Parameter Estimation of Lunar Regolith from Lunar Penetrating Radar Data

Parameter estimation of the lunar regolith not only provides important information about the composition but is also critical to quantifying potential resources for lunar exploration and engineering for human outposts. The Lunar Penetrating Radar (LPR) onboard China’s Chang’E-3 (CE-3) provides a unique opportunity for mapping the near-surface stratigraphic structure and estimating the parameters of the regolith. In this paper, the electrical parameters and the iron-titanium content of regolith are estimated based on the two sets of LPR data. Firstly, it is theoretically verified that the relative dielectric constant can be estimated according to the difference of the reflected time of two receivers from a same target. Secondly, in order to verify the method, a parameter estimation flow is designed. Subsequently, a simple model and a complex model of regolith are carried out for the method verification. Finally, on the basis of the two sets of LPR data, the electrical parameters and the iron-titanium content of regolith are estimated. The relative dielectric constant of regolith at CE-3 landing site is 3.0537 and the content of TiO2 and FeO is 14.0127%. This helps us predict the reserves of resources at the CE-3 landing site and even in the entire Mare Imbrium.


Introduction
As said in Reference [1], the surface of the Moon has collided with small and large asteroidal and cometary materials for 25 billion years. As time passes, these collisions comminute the surface materials and 'gardened' a fine-grained layer termed "regolith."The thickness of the regolith is between 2 and 20 m, beneath the younger maria and below the oldest surface of the lunar highland separately. Parameter estimation of the lunar regolith not only provides important information about its composition but is also critical for quantifying potential resources for lunar exploration and engineering for human outposts.
The regolith parameter could be determined by the experiments of return samples from the lunar surface. The Luna and Apollo programs have collected lunar regolith and rock samples on the lunar surface and were returned to Earth. Laboratory experiments of these samples have been conducted to reveal the nature of lunar regolith. Basu et al. [2] estimated the average chemical composition and mineral composition of lunar regolith in each sampling area, according to the lunar samples of the Apollo and Luna projects; Carrier et al. [3] and Gromov et al. [4] calculated the relationship between by using the various information of the traces, such as the travel time [33], phase [34], amplitude [35], frequency spectrum [36], reflection coefficient [37], direct wave [38] and so on.
Some pre-research used LPR data to estimate the parameters of lunar regolith. Dong et al. [39] calculated the parameters of the regolith by relative reflection amplitudes; Feng et al. [40] estimate the radar velocity and the parameters of the regolith by hyperbolic matching in the CH-2B radar-gram.
In this paper, two sets of data (CH-2A and CH-2B) are used to estimate the dielectric constant of lunar regolith and the iron-titanium content. Firstly, we introduce the principle and the formula derivation of the method. Secondly, we design a flow for model verification and parameter estimation. Thirdly, a complex regolith model is carried out for the verification of this method in lunar regolith estimation. Finally, the dielectric constant and the content of TiO 2 and FeO in lunar regolith are estimated from LPR data.

Methodology
In this paper, the parameter estimation method uses two sets of radar data with different offsets. According to the difference of the arrival time of the two sets of data from the same target, we can estimate the dielectric parameters.
This section will divide into two cases. The first case: There is no space between the radar and the ground. In the second case, there is space between the radar and the ground. Figure 1 illustrates the geometric propagation paths of electromagnetic waves when the two sets of radar transmitters (T 1 &T 2 ) and receivers (R 1 &R 2 ) with different offsets (L 1 &L 2 ) are close to the ground. The dielectric constant ε r of the medium is estimated by the different arrival time ( T 1 → R 1 & T 2 → R 2 ) of the reflected waves from a same anomalous body. authors estimate parameters by using the various information of the traces, such as the travel time [33], phase [34], amplitude [35], frequency spectrum [36], reflection coefficient [37], direct wave [38] and so on. Some pre-research used LPR data to estimate the parameters of lunar regolith. Dong et al. [39] calculated the parameters of the regolith by relative reflection amplitudes; Feng et al. [40] estimate the radar velocity and the parameters of the regolith by hyperbolic matching in the CH-2B radar-gram.
In this paper, two sets of data (CH-2A and CH-2B) are used to estimate the dielectric constant of lunar regolith and the iron-titanium content. Firstly, we introduce the principle and the formula derivation of the method. Secondly, we design a flow for model verification and parameter estimation. Thirdly, a complex regolith model is carried out for the verification of this method in lunar regolith estimation. Finally, the dielectric constant and the content of TiO2 and FeO in lunar regolith are estimated from LPR data.

