3.4. 1. 1 Distribution characteristics of karst groundwater level observation network in Guilin city
Fig. 3. 1 1 Schematic diagram of distribution of karst groundwater level observation network in Guilin.
(According to the water level data in September 1990)
By 1990, there are 60 water level observation wells in different groundwater types, environmental geological sub-regions and areas with prominent environmental geological problems, including 35 drilled wells, 4 natural water intake points 14 and large wells 1 1 mouth. Pore water 1 1, karst groundwater 49; The distribution of karst groundwater level observation network in Guilin is shown in Figure 3. 1 1. As can be seen from the distribution map, observation wells are mainly distributed in the city center, and there are few or no observation wells around the study area.
3.4. 1.2 Main problems existing in Guilin karst groundwater level observation network
Guilin karst groundwater level observation network was established in 198 1. At that time, many observation wells originated from geological exploration holes or construction drilling holes for other purposes, and there were some unreasonable and imperfect places. The main problems are as follows:
1) The information and data provided by the observation network are redundant. According to the analysis of water level data and distribution map, the water level and its dynamic changes of some adjacent observation wells are basically the same. Such as observation wells G Ⅲ 46 and G Ⅲ 47 (Figure 3. 12).
Fig. 3. Dynamic curve of groundwater level in121990 observation well.
2) The distribution of monitoring points is unreasonable. There are too many observation well in the basin and too few observation wells in some runoff areas. For example, in the north of Chaoyang and the southwest of the study area, the groundwater runoff area, the water level changes greatly, the hydraulic gradient changes greatly and there are too few observation wells.
3) Water level monitoring and water quality monitoring are not synchronized in groundwater monitoring. For a long time, the heavy water level is lighter than the water quality, which cannot meet the requirements of comprehensive evaluation of groundwater resources quality.
3.4.2 Research on Optimization of Karst Groundwater Level Observation Network in Guilin City
Research objects and objectives
Guilin karst groundwater is widely distributed and has good water quality, which is the main mining layer at present. However, in areas with strong karst development, over-exploitation of karst groundwater often leads to geological disasters such as land subsidence and collapse. Therefore, it is necessary to establish a reasonable groundwater observation network, observe the karst groundwater level in real time, realize the rational exploitation of groundwater, prevent the diffusion of funnel area and protect the quality of groundwater. Taking karst groundwater level as the research object, the observation network of karst groundwater level in Guilin city is optimized.
The standard deviation of groundwater level estimation error is selected as the replacement objective function, and the average standard deviation of the estimation error of the whole observation network is used as the standard to evaluate the overall accuracy of the observation network. The optimization goal is to make the average standard deviation of the optimized estimation error not much different from the original observation network under the condition of saving part of the cost.
Layout principle of groundwater observation network in 3.4.2.2
The general layout principles of observation wells are as follows: for a large monitoring area, the monitoring network should be mainly arranged along the groundwater flow direction, supplemented by the vertical flow direction of groundwater; For the monitoring area with small area, control monitoring points are arranged according to the recharge, runoff and discharge conditions of groundwater.
1) The national regional groundwater monitoring points should be arranged according to the geological environment background and hydrogeological conditions on the basis of the division of hydrogeological units and aquifer sequences. Mainly arranged in:
A. Karst water has water supply significance, and karst collapse has occurred or may occur.
B areas where regional environmental geological problems have been or will be formed.
2) The groundwater monitoring points in provincial cities should be arranged on the basis of the groundwater monitoring network in national cities. The following aspects should be considered when arranging:
Monitoring points should be set up in the recharge area, runoff area and discharge area of urban water supply sources, near pollution sources and water source protection areas.
B two monitoring lines should be laid in the water source area, parallel to and perpendicular to the direction of groundwater flow, to monitor the formation and development trend of the groundwater level drop funnel.
C. When finding out the mutual influence between water sources or the influence of drainage from nearby mining areas on water sources, monitoring points should be set up in the connecting zone between the two mining areas.
D. In order to establish the urban groundwater equilibrium calculation model or groundwater management model, control monitoring points can be arranged in the boundary and calculation area.
