## 1 Introduction

The spatial correctness of any investment project depends on the quality of the control network. On the territory of Poland, in accordance with technical regulations in practice, class 3 control networks are usually used [1,2]. Geodetic control points are physically represented in the field by appropriate monuments that ensure the stability of their coordinates. The coordinates, recorded in the National Geodetic and Cartographic Database on the basis of the original measurement, are then used for years in the unchanged form. In terms of accuracy, class 3 control network is considered superior for all measurement tasks, including the cases when the investment project is implemented on more accurate autonomous networks. The issue of aligning control networks of variable accuracy is solved with the use of Hausbrandt’s posttransformation correction [3,4,5]. It should also be noted here that any corrections of the coordinates of individual geodetic control points is only possible if the entire control network undergoes an upgrade or an update measurement [1].

Class 3 control network is functional if it meets the requirements related to the accuracy of coordinates of control points and the density of monuments in the field. According to the applicable technical regulations [1], the mean error of adjusted position of a class 3 network point for control formerly classified as class II should be *m*_{p} ≤ 0.05 m and for control points formerly classified as class III, *m*_{p} ≤ 0.10 m. The coordinates of control points in PL-2000 coordinates system were determined by transforming the “1965” coordinates system. Due to local deformations and errors of the original system, as well as its operating time, the currently available data may be – and, as practice shows, frequently are – substandard.

The requirement on the density of monuments provided in the relevant technical regulations is imprecisely formulated. On the developed area, the density of class 3 control network points should be at least 1 point on 20 hectares [2]. In practice, control points are expected to be placed not further than 200–300 m from each other. The actual condition is verified as a part of the control network upgrades. The recommendation provided in the applicable technical regulations that the control network’s condition should be inspected and the control points’ coordinates should be updated on as-needed basis is correct from a factual and technical standpoint, but ambiguous. The lack of a clearly defined criterion is often exploited by the local authorities. The inspection intervals are extended for economic reasons.

The assessment of functionality of class 3 control networks requires

- formulating the measures of a global assessment of the control network functionality,
- standardizing the procedure of examining the control network degradation process, and
- developing a method of predicting the control network degradation process and planning the update measurements.

The model of assessment of control network functionality presented herein is based on the reliability approach. Only the reliability-based model allows combining variables of different properties. In the issue under consideration herein, accuracy is provided in metric units, while the technical condition and the availability of monuments are without unit. The theoretical background of the functionality model is provided in ref. [6]. An example of the application of the model in studying vertical control networks is provided in ref. [7]. In this study, the reliability model is used in the analysis of the horizontal controls degradation process. As in all considerations of reliability of a structure, an important element of the study is the statistical analysis of functionality of horizontal controls. The authors have been conducting this study since 2016 on control nets located in different parts of the Lodz and Kielce regions.

## 2 Data acquisition in horizontal control network functionality studies

In the proposed functionality analysis method, the control network “wear” process is described by two variables: usability and stability [6,7]. The geodetic control points stabilized in the field with monuments perform their task provided that they are in good working condition and the coordinates are stable. In practice, good working condition does not always qualify a control point as meeting the measurement requirements. This is the case for control points established close to the walls, in forests, shrubbery, marshes, restricted areas, on considerable depth, etc. It should be pointed out that in the case of doubt or the need to conduct additional measurements using traditional methods, a surveyor usually seeks another control point. In this context, usability is a more generalized feature than the working condition, as it provides for both the condition of the monument and the possibility of taking a measurement. Consequently, for the reliability-based approach, qualification is arbitrary and, in each case, binary.

Stability data are acquired from the results of the control measurement. This type of measurement is usually taken using kinematic techniques (RTK and RTN) on at least two geodetic control points located not farther than 5 km from the measured control points. The line vector of coordinate’s difference *f*_{XY}, determined based on the control measurement, must not exceed 0.12 m [2]. The result of the control measurement is affected by the error or primary and current measurements and various other factors that are usually impossible to identify. In the study results presented herein, the impact of the measurement error was ignored, arguing that for the RTK GNSS measurement, the error does not exceed 1–3 cm, with the acceptable *f*_{XY} deviation of 0.12 m.