Methodology
In this paper, the parameter estimation method uses two sets of radar data with different offsets. According to the difference of the arrival time of the two sets of data from the same target, we can estimate the dielectric parameters.
This section will divide into two cases. The first case: There is no space between the radar and the ground. In the second case, there is space between the radar and the ground. Figure 1 illustrates the geometric propagation paths of electromagnetic waves when the two sets of radar transmitters (  TR  ) of the reflected waves from a same anomalous body. According to the geometric relationship of Figure 1, it can be known that: According to the geometric relationship of Figure 1, it can be known that: where t 1 is the time of the electromagnetic wave from T 1 to R 1 when the offset is L 1 , t 2 is the time of the electromagnetic wave from T 2 to R 2 when the offset is L 2 and v is the velocity of the electromagnetic wave in the medium. The relationship between the electromagnetic wave propagation velocity and the electrical parameters (dielectric constant and magnetic permeability) is known: where ε r is the dielectric constant and µ r is the magnetic permeability, c is the velocity of electromagnetic waves in a vacuum. Obtained from Equations (1) and (3) and Equations (2) and (3): and from Equations (4) and (5): for two sets of data with different offsets, when: (1) the transmitter and receiver are close to the ground; (2) the offsets (L 1 &L 2 ) are known; (3) the reflected time (t 1 &t 2 ) is known; and (4) the medium is a non-magnetic medium (µ r = 1), According to Equations (6) and (7), the position of the abnormal body and the dielectric constant of the medium can be estimated.
In the actual situation, the radar transmitter and receiver are difficult to get close to the ground. Generally, there is a certain height from the ground to the radar. Figure 2 illustrates the geometric propagation paths of electromagnetic waves when the two sets of radar transmitters (T 1 &T 2 ) and receivers (R 1 &R 2 ) with different offsets (L 1 &L 2 ) are at a height (h) over the ground.
Take the path from T 1 to R 1 as an example: Snell's Law: where θ 1 is the incident angle of the electromagnetic wave, θ 2 is the angle of refraction, n 0 and n r are the refractive index of the vacuum and the medium, respectively. The velocity of electromagnetic waves in the medium: n r is the refractive index of the medium, ε r and µ r are the relative permittivity and the relative permeability of the medium, c is the velocity of electromagnetic waves in vacuum. Take the path from 1 T to 1 R as an example: Snell's Law: According to the geometric relationship which can be seen in Figure 2: Bringing Equations (9)-(12) into Equation (8) and sort out: According to the electromagnetic wave path and the velocity, the time from 1 T to 1 R : According to the geometric relationship which can be seen in Figure 2: Bringing Equations (9)-(12) into Equation (8) and sort out: According to the electromagnetic wave path and the velocity, the time from T 1 to R 1 : In the same way, we can get the relationship from T 2 to R 2 : In the four Equations (13)- (16), there are 11 parameters (t 1 , l 1 , L 1 , t 2 , l 2 , L 2 , h, H, ε r , µ r , c). When: (1) the offsets (L 1 &L 2 ) are known; (2) the height (h) of the radar from the ground is known; (3) the reflected time (t 1 &t 2 ) is known; and (4) the medium is a non-magnetic medium (µ r = 1).
The four remaining unknown parameters can be obtained according to the four Equations (13)- (16). Thereby, we can estimate the position of the abnormal body and the dielectric constant of the medium.

Permittivity Estimation Procedure
According to the following steps, the same model (Figure 3a and Figure 5a) is used to describe the above two cases.    Figure 3d,e.
Then, we get 0.755 ns t  from the wavelet waveform ( Figure 4). Finally, the true arrival time 1 27.105 ns t  and 2 28.885 ns t  are obtained. Step 1: Simulation. Setting different parameters perform simulations of the model and we can obtain two sets of radar profiles with different offsets.
Step 2: Obtain t 1 and t 2 . Read the reflection arrival time of the two sets of data from the same target. Specific operations of step 2: Step 2.1. Firstly, read the first significant extreme point of the reflected wavelet and get t 1 , t 2 .
Step 2.2. Then, extract a reflected wavelet from the radar profile and read the time from the jump point to the first significant extreme point in the wavelet.
Step 2.3. Finally, the true reflected wave arrival time of the data with different offsets is obtained Step 3: Parameter estimation. The dielectric constant is obtained according to the Equations (6) and (7) or Equations (13)-(16) for different situations.