Study on Optimal Design of Karst Groundwater Level Observation Network in Guilin City, 3.4.2.3
Semi-quantitative analysis of karst hydrogeology (1)
Considering the inaccuracy of increasing or decreasing observation wells by Kriging method, a semi-quantitative analysis method of karst hydrogeology is put forward from the perspective of main factors (karst groundwater protection conditions, karst groundwater abundance, karst development characteristics, karst groundwater exploitation conditions, etc.). ) affect the optimal design and layout of the observation network. According to the semi-quantitative zoning map and the contour map of the error variance of water level estimation, the comprehensive map is synthesized, and the observation wells are increased or decreased through the comprehensive map, which can not only arrange the observation wells more accurately, but also improve the optimization efficiency. The method is introduced as follows.
1) assignment. According to the influence of various factors on the layout of karst groundwater level observation network in Guilin, it is divided into four grades: large, medium, small, large, medium and small. According to the actual situation in Guilin, the main influencing factors selected and their assignment criteria are shown in Table 3.3.
Table 3.3 Assignment Criteria of Influencing Factors
2) Determination of weight. Weight reflects the relative importance of each influencing factor, which is not only the subjective evaluation of decision makers, but also the objective reflection of the essential physical attributes of indicators, and the result of subjective and objective comprehensive measurement. The weight mainly depends on two aspects: first, the role of the index itself in decision-making and the reliability of the index value; Second, the degree to which decision makers attach importance to this indicator. At present, there are many subjective arbitrariness in determining the index weight, which seriously affects the objectivity of the evaluation results. The optimization model is established, and the subjective and objective weights are obtained by linear weighting method. The expression is as follows
Study on the particularity of groundwater and environment in karst area
Solve the above model to get the comprehensive integration weight coefficient:
Study on the particularity of groundwater and environment in karst area
Among them: the weight of each factor in expert evaluation; Hk is the importance of expert evaluation results, hk= 1, and Q is the number of experts; α and β respectively represent the relative importance of subjective and objective weighting methods, α+β =1; M is the number of evaluation schemes; αij is the j th factor value of the i th evaluation scheme; It is the optimal value of the j-th factor value in each evaluation scheme, that is, the optimal value in α, and generally takes its average value; Ij is the weight vector after comprehensive integration.
According to the initial weight of each influencing factor suggested by several water conservancy experts who have a deep understanding of the hydrogeological conditions in the study area, the comprehensive weight of each influencing factor is obtained as shown in Table 3.4.
Table 3.4 Weight Distribution Table of Various Influencing Factors
3) Comprehensive evaluation. Based on the multi-layer vector data superposition function of MapGIS software with arbitrary graphic boundaries, the zoning map of Guilin karst groundwater protection conditions, karst development intensity and other factors affecting the optimal design of karst groundwater level observation network is layered, and the comprehensive evaluation value of each partition is calculated by using the weighted average model in the small partition formed after superposition. The mathematical model expression is as follows
Study on the particularity of groundwater and environment in karst area
Where: r is the comprehensive evaluation value of each district; Is the weight of each zone factor; Yj is the factor assignment of each area; N is the number of all factors in each region.
According to the calculation results and comprehensive evaluation values, a semi-quantitative analysis map of karst hydrogeology is generated based on MapGIS software. Its formation process is as follows:
A. determination of comprehensive weight value of each influencing factor. According to the weight values of various influencing factors suggested by many water conservancy experts, the comprehensive weight values of various influencing factors are obtained, and the factor values and comprehensive weight values are given to each partition in the zoning map of the main influencing factors such as the abundance of karst groundwater, the protection conditions of karst groundwater and the intensity of karst development.
B. Based on MapGIS layer superposition function, each partition map is superimposed into a comprehensive layer.
C. Derive the attribute values (factor values and weights of each partition map) of the comprehensive map, calculate the comprehensive evaluation value (r) of each small partition in the comprehensive map by using the comprehensive evaluation model, and then import it into the overlay layer. The calculation results are between 0 and 7, divided into 7 intervals, and each interval is given a color and represented in the comprehensive map, so as to obtain the semi-quantitative analysis map of karst hydrogeology, as shown in Figure 3. 13.
Fig. 3. 13 Schematic diagram of semi-quantitative analysis of karst hydrogeology
4) Evaluation results and analysis. It can be seen from the semi-quantitative analysis chart of karst hydrogeology that most of the observation wells in the original observation network are distributed in areas with high evaluation value, and there are 4 1 observation well in areas with evaluation value above 3 (Table 3.5). The higher the evaluation value, the greater the hole distribution rate.
