwise regression model for the vegetation cover degree and biomass with the 0–20 cm soil mass water content(as shown in Tables 5 and 6). Table 5 shows the F inspection value and significance level for the regression fitting and variance analysis of the duration of sand-dust weather with the precipitation and underground water depth. The general judgment criteria for the relationship coefficient are that |R|＞0.95 signifies a significant correlation; 0.5≤|R|＜0.8 signifies a moderate correlation; 0.3≤|R|＜0.5 signifies a low correlation; and |R|＜0.3 signifies an extremely weak relationship or non-correlation. The duration of sand-dust weather had a moderate positive correlation with the precipitation and underground water depth，and the model passed the R fitting inspection and achieved an ideal fitting effect. Because the multiple determination coefficient signifies that the independent variable can explain the weight of the dependent variable variation，the precipitation and underground water depth could explain 23.9% of the variation of sand-dust weather duration. The remaining parts must be explained by other factors，such as the micro-topographic changes in the land. The standard error indicates the average difference between the predicted value and the actual measured value; a smaller standard error signifies a more ideal fitting degree. The F value in Table 5 is the variance analysis and test value，and is the ratio between the regression mean square error and residual mean square error. As we expected，the larger regression mean square error was more ideal and the smaller mean square error was more ideal，while the larger F value signified that the model prediction result is more ideal. In accordance with Table 5，the significance level of F was ＜0.5; thus，the critical value F_a of F_0.5(2，6)at the a = 0.5(p＜0.5)level was 0.780，the F test value was much larger than the critical value Fa，and the model passed the variance inspection of F. 3.4.2  Regression coefficient analysis A regression coefficient analysis is mainly performed to inspect the significance level of the correlation coefficient. The p-value in Table 6 indicates the significance level of the regression coefficient; the smaller the p value is，the more significant the change of the regression coefficient will be. In accordance with the stepwise regression analysis method，the independent variables of which the p-value is smaller than 0.5 in Table 6 only include the constant term and underground water depth; thus，the critical value of t_0.5(8)at the level of a = 0.5 is 0.706. In Table 6 the absolute value of its corresponding t inspection value is larger than the critical value，which demonstrates that the constant and soil mass water content are extremely significant at the level of a = 0.5(p＜0.5)，and the confidence level is 95%. From the above R fitting inspection，F variance inspection，and t regression coefficient inspection，the regression model for the duration of sand-dust weather is:

$T=31.83-0.04D\left( {{R}^{2}}=0.429,p＜0.5 \right)#

(1)

where T is the duration(h)of sand-dust weather，and D is the underground water depth(cm). 4 Discussion

The Heihe Basin area usually has the problems of fair and reasonable water resource allocations among the upper，middle，and lower reaches，as well as to the departments. For example，the State Development Planning Commission and the Ministry of Water Resources successively approved the water distribution program in the Heihe Basin in 1992 and 1997. Hydroelectric power stations are usually constructed in the upper reaches of the basin to supply water to the upper and lower reaches by means of artificial regulation. In the regulation and management of these water resources，the precipitation，soil water content，and annual and inter-annual variation laws of underground water can provide scientific support and reference data. In addition，the sand-dust weather that is caused by meteorological factors is also closely related to the land desertification and vegetation growth in the basin，while the growth of vegetation is in turn closely related to the environmental moisture(Niu et al.，2015). Therefore，the changes in sand-dust weather(frequency，duration，and intensity)are important indicators for the prevention and control of desertification.

Zhao et al.(2009)found that the underground water level in 2006 was obviously decreased compared with the level before the water diversion on the Heihe middle and lower reaches in 2000，and the average decrease was 0.96 m. This research showed that the water level in the middle and lower reaches still tended to decrease after 2006，and this would be very evident in a short period. Liu et al.(2009)researched the precipitation characteristics in the desert area of the Heihe Basin as well as the response of its soil water content to the precipitation pulse. They found that the soil water content exhibits an obvious response phenomenon to the precipitation pulse，and that precipitation of less than 5 mm was of great significance to the survival of shallow-rooted plants such as annual herbs. At the same time，precipitation of greater than 5 mm can effectively make up the water content in the root layer soil so as to allow the desert plants to survive and grow under the dry conditions. This research showed that precipitation has a moderately positive correlation with the soil water content，which is consistent with our research conclusions.

Sand-dust weather is divided into sandstorms，blowing dust，and floating dust. In terms of the regional characteristics of the sand-dust weather in China，the areas with an average of more than 10 days of sandstorms are mainly distributed in the southern Xinjiang Basin，the Hexi area，and the northeast Qinghai-Tibet Plateau(Zhang and Yang，2006). The spatial distribution of sandstorms in China is basically consistent with the distribution of northern desert and desertified land. Although the artificial forest network in the oasis area of Heihe has been a great development，which can effectively prevent the intrusion of desert in local areas and successfully revert some desertified land to artificial oasis areas，the desertification in the middle reaches still continues to develop. Thus，the land desertification and the deterioration of oasis land provide abundant sand and dust sources for the sand-dust weather. Therefore，the most fundamental and necessary work for the prevention and control of sandstorms is to establish a complete forest protection system and protect the natural Populus euphratica forests on the river banks of the lower reaches; the core task is the scientific management of water resources.

5 Conclusions

1)During 2006–2014 the precipitation in the middle reaches of the Heihe River generally fluctuated within the range of 104.33~174.89 mm，and the average amount was 139.61 mm. The mass water content of each of four studied soil layers generally fluctuated within the range of 2.69%~5.52%，and the average value was 3.97%. There were a total of 86 sand-dust weather occurrences，which usually fluctuated within the range of 4.77~14.34 times/a; the average frequency was 9.56 times/a. The duration of sand-dust weather usually fluctuated within the range of 8.31~181.79 h/a，and the average duration was 95.05 h. The frequency and duration of sand-dust weather had an obvious tendency of fluctuant increase，while the magnitudes of variation of precipitation，soil water content，and underground water level were smaller.

2)During the March–November period of the vegetation growing season，the changes of water content in each studied soil layer were basically consistent，as were the changes of frequency and duration of sand-dust weather，and the changes of precipitation and underground water depth. The sand-dust weather had the largest magnitude of variation，followed by the precipitation and underground water，while the variation of soil water content was most stable.

3)We established a regression model for the duration of sand-dust weather，T = 31.83−0.04D(R² = 0.429，p＜0.5)，and this model can be used to forecast 42.9% of the variation of duration of sand-dust weather.

Acknowledgments:

This work was supported by the Science and Technology Innovation Service Platform of Qilian mountains in Gansu Province(No. 144JTCG254)，the Innovation Groups of Basic Research of Gansu Province(No. 145RJIG337)，and the National Natural Science Foundation of China(No. 41461004).