The two cases are specifically described below. Firstly, the simulation of the forward modeling ( Figure 3a) have been done. FDTD (Finite-Difference Time-Domain) is applied for the simulation of the simple model. The FDTD main code was written by Irving and Knight [41] but we modified some parts of the code (the import of the model, the export of the wavefield snapshot and so on). Some key simulation parameters are presented in Table 1 and we obtain two sets of radar profiles with different offsets (Figure 3b,c). According to step 2, we firstly read out t 1 = 27.860 ns and t 2 = 29.640 ns in Figure 3d,e. Then, we get ∆t = 0.755 ns from the wavelet waveform ( Figure 4). Finally, the true arrival time t 1 = 27.105 ns and t 2 = 28.885 ns are obtained.  Consistent with the second case, the model ( Figure 5a) is simulated firstly and the parameters are shown in Table 1. Two sets of radar profiles with different offsets are obtained (Figure 5b,c). What is different from the first case is that the radar transmitter and receiver is of 0.5m above the ground. Consistent with the second case, the model ( Figure 5a) is simulated firstly and the parameters are shown in Table 1. Two sets of radar profiles with different offsets are obtained (Figure 5b,c). What is different from the first case is that the radar transmitter and receiver is of 0.5 m above the ground.
According to step 2, we firstly read out t 1 = 31.015 ns and t 2 = 32.320 ns in Figure 5d,e. Then, we get ∆t = 0.755 ns from the wavelet waveform ( Figure 6). Finally, the true arrival times t 1 = 30.260 ns and t 2 = 31.565 ns are obtained.  Consistent with the second case, the model ( Figure 5a) is simulated firstly and the parameters are shown in Table 1. Two sets of radar profiles with different offsets are obtained (Figure 5b,c). What is different from the first case is that the radar transmitter and receiver is of 0.5m above the ground. According to step 2, we firstly read out 1 ' 31.015 ns t  and 2 ' 32.320 ns t  in Figure 5d,e.
Then, we get 0.755 ns t  from the wavelet waveform ( Figure 6).Finally, the true arrival times   Finally, following the third step, as we have already known some parameters, the quaternary equations shown below are obtained by using the Equations (13)- (16).  Finally, following the third step, as we have already known some parameters, the quaternary equations shown below are obtained by using the Equations (13)- (16).
The solution, H = 2.296 m, ε = 2.991, which is solved by least squares, is consistent with the model. Thus, the method is feasible.

Model Experiment
In order to verify the feasibility of the method on the moon, this section conducts model tests. Two models are established: the first one is a simple model with different anomalous bodies and the second one is a complex regolith model. Verification shows that this method is feasible to estimate the electrical parameters of the regolith.
The simple model is built as shown in Figure 7a, in which there are five anomalous bodies with different shapes. The relative dielectric constant of the anomalous body is 6. Figure 7b,c shows the forward results with different offsets of 1 m and 2 m, respectively. The forward parameters are shown in Table 2. the forward results with different offsets of 1m and 2m, respectively. The forward parameters are shown in Table 2.  According to the above steps of this method, we can read the arrival time at each position (Table 3). We know 0.76 ns t    According to the above steps of this method, we can read the arrival time at each position (Table 3). We know ∆t = 0.76 ns from Figure 8. The height of the anomalous bodies and the relative dielectric constant (Table 3) can be estimated by the Equations (13)-(16).    We consider several approaches to obtain the final estimated dielectric constant: (1) The final result is calculated by the mean of the recovered permittivity values: where ε is the final estimated dielectric constant, n is the number of the recovered permittivity values and ε i is the ith recovered permittivity value. (2) A weight based on the amplitude of the measured echoes is set: where ε is the final estimated dielectric constant, n is the number of the recovered permittivity values and ε i is the ith recovered permittivity value. A i is ith amplitude. (3) A weight based on the reciprocal of the estimated height of each anomalous body is set: where ε is the final estimated dielectric constant, n is the number of the recovered permittivity values and ε i is the ith recovered permittivity value. H i is ith height.
After a comprehensive consideration, we set a weight based on the reciprocal of the estimated height of each anomalous body as the final result. Since if the standard deviation of the recovered permittivity values is quite large, calculating the mean of the recovered permittivity values is too simple and is not a reliable methodology. However, if the radar data has a low signal-to-noise ratio and a high energy attenuation, it is too difficult to obtain the real amplitude of the reflection waves. If the reference target is particularly deep, the estimation could become sensibly inaccurate, so setting a weight based on the reciprocal of the estimated height of each anomalous body as the final result could be an easy and reliable method.