Table 3.5 Distribution area and hole distribution rate of each interval of the original observation network
(2) Quantitative analysis of optimal design of karst groundwater level observation network in Guilin.
1) Analysis of karst hydrogeological conditions. In karst areas, the strong heterogeneity of groundwater-bearing medium brings certain difficulties to the layout of groundwater observation network and the selection of optimization methods. After analyzing and summarizing the hydrogeological conditions in the study area, Kriging method is suitable for the optimal design of karst groundwater observation network in Guilin, and its basis is as follows:
A. Observation wells are concentrated in LAM Raymond Plain or on both sides of Lijiang River, and the hole depths are all within 100m. However, within the depth range, karst develops strongly, and karst caves are connected with each other through cracks and micro-cracks, so that karst groundwater keeps close hydraulic contact and forms a unified karst aquifer.
B. The water-bearing media in the groundwater system in the study area are mainly karst fractures and caves, and pipelines are developed locally, which is heterogeneous, but not strong; The movement mode of groundwater is mainly scattered flow, and the flow pattern of water is mainly laminar flow, which is anisotropic, but not strong. In the eastern part of the study area, the widely developed caves and fractures form an "underground cave-fracture network", which makes the karst aquifer behave as a porous aquifer and can be approximately simplified as a relatively homogeneous anisotropic aquifer.
To sum up, the groundwater system in the study area is heterogeneous and anisotropic, but it is not strong, and even shows homogeneity in a local range. Therefore, Kriging method is suitable for the study of optimal design of karst groundwater observation network in Guilin.
2) The application of Kriging method in the optimal design of karst groundwater observation network in Guilin.
A coordinates and water level values of observation wells of karst groundwater level in Guilin. Table 3.6 shows the average water level of observation wells in Guilin karst groundwater level observation network in September (1990).
Table 3.61September 1990 The average water level of observation wells in Guilin Karst Water Level Observation Network is 65438+
B. calculation of experimental variogram. Use the experimental variation function formula to calculate the experimental variation function values in all directions. Through analysis, the values of experimental variogram calculated along the four directions of north-south, north-south, northeast-southwest and northwest-southeast are similar, which also shows that the heterogeneity of water-bearing media in groundwater system is not strong within the distribution range of observation wells.
Because of the irregular distribution of observation wells, the average distance method can be used to calculate the experimental variogram. First, calculate the distance between each observation point and other observation points, and divide them into 8 categories. Then calculate the number and average value of each category based on Matlab programming, and then calculate the average experimental variation function (H) as shown in Table 3.7.
Table 3.7 Average value and average experimental variogram value of each type of hij in the original observation network
According to the average value of each type of hij and the corresponding average experimental variogram value (h), the experimental variogram curve is made, which is the dotted line in Figure 3. 14.
Fig. 3. Curve fitting diagram of14 variogram
C. fitting of experimental variogram. The weighted linear programming method is used to automatically fit the variogram, and the spherical model is selected as follows
Study on the particularity of groundwater and environment in karst area
The fitting problem of variogram of spherical model is transformed into multiple linear regression problem.
According to the weighted linear programming method, b0= 1.86, b 1=0.6 1, b2 = 0.02 are all calculated by Matlab programming. Then we calculate C0= 1.86, a=3. 17, C= 1.29, and the prior standard deviation C(0)=C0+C=3. 15.
Table 3.8 Weighted linear programming values of x 1j, x1j and ωj
sequential
Note: A is the amplification factor, which can be set according to the man-machine interaction mode during fitting.
According to the calculation results in Table 3.8, the theoretical variogram of the best fit can be obtained as follows.
Study on the particularity of groundwater and environment in karst area
D. calculation of standard deviation of estimation error and water level estimation. After determining the variogram model, kriging equation can be used to calculate the estimation error of each observation well and the standard deviation of water level estimation.
Firstly, the estimation error and standard deviation of water level estimation are calculated by ordinary kriging equation. In the calculation, because the kriging coefficient is negative, the improved kriging model is used for calculation. As can be seen from Figure 3. 13, when the distance between two observation well points is greater than the range of 3. 17km, the variation function γ(h) is almost unchanged. This shows that when the distance between two points is greater than this range, other observation wells in the observation network have an influence on the estimation of the estimated points, but the influence is not significant and can be ignored. In order to make the calculation simple and meet the accuracy requirements, all observation wells whose distance is less than the variation range are selected to participate in the calculation of Kriging weight coefficient and estimation error variance. The calculation process of kriging weight coefficient and estimation error variance of observation wells Gⅲ 1 and Gⅲ63 is introduced in detail below.