## 3 Theoretical background of the functionality study outcomes analysis

In the technical analysis, reliability is usually defined as the probability that an object meets specified criteria over a period of time [*t*_{0}, *t*] of system operation [8,9,10]. The concision of assessment of the quality of the whole system with a single characteristic is one of the key advantages of the reliability-based model. A less advantageous feature of this approach is the need to gather data through research tests conducted on large sets.

The functionality model is defined by three functions, i.e., the reliability function *R*(*t*), the failure risk function (hazard function) *F*(*t*). The reliability function *R*(*t*) defines the probability of the correct operation of the object’s system in the period of 〈0, *t*〉, i.e., from the moment of handing over the object for operation.

*R*(

*t*) is related to function

*F*(

*t*), which defines the random stability life of the object [6].

*t*

_{N}is the time of operation of the object. By definition, the function takes the following form:

*λ*(

*t*) is defined by the relation:

*R*(

*t*),

*F*(

*t*) or

*λ*(

*t*). The most frequently determined function is the risk function

*λ*(

*t*), whereby in practice, it is usually assumed that:

*λ*= const is a simplification, as the intensity of the destruction process varies in different periods of time. However, in the case of a study of the control network functionality, we have only one information, i.e., the total number of geodetic monuments that meet the usability and stability criteria in the period of time since the last upgrade. The value of variable

*λ*is determined after transforming the survival function

*R*(

*t*).

*λ*is determined from the following relation:

*T*

_{p}is the period of study from baseline (no. of years),

*N*

_{p}is the number of geodetic control points according to the catalogue, and

*n*

_{p}is the number of control points that meet the requirement of both usability and stability.

An alternative method is based on separating the stability and usability tests [6,7]. In such case, the outcome of the test are two independent data sets and the probability of meeting the criteria of usability *P*_{u} and stability *P*_{s} determined on the basis thereof. Since the system is functional, if both conditions are met at the same time, the following equation is obtained [10]:

*λ*

_{p}is determined as the sum of the components

*λ*

_{u}(usability) and

*λ*

_{p}(stability).

*λ*

_{u}and

*λ*

_{s}are determined separately based on the probabilities

*P*

_{u}and

*P*

_{s}.

From a practical standpoint, the key element of the functionality analysis is the determination of the number of years *T*_{kr} after which the control network no longer meets the functionality criteria. In the authors’ opinion, two critical functionality states should be distinguished, i.e., the first “warning” state when *R*(*T*_{kr1}) = *P*_{kr1} = 75% and the second when *R*(*T*_{kr2}) = *P*_{kr2} = 50%. The value of *T*_{kr} is determined by the following relation:

## 4 Example of functionality analysis of the class 3 control network

### 4.1 Survey results

The analysis procedure was illustrated using the example of a class 3 control network located in the southern part of the Lodz Region. The upgrade and update measurement of the network were performed in 2007. The functionality analysis tests were performed in 2017 on 124 control points. The results of the survey for the purpose of the usability assessment are presented in Table 1.

Class 3 control network survey results

No. of points according to catalogue | Control points existing in the field | Destroyed control points | ||
---|---|---|---|---|

Identified control points | Unavailable control points | Control points in good working condition | ||

124 | 90 | 0 | 90 | 34 |

On the geodetic control points, which were in good working condition, the control measurement of coordinates was performed using the RTK GNSS method and the VRSnet reference stations net. It was determined that a substantial portion of the control network, i.e., 26 control points, failed to meet the criterion of acceptable deviation *f*_{XY} ≤ 0.12.

### 4.2 Control network functionality prediction and critical states

Of the initial number of control points *N*_{p} = 124, the requirement of both usability and stability was met by 64 points. The risk value *λ* was determined from formula (7) for *T*_{p} = 10 years, *N*_{p} = 124, *n*_{p} = 64, *λ* = 0.0631/year.