The estimated dielectric constant is: which is consistent with the model. In order to verify whether this method is suitable for LPR data, we build a complex model (Figure 9a). This model considers many factors: random medium, undulating interface and anomalous body. The modeling method is referenced in References [42][43][44]. According to the actual acquisition parameters of LPR [13], the simulated parameters are shown in Table 2. Two sets of forward results with different offsets are obtained in Figure 9b,c. According to the above method, reading from the two sets of radar data, t 1 and t 2 are shown in Table 4. ∆t = 1.2535 ns is from Figure 10. The height and the relative dielectric constant (Table 4) are estimated according to the Equations (13)- (16). anomalous body. The modeling method is referenced in References [42][43][44]. According to the actual acquisition parameters of LPR [13], the simulated parameters are shown in Table 2. Two sets of forward results with different offsets are obtained in Figure 9b,c. According to the above method, reading from the two sets of radar data, 1 ' t and 2 ' t are shown in      It can also be found that for this kind of complex data the variance of the result is relatively large (the standard deviation is quite large) due to the non-uniformity of the medium, so, as mentioned above, we set a weight based on the reciprocal of the estimated height of each anomalous body as the final result. The estimated final dielectric constant is:   It can also be found that for this kind of complex data the variance of the result is relatively large (the standard deviation is quite large) due to the non-uniformity of the medium, so, as mentioned above, we set a weight based on the reciprocal of the estimated height of each anomalous body as the final result. The estimated final dielectric constant is: This is consistent with the forward complex model (ε r1 = 2.06 ± 0.22).

Result
The Yutu rover released by CE-3 was the first soft landing on the Moon since the Soviet Union's Luna 24 mission in 1976. To be specific, the Yutu rover explored the surface and subsurface of the landing site in the northern part of Mare Imbrium using its four main instruments: The Panoramic Camera, Lunar Penetrating Radar (LPR), Visible-Near Infrared Spectrometer (VNIS) and Active Particle-Induced X-ray Spectrometer (APXS). Its track extends to 114.8 m ( Figure 11) near a young crater. In this part, the data processing results of the LPR are reported.
Aiming at the near-surface stratigraphic structure of the regolith, the CH-2 data is selected. TheCH-2 has two receiver antennas (CH-2A and CH-2B).The LPR data processing pipeline is designed according to the acquisition parameters, the actual situation and the data quality (Table A1 in Appendix A). Two radar images with high resolution (Figure 12) are accessible after data editing and processing. The IDs for the data from Lunar Penetrating Radar are listed in Table A2 in the Supporting Materials.
At first, the reflected waves caused by the same basalt block should be found in CH-2A and CH-2B data ( Figure 13). Reading the t 1 and t 2 of each position follows according to Figure 14, ∆t = 1.1500 ns. Equations (13)-(16) estimate the height and the relative dielectric constant of each anomaly body, which are shown in Table A3.
The Yutu rover released by CE-3 was the first soft landing on the Moon since the Soviet Union's Luna 24 mission in 1976. To be specific, the Yutu rover explored the surface and subsurface of the landing site in the northern part of Mare Imbrium using its four main instruments: The Panoramic Camera, Lunar Penetrating Radar (LPR), Visible-Near Infrared Spectrometer (VNIS) and Active Particle-Induced X-ray Spectrometer (APXS). Its track extends to 114.8 m (Figure 11) near a young crater. In this part, the data processing results of the LPR are reported. Aiming at the near-surface stratigraphic structure of the regolith, the CH-2 data is selected. TheCH-2 has two receiver antennas (CH-2A and CH-2B).The LPR data processing pipeline is designed according to the acquisition parameters, the actual situation and the data quality (Table  A1 in Appendix A). Two radar images with high resolution (Figure 12) are accessible after data editing and processing. The IDs for the data from Lunar Penetrating Radar are listed in Table A2 in the Supporting Materials.
At first, the reflected waves caused by the same basalt block should be found in CH-2A and CH-2B data ( Figure 13). Reading the 1 ' t and 2 ' t of each position follows according to Figure 14 Table A3.