A.gⅲ 1. According to the calculation, the distance between observation wells Gⅲ3, Gⅲ32, Gⅲ35 and Gⅲ38 and Gⅲ 1 is less than 3. 17km, so only four observation wells are involved in the calculation. See table 3.9 for the variation function values between them and between them and observation well G Ⅲ/Kloc-0.
Table 3.9 Variation function γi 19, γij, weight coefficient λi and Lagrange multiplier μ
Using the data in the table, calculate the variance equation of G Ⅲ1estimation error:
Study on the particularity of groundwater and environment in karst area
The estimated water level of Gⅲ 1 is
Study on the particularity of groundwater and environment in karst area
The difference between the estimated water level and the actual water level is 0.03m.
B.g Ⅲ 63. According to the calculation formula of variogram, the variogram values of all observation wells with the distance less than this range, and the variogram values of these observation wells and G Ⅲ 63 are calculated, as shown in Table 3. 10.
Table 3. 10 variogram γi63 and γij, weight coefficient λi and Lagrange multiplier μ
According to the variation function value in the table, the variance of the estimation error of G Ⅲ 63 water level is 4.8852, and the estimated water level value obtained by substituting the weight coefficient into the water level estimation formula is 144.95m, which is 0.90m different from the actual situation.
The calculation process of other observation wells in the original karst groundwater level observation network is the same as that of observation wells Gⅲ 19 and Gⅲ63. See Table 3. 1 1 for the error of water level estimation and the standard deviation of observation well water level estimation.
As can be seen from Table 3. 1 1, the standard deviation of the average estimation error of the groundwater observation network in the study area is 3.9820. Therefore, the critical value of the standard deviation of a given estimation error is 3.9820. According to the contour map of standard deviation of estimation error, appropriately increase observation wells in the area with large standard deviation of water level estimation error; On the contrary, in the area where the standard deviation of water level estimation error is small, the number of observation wells decreases. Then several optimization schemes are drawn up, and the average estimation error and standard deviation of water level estimation under each scheme are calculated. After precision comparison and cost analysis, the best scheme is selected.
Table 3. 1 1 Standard deviation of estimation error of each observation well in the original observation network
Note: δh refers to the deviation between the estimated water level and the actual water level. Because there are no observation wells within the range of 3.1and gⅢ 71around, they cannot participate in statistical calculation.
In addition, from the calculated estimated water level of observation wells, it can be known that the estimated water level is more accurate in areas with dense observation networks, while in some surrounding areas or areas with large water level changes, the estimated water level is inaccurate or even biased because of the small distribution of observation wells. For example, only observation well GⅢ 45 is distributed in observation well GⅢ 41,so the estimated water level value of GⅢ 45 is115m different from the actual value. The observed water levels of observation well 16, such as gⅢ11,gⅢ14, gⅢ 23 and gⅢ 25, are all in the range of observation well gⅢ13, which are higher than those of observation well, and the water level of observation well gⅢ 57 is the smallest, which is still higher than that of observation well. The estimated value will naturally not be less than the lowest water level value in the observation well, that is, the deviation between the estimated water level value and the actual observed water level value should be above 0.72m m. Therefore, the observation well should be arranged around the observation well where the karst groundwater level changes greatly, so as to extract the hydrogeological information in the study area more comprehensively.
According to the calculated standard deviation and water level estimation value of the estimation error of karst groundwater level in the original observation network, the standard deviation contour map of the estimation error is generated based on MapGIS software by using Kriging spatial interpolation technology, as shown in Figure 3. 15 and Figure 3. 16. The results show that the standard deviation isoline of estimation error generated by kriging spatial interpolation method in MapGIS can truly reflect the actual situation. Qifeng Town and observation wells in the northeast corner of the study area are too far away from the surrounding observation wells to calculate the standard deviation of estimation error, so there is a lack of isoline distribution, and observation wells need to be added to reflect the dynamics of karst groundwater level in Guilin more comprehensively. There are many observation wells near Karst Research Institute and Gui Xiang Railway, and the standard deviation of estimation error is also small, so a certain number of observation wells can be reduced.