Based on the survival function (3)

Destruction process prediction for the control network in question and the critical states of functionality thereof

Initial state | Control measurement | Functionality prediction | Critical states | ||||
---|---|---|---|---|---|---|---|

2007 | 2017 | 2022 | 2027 | 2032 | P_{kr1} | P_{kr2} | |

1 | 2 | 3 | 4 | 5 | 6 | 7 | |

T [years] | 0 | 10 | 15 | 20 | 25 | ||

R(t) | 0 | 0.532 | 0.388 | 0.283 | 0.207 | 0.75 | 0.50 |

Number of functional control points | 124 | 66 | 48 | 35 | 26 | 93 | 62 |

### 4.3 Statistical interpretation of the control measurement results

The statistical analysis of the results of measurement of the coordinates of the control points is not an integral element of the functionality assessment. It is cognitive in nature, and its purpose is to identify the statistical properties in the set of deviations *f*_{X}, *f*_{Y} determined as the differences between the PL-2000 system coordinates and the results of the control measurement.

*f*

_{XY}≤ 0.12. The values that failed to meet the requirement were removed from the data set. Table 3 presents the frequency distribution of linear deviation

*f*

_{l}vectors after separating eight class intervals.

Summary of the mean values of azimuths and modules of linear deviations *f*_{XY}

Interval | Interval range (g) | Number of data points in an interval | Mean azimuth value (g) | Mean f_{XY} module (m) |
---|---|---|---|---|

1 | 2 | 3 | 4 | 5 |

1 | 0–49.9 | 4 | 31.4717 | 0.0562 |

2 | 50–99.9 | 5 | 78.7921 | 0.0774 |

3 | 100–149.9 | 6 | 123.4944 | 0.0704 |

4 | 150–199.9 | 6 | 176.8879 | 0.0595 |

5 | 200–249.9 | 19 | 219.5562 | 0.0774 |

6 | 250–299.9 | 26 | 272.7967 | 0.0754 |

7 | 300–349.9 | 22 | 318.4120 | 0.0603 |

8 | 350–399.9 | 2 | 358.7792 | 0.0750 |

The data presented in Table 3 and the radar charts plotted on the basis thereof (Figure 1) show the distribution of the vectors of linear deviations *f*_{XY}. The distribution of linear deviations *f*_{XY} vectors is not even; the number of observations in the 200–300 G azimuths interval is much higher than in other sectors. Such a prominent difference in the number of data points indicates a systematic error. However, the conclusion of the significant impact of the systematic factor is not supported by the radar chart. The mean values of *f*_{XY} modules calculated for individual sectors do not show significant differences (Table 3, column 5).

To identify the statistical properties of the data set, a correlative dependence analysis was performed. The following were determined:

- mean values and standard deviations of variables
*f*_{X},*f*_{Y}, - correlation coefficient from the value of variables
*f*_{X},*f*_{Y,} - linear regression functions, and
- standard deviation ellipse.

The results of the analysis are presented in Table 4 and Figure 2.

Statistical analysis of the results of the control measurement

Statistical value | f_{X} (m) | f_{Y} (m) |
---|---|---|

Mean value (m) | −0.012 | −0.020 |

Standard deviation (m) | 0.044 | 0.057 |

Linear regression function coefficients | 0.0852 | 0.14445 |

Correlation coefficient | 0.112 | |

Standard deviation ellipse (m) | b = 0.045, a = 0.055 |

The analysis shows that deviations *f*_{X} and *f*_{Y} are not correlated although the observations have a minor systematic error. Once the error has been filtered out, the data will have the properties of random variables. The standard deviation ellipses before and after transformation are virtually identical.

## 5 Class 3 control networks functionality studies in the Lodz and Kielce Regions

The class 3 control network functionality studies conducted by the authors aims, on the one hand, to determine the scale of the destruction issue, and, on the other hand, to identify the patterns of this process relevant for surveying practice. The studies are conducted since 2017 in the Lodz and Kielce Regions (Figure 3) and follow a single program using a sample of 70–170 control points. Table 5 provides the basic information on the studied control networks, i.e., the object code (col. 1), measurement dates: initial and control measurement (col. 2), the number of control points: the number of points in the catalogue/number of points on which the control measurement was performed/the number of points meeting the requirements of *f*_{XY} ≤ 0.12 (col. 3), the urbanization of the area in which the studied control network is located (col. 4) and the reference stations net (col. 5).

Study objects

Object code | Dates of control network measurement | Number of control points | Area urbanization | Reference stations network |
---|---|---|---|---|

1 | 2 | 3 | 4 | 5 |

A | 2012/2018 | 70/62/49 | Municipal area | ASG-EUPOS |

B | 2007/2017 | 124/90/64 | Rural area | VRSnet |

C | 1981/2016 | 169/67/50 | Municipal area | TPI Net |

D | 2006/2018 | 107/66/59 | Municipal area | ASG-EUPOS |

E | 1977/2016 | 149/84/76 | Rural area | ASG-EUPOS |

F | 1998/2018 | 121/65/60 | Municipal area | ASG-EUPOS |

G | 1997/2017 | 156/100/59 | Rural area | VRSnet |

H | 2000/2017 | 158/80/65 | Municipal area | VRSnet |

The control measurements were performed using the RTK GNSS method, referencing various reference stations networks, as per the recommendations provided in the technical regulations [2]. In the event of unfavorable conditions, which could affect the reception of satellite signals, the measurement session was extended as appropriate. Furthermore, for the study purposes, for a number of control points, measurements were taken with reference to the reference stations net other than provided in Table 5. Both in the case of extended measurement sessions and the use of different reference systems, the maximum discrepancies of the satellite measurement results did not exceed 0.03 m.

The observation results then underwent analysis identical in terms of scope and method as applied to object B presented in item 4. First, the observations exceeding the linear deviation of 0.12 m were excluded from the analyzed set. It should be emphasized that such deviations were found in all analyzed data sets. The number of control points provided in column 4 in Table 5, e.g., 149/84/76 ,means that the control measurement could be performed only on 76 points.

The analysis results summarized in Table 6 and shown in Figure 4 show a number of statistical patterns:

- In all cases, the mean values of deviations
*f*_{Y}are negative. The data sets are shifted slightly westward. The result correlates with the size charts provided on the radar charts (Figure 4), whereby on the latter one, the systematic impact is much more prominent. - The variables
*f*_{X}and*f*_{Y}are not correlated. The correlation coefficient did not reach the value, indicating the presence of a correlative dependence in any of the analyzed samples. - The standard error calculated for variables
*f*_{X}and*f*_{Y}, and ellipse semi-axes show only minor differences.

Statistical analysis of the results of the control measurement

Object no. | Number of points | Mean values | Standard deviation | Correlation coefficient | Azimuth of semi-axis A (g) | Semi-axes of ellipses | |||
---|---|---|---|---|---|---|---|---|---|