Fig. 3. 15 Schematic diagram of standard deviation isoline of water level estimation error of original karst groundwater observation network
It can be seen from the groundwater flow field map made by using the estimated water level and the original groundwater flow field map that the flow field in Chaoyang changes greatly. Because there are few observation wells, only observation wells G Ⅲ 41and G Ⅲ 45 estimate the water level each other, which leads to a large error in water level estimation, which makes the estimated groundwater flow field unable to truly reflect the actual situation, and also shows that the original observation network is unreasonable.
E. draw up the selected scheme. Due to the sparse layout of karst groundwater level observation network in Guilin, adjustment is the main method under the condition of appropriately reducing a few observation wells. The contour map of standard deviation of water level estimation error of the original observation network and the semi-quantitative analysis map of karst hydrogeology are combined into a comprehensive map. In the area where the standard deviation of water level estimation error is large, after analyzing the hydrogeological conditions, observation wells are added in the districts with large evaluation value in this area; In the area where the standard deviation of estimation error is small, and in the sub-area where the evaluation value is small, the number of observation wells will decrease. According to this principle, two optimization schemes are drawn up. Then the estimation error and standard deviation of water level estimation of each scheme are calculated by improved kriging method. Finally, according to the cost analysis and precision comparison, a better scheme is selected.
Fig. 3. 16 schematic diagram of water level estimation isoline of the original karst groundwater observation network
(According to the water level data in September 1990)
Scheme 1:
A. The standard deviation of estimation error of observation wells distributed in Gui Xiang railway is small, and the maximum standard deviation of estimation error is 3.794 1. Therefore, according to the evaluation value of 1 ~ 2 in the karst hydrogeological evaluation zoning map, the observation wells G Ⅲ 47 and G Ⅲ 48 are reduced, and the observation well G Ⅲ 46 which provides the same water level information as G Ⅲ 47 is retained. In the range of 2 ~ 3, the G Ⅲ 42 in the observation well decreases. Because of the dense distribution of groundwater level isoline and large hydraulic gradient, it is a groundwater runoff area, so observation wells Gⅲ24 and Gⅲ42 are still retained. Only observation well G Ⅲ 6 is distributed in Changhai Machinery Factory and its vicinity, and the standard deviation of its estimation error is 5.5694. Increase observation well Z 1 in the range of 6 ~ 7. Although the karst research institute and the banks of Lijiang River in the city center are groundwater discharge areas, there are many observation wells, and the standard deviation of estimation error is very small. Therefore, the observation wells G Ⅲ 53, G Ⅲ 54, G Ⅲ 55, G Ⅲ 59, G Ⅲ 60 and G Ⅲ 68 distributed in the area with high evaluation value can decrease at certain intervals.
B. Tuomu Town is a groundwater runoff area, and observation well Z2 is added in the area with high evaluation value; Wayao is a densely populated area, with high density of karst collapse disaster and large standard deviation of estimation error, so observation well Z3 is added in the area with evaluation value of 5 ~ 6; North of Chaoyang is a groundwater discharge area, and the estimated water level flow field in this area can't really reflect the actual situation, and the standard deviation of the estimation error is also large, so observation well Z4 is added in the interval of evaluation value 5 ~ 6.
C observation wells gⅢ 20 and gⅢ 21are far away from the surrounding observation wells, so they are slightly moved to participate in the calculation of standard deviation of estimation error and water level estimation; The observed water levels of G Ⅲ 29 and G Ⅲ 71are quite different from those of other observation wells. In order to reflect hydrogeological information more comprehensively, an observation well Z5 was added in Tangjiawan between them.
In addition, because observation wells G Ⅲ17 are very close to observation wells G Ⅲ 65, G Ⅲ 31and G Ⅲ 64, the water level values and dynamic changes are basically consistent, and the distribution of observation wells G Ⅲ17 and G Ⅲ 31with low evaluation values is reduced. In this scheme, on the basis of the original observation network, observation wells 1 1 are reduced, moved by two places, and 5 observation wells are added. After optimization, there are ***43 observation wells. 1 See Figure 3. 17 for the distribution of karst groundwater level observation network in Guilin.