f_{X} (m) | f_{Y} (m) | σ_{jX} (m) | σ_{jY}(m) | A (m) | B (m) | ||||

A | 49 | −0.005 | −0.006 | 0.022 | 0.023 | −0.12 | −7.87 | 0.022 | 0.022 |

B | 64 | −0.012 | −0.020 | 0.044 | 0.057 | 0.11 | 7.31 | 0.045 | 0.055 |

C | 50 | 0.023 | −0.014 | 0.054 | 0.049 | 0.165 | 10.46 | 0.055 | 0.048 |

D | 59 | 0.013 | −0.008 | 0.049 | 0.043 | −0.19 | −12.28 | 0.050 | 0.041 |

E | 76 | 0.023 | 0.003 | 0.036 | 0.040 | 0.19 | 12.27 | 0.038 | 0.039 |

F | 60 | 0.005 | −0.029 | 0.035 | 0.039 | −0.10 | −6.07 | 0.035 | 0.038 |

G | 59 | 0.006 | −0.050 | 0.041 | 0.039 | 0.01 | 0.21 | 0.040 | 0.039 |

H | 65 | 0.012 | −0.030 | 0.039 | 0.038 | 0.12 | 7.48 | 0.040 | 0.037 |

## 6 Prediction of the functionality destruction process of the analyzed class 3 control networks

Table 7 summarizes the outcomes of the functionality prediction calculated for three subsequent 5 year operation periods and the years in which functionality reaches critical values (11). It should be noted that the first state, i.e., the alarm critical state, was not exceeded only for object A. The second critical state was exceeded for all analyzed data sets.

Prediction of critical functionality states of analyzed objects

Object | λ | Initial measurement | Number of functional control points | Critical states | |||||
---|---|---|---|---|---|---|---|---|---|

Date | Number | 2018 | 2023 | 2028 | 2033 | 75% | 50% | ||

A | 0.059 | 2012 | 70 | 49 | 37 | 24 | 18 | 2020 | 2024 |

B | 0.063 | 2007 | 124 | 62 | 45 | 33 | 24 | 2015 | 2019 |

C | 0.035 | 1981 | 169 | 46 | 39 | 32 | 27 | 1995 | 2002 |

D | 0.056 | 2006 | 107 | 55 | 41 | 31 | 24 | 2015 | 2019 |

E | 0.017 | 1977 | 149 | 74 | 68 | 63 | 58 | 2006 | 2021 |

F | 0.056 | 1998 | 121 | 40 | 30 | 23 | 17 | 2007 | 2011 |

G | 0.047 | 1997 | 156 | 58 | 46 | 36 | 29 | 2008 | 2013 |

H | 0.054 | 2000 | 158 | 60 | 46 | 35 | 27 | 2009 | 2014 |

Based on the data in Table 6, a general conclusion can be drawn on the condition of the control networks and a detailed one on the intensity of the destruction process in different regions. The risk variable *λ* fluctuates around the mean value of 0.046 between the extreme values of 0.063 and 0.017. With the exception of the extreme values, the risk variable does not show significant variation. The significant deviation of object **E** can be explained by the exceptionally high local investment activity (central districts of Lodz city).

## 7 Final remarks and conclusions

### 7.1 Identification of the functionality model

The presented method of functionality assessment is a coherent solution in terms of the test purpose, an unambiguous criteria expressed in the form of probability interpreted according to the reliability engineering theory and data acquisition procedures as well. Only the reliability approach integrates different properties such as usability and stability in one model.

The critical state level is a questionable element of the method. The correct indication of critical probabilities provides a rational basis for the decision concerning the date of maintenance actions. The authors suggest introducing two critical functional states at the levels *P*_{cr1} = 75% and *P*_{cr2} = 50%.

The destruction of the control network is influenced by two factors. The main one is the physical destruction of network benchmarks. Since the cause of destruction is the investment activity necessary for the functioning of the economy, therefore, the solution to this problem depends on decisions made by regional surveying departments and local administration. Technically, the problem is solved by supplementing the existing network and new measurements.

The presented method of evaluating functionality can be used regardless of the structure and the number of points. An optimal solution is to make an inventory and control check measurement for all control network points stabilized in a specified area, e.g., in the area of a commune. In case of significant number of points, the survey may be of statistical nature. Authors' experience has shown that achieving a reliable result requires testing a minimum of 100 points. The measurement should be carried out with a standard error *σ*_{p} ≤ 0.03 m, i.e., at least twice as high as the accuracy of the coordinates specified in technical regulation [1,2].

According to the authors, the scope and the detail of the research carried out entitles to conclude that the presented procedure can be adopted as a standard procedure, useful in the assessment of the quality of regional geodetic resource databases.