Fig. 3. Schematic diagram of 1 7 scheme1standard deviation isoline of water level estimation of karst groundwater observation network
After adjustment and optimization, the estimated variance, standard deviation and water level estimation of each observation well are calculated. Because the spherical model describes the change of regional variable structure, it is only a local adjustment on the basis of the original observation network, and the spatial structure of the whole observation network has not changed much, so the adjusted and optimized observation network still adopts the spherical model fitted by the original observation network. According to the coordinates and water level values of each observation well in the adjusted observation network, the standard deviation of estimation error and water level estimation calculated by using the improved Kriging program model compiled by Matlab are shown in Table 3. 13, and the standard deviation of estimation error and isoline of estimated water level generated by MapGIS software are shown in Figures 3. 17 and 3. 18.
Figure 3. 18 Scheme 1 Schematic Diagram of Water Level Estimation Isogram of Karst Groundwater Observation Network
(According to the water level data in September 1990)
Option 2:
In the scheme 1, the standard deviation of estimation errors of gⅢ 41,gⅢ 65, gⅢ 71and gⅢ 20 is still relatively large. Therefore, in Scheme 2, based on Scheme 1, observation wells Z6, Z7, Z8 and Z9 are added in these four areas with high comprehensive evaluation values near the observation wells. The calculation results are shown in Table 3. 12, and the standard deviation of estimation error and the contour map of estimated water level generated according to the calculation results are shown in Figure 3. 19 and Figure 3.20.
F. analysis and determination of alternatives. According to the calculation results of the original observation network, scheme 1 and scheme 2, the standard deviation of estimation error and the contour map of estimated water level, it can be seen that although the number of observation wells is reduced after optimization, the standard deviation of average estimation error is increased, while the average deviation between estimated water level and actual water level is reduced. The main reason is that most of the original observation networks are located in the center of the city, and there are few observation wells around them, and some of them can't even be statistically calculated, which makes the standard deviation of the average estimation error of the original observation network smaller. For the whole study area, the scheme 1 and scheme 2 are optimized and adjusted. In order to fully reflect the water level information of karst groundwater, observation wells are reduced in areas where observation wells are concentrated, and observation wells are increased in surrounding areas and areas where water level changes greatly, so that all observation wells participate in statistical calculation after optimization, the observation range of observation network is also expanded, and the number of observation wells is reduced, so the standard deviation of estimation error is also increased accordingly. Due to the addition of observation wells in areas with large water level changes, the deviation between the estimated water level in observation wells and the actual water level becomes smaller, and there will be no big deviation in water level estimation. Therefore, the estimated deviation of the average water level of the optimized observation network is lower than that of the original observation network.
Fig. 3. 19 schematic diagram of standard deviation isoline of water level estimation error of karst groundwater observation network in scheme 2.
The optimized observation network has observation wells in the direction and vertical direction of groundwater flow, observation wells in groundwater recharge area, runoff area and drainage area, observation wells providing redundant information in drainage area of Lijiang River basin are reduced, and observation wells are added in areas with large hydraulic gradient, so the optimized scheme is more reasonable in spatial layout than the original observation network. From the analysis of flow field diagram, the water level flow field estimated by the original observation network has great changes with the original observation network, while the water level flow field estimated by the optimized scheme is more in line with the reality.
Fig. 3.20 Schematic diagram of standard deviation isoline of water level estimation error of karst groundwater observation network in Scheme II
(According to the water level data in September 1990)
Table 3. Calculation Results of Scheme 1 21and Scheme 2
sequential
See Table 3. 1 3 for the accuracy and cost analysis and comparison of the original observation network, scheme1and scheme 2.
Table 3. 13 Scheme 1 Comparison with Scheme 2
Note: It costs about 80,000 yuan to drill a new well. The operation management fee and repair fee of each well are calculated at 5% of the annual project investment, and the design life of each scheme is 10 year.
It can be seen from the table that the scheme 1 can save 240,000 yuan, but the standard deviation of the average estimation error is larger, which is 15.8% higher than the original observation network, indicating that the layout of the whole observation network after optimization is sparse and cannot meet the accuracy requirements; Scheme 2 adds 4 observation wells on the basis of scheme 1. Although the cost is only 80 thousand yuan, the observation accuracy is improved a lot. According to the relatively sparse layout density of the existing observation network in Guilin, the second scheme is preferred to ensure the observation accuracy.