### 7.2 Statistical properties of the destruction process of the class 3 control networks

The problem of evaluation of horizontal control networks functionality in the article is presented on the example of class 3 control networks, which in Poland are stabilized by means of permanent markings. The local characteristics of these networks follow from the fact that the current coordinates of points in the “2,000” system are the result of the transformation of coordinates originally calculated in another reference system (“65”). The main conclusion is that the level of network destruction is significant and is dependent on the exploitation period. Changes in coordinate values exceeding the critical values usually affect some percentage of points. But the resultant value of the risk parameter *λ* is always significant, on average *λ* = 0.046.

The measurement results presented in Table 6 and Figure 3 indicate that the coordinates are influenced by a systematic factor. Asymmetry of the distribution of the number of deviations *f*_{XY} is visible in all test results. The direction of the translation vector is approximately consistent with the *Y*-axis. The value of translation and standard errors of the variables *f*_{X}, *f*_{Y} confirm the opinion that the accuracy of transformation of the primary system to the system “2,000” is within the limits of 0.05 m. Since it is impossible to indicate whether the observed coordinate differences are the result of an invalid primary measurement or due to unrecognized factors, it is optimal to update the network in larger areas. Currently, such corrections are carried out locally in accordance to the investment tasks.

In the analysis of the functionality issue, a research workshop is important. The coordinates of network points were determined using the GPS kinematic technique combined with ASG-EUPOS, VRSnet and TPInet reference station networks. To estimate the accuracy of the determined coordinates, some of the points were observed with extended observation time or in connection with another reference network. The standard error determined on this basis is similar for all objects, approximately 0.02 m.

### 7.3 Modernization of control networks by means of satellite measurement techniques

The high level of destruction and maintenance costs of networks on the one side and the effectiveness of satellite measurements on the other become rational arguments for the lack of acceptance for static control nets [11,12]. Currently, in the opinion of the authors, this is a premature opinion, although optimization of the methods of design and maintenance of class 3 control networks is necessary and possible taking into account local conditions. On the one hand, databases of class 3 control networks available in Polish regional documentation centers are components of the state surveying and cartographic resource [3,4]. These databases have an important role in the functioning of various segments of economy and administration. On the other hand, from a technical point of view, permanent signs are necessary in areas beyond the range of permanent reference stations, in densely built-up areas, in areas where it is not possible to effectively distribute GPS adjustment corrections via the GSM network or due to significant land surpluses also the UHF radio communication, etc.

The results of this study confirm the necessity to rethink the approach to the problem of networks' functioning and indicate the directions of these changes. In the opinion of the authors, the modified approach should be complementary, i.e., take into account the network structure, number of points and coordinate measurement methods. The main postulate of the proposed concept is to replace the network with a set of autonomous points. Local connections of adjacent points would be made by means of classical measurements, but the latter would only have a check character. The planning of location and number of points should take into account the process of destruction. The level of destruction can be assumed on the basis of the analysis of critical states determined by means of a functionality model. Such assumption will significantly increase the number of network points.

An important element of the proposed concept is to measure the coordinates of network points. Within this task it is possible to use differential GNSS positioning, precise point positioning (PPP), linear-angle measurements and hybrid methods integrating different techniques. The presented research results show that at the stage of control measurements sufficient accuracy is provided by the kinematic technique. When establishing a network, measurements should be carried out using static methods in accordance with the recommendations [1], but practice proves that under favorable conditions, kinematic techniques are also satisfactory. In Poland, measurements using differential GNSS method are possible thanks to several systems of reference stations, mainly the ASG-EUPOS system.

An alternative to differential measurements is technique of precise satellite positioning using a global fixed station infrastructure [13,14,15]. The advantage of the absolute PPP method is its autonomy. The measurement is performed without the need to relate to regional reference stations, while access to the GNSS (IGS) data on GNS and GLONASS is necessary. The PPP method is still a subject of research, and its accuracy depends on the length of the session and IGS data [16] The results in the range of 0.02–0.03 cm achieved with sessions of about 2–3 h can be assumed as sufficient for class 3 network measurements. The accuracy requirements for coordinate measurements are also met by Smart Station technology integrating satellite measurements with classical measurements [17]. It can be assumed that this technique will be particularly useful at the stage of control measurements, as it gives the possibility to control both the station and the adjacent points.

The project is supported by the program of the Minister of Science and Higher Education under the name: “Regional Initiative of Excellence” in 2019–2022 project number 025/RID/2018/19 financing PLN 1,20,00,000.

## References

- [1]↑
Regulation of Minister of Interior and Administration in case of geodetic, gravimetric and magnetic control network, (Rozporządzenie Ministra Administracji i Cyfryzacji z dnia 14 lutego 2012 r. w sprawie osnów geodezyjnych, grawimetrycznych i magnetycznych), Journal of Laws No 352, Warsaw: Government Legislation Centre; 2012.

- [2]↑
Regulation of Minister of Interior and Administration – in case of technical standards of performing detailed surveys and working out and sending results of these surveys to National Geodetic and Cartographic Database, (Rozporządzenie Ministra spraw wewnętrznych i administracji), Journal of Laws No. 263. Warsaw: Government Legislation Centre; 2011.

- [3]↑
Doskocz A. The current state of the creation and modernization of national geodetic and cartographic resources in Poland. Open Geosci. 2016;8(1):579–92.

- [4]↑
Banasik P, Ligas M, Kudrys J. Analysis of the local HiL coordinate system for a coordinate transformation to the PL-2000. Geomat Environ Eng. 2016;10(4):15–26.

- [5]↑
Kadaj R. Polish coordinate systems. Transformation formulas, algorithms and programs, (Polskie układy współrzędnych. Formuły transformacyjne, algorytmy i programy), ALGORES-SOFT Do usunięcia, www.geonet.net.pl.

- [6]↑
Wolski B. Operational reliability of geodetic control points, Geodesy and Cartography. Committee for Geodesy. Pol Acad Sci. 2007;56(2):83–94.

- [7]↑
Wolski B, Toś C. Probabilistic model of assessment of level network functionality. Geomat Environ Eng. 2017;11(2):73–83.

- [8]↑
Prószyński W, Kwaśniak M. Reliability of geodetic network (in polish). Warsaw: Warsaw University of Technology; 2002.

- [9]↑
Migdalski J. Reliability engineering (Inżynieria niezawodności). Warsaw: University of Technology and Agriculture in Bydgoszcz ATR, ZETOM; 1992.

- [11]↑
Markku P, Pasi H. Future of National reference frames – from static to kinematic? Geodesy Cartograph Pol Acad Sci. 2018;67(1):117–29.

- [12]↑
Rizos C, Janssen V, Roberts C, Grinter T. Precise point positioning: is the era of differential GNSS positioning drawing to an end? Proceedings of FIG Working Week 2012 “Knowing to manage the territory, protect the environment, evaluate the cultural heritage”, Rome, 2012

- [13]↑
Choy S. High accuracy precise point positioning using a single frequency GPS receiver. J Appl Geodesy. 2011;5:59–69.

- [14]↑
Martín A, Anquela B, Capilla R, Berné JL. PPP technique analysis based on time convergence, repeatability, IGS products, different software processing, and GPS+GLONASS constellation. J Surveying Eng. 2011;137(3):99–108.

- [15]↑
Malinowski M, Kwiecień J. A comparative study of precise point positioning (PPP) accuracy using online services. Rep Geodesy Geoinformatics. 2016;102(2016):15–31.

- [16]↑
Stępniak K, Wielgosz P, Paziewski J. Analysis of PPP accuracy depending on observing session duration and GNSS systems used, (Badania dokładności pozycjonowania techniką PPP w zależności od długości sesji obserwacyjnej oraz wykorzystanych systemów pozycjonowania satelitarnego). Biul WAT. 2012;1:429–50.

- [17]↑
Doskocz A, Uradziński M. Position determination of control network points in the Smart Station Technology using ASG-EUPOS System. Rep Geodesy. 2012;92(1):155–61.