In order to support the conclusion in this paper, deep statistical analysis of the results should be conducted. A histogram of the recovered permittivity is shown in Figure 15.      In order to support the conclusion in this paper, deep statistical analysis of the results should be conducted. A histogram of the recovered permittivity is shown in Figure 15.  However, as with the previous analysis, we believe that if the weight based on the reciprocal of the estimated height of each anomalous body is applied, the estimated dielectric constant is much more accurate: If we only focus on the analysis of the recovered permittivity data (ε in Table A3), the mean value of the estimated dielectric constant is µ = 3.0537, the standard deviation is σ = 0.5923. However, as with the previous analysis, we believe that if the weight based on the reciprocal of the estimated height of each anomalous body is applied, the estimated dielectric constant is much more accurate: and the standard deviation: We believe that the recovered permittivity obeys the normal distribution, the estimated final dielectric constant is ε = 3.0109 ± 1.1538 (confidence level at 95%).
According to the above result, a reasonable dielectric constant result has been given but it is found from Figure 15 that the recovered dielectric constant distribution is not a standard normal distribution and the standard deviation is also relatively large. We consider it is due to the heterogeneity of the regolith medium. The recovered dielectric constant is not only related to the location but also related to the depth. An analysis between permittivity and the depth (or location) of the target should be performed to improve data post-processing. Figure 16a is scatterplot of permittivity and depth and at each meter, we calculate the mean values and the standard deviations (Figure 16b). The figure shows us a relationship between permittivity and depth. There is a maximum value at~2m, which is due to the stratigraphic structure at the CE-3 landing site. At the same way, the analysis between permittivity and the location of LPR has been performed in Figure 17. The permittivity does not change drastically with location.
The lab test results from lunar regolith samples on Apollo and Luna era show that the dielectric constant of lunar regolith is related to the density of lunar regolith [1]. The density is yielded by the dielectric constant as follows: The loss tangent (tan δ) can be easily yielded by the density as follows: and: On the basis of the obtained relative dielectric constant, On the basis of the obtained relative dielectric constant, the content of TiO 2 and FeO on CE-3 landing site were estimated (the results are shown in Figure 18).
After the analysis, we consider the estimated relative permittivity of lunar regolith at 3.0109 and the content of TiO 2 and FeO is 14.0127%, which is consistent with other methods [3,39,40]. dielectric constant is 3.0109 1.1538   (confidence level at 95%).
According to the above result, a reasonable dielectric constant result has been given but it is found from Figure15 that the recovered dielectric constant distribution is not a standard normal distribution and the standard deviation is also relatively large. We consider it is due to the heterogeneity of the regolith medium. The recovered dielectric constant is not only related to the location but also related to the depth. An analysis between permittivity and the depth (or location) of the target should be performed to improve data post-processing. Figure 16a is scatterplot of permittivity and depth and at each meter, we calculate the mean values and the standard deviations (Figure 16b). The figure shows us a relationship between permittivity and depth. There is a maximum value at ~2m, which is due to the stratigraphic structure at the CE-3 landing site. At the same way, the analysis between permittivity and the location of LPR has been performed in Figure 17. The permittivity does not change drastically with location.   According to the above result, a reasonable dielectric constant result has been given but it is found from Figure15 that the recovered dielectric constant distribution is not a standard normal distribution and the standard deviation is also relatively large. We consider it is due to the heterogeneity of the regolith medium. The recovered dielectric constant is not only related to the location but also related to the depth. An analysis between permittivity and the depth (or location) of the target should be performed to improve data post-processing. Figure 16a is scatterplot of permittivity and depth and at each meter, we calculate the mean values and the standard deviations (Figure 16b). The figure shows us a relationship between permittivity and depth. There is a maximum value at ~2m, which is due to the stratigraphic structure at the CE-3 landing site. At the same way, the analysis between permittivity and the location of LPR has been performed in Figure 17. The permittivity does not change drastically with location.
On the basis of the obtained relative dielectric constant, On the basis of the obtained relative dielectric constant, the content of TiO2 and FeO on CE-3 landing site were estimated (the results are shown in Figure 18). After the analysis, we consider the estimated relative permittivity of lunar regolith at 3.0109 and the content of TiO2 and FeO is 14.0127%, which is consistent with other methods [3,39,40].

Discussion
The method for permittivity estimation in this paper has the advantages compared with the method mentioned in the introduction: (1) It is easier to understand and operate. This method uses two sets of data with different offsets to estimate the permittivity. The main work is reading the arrival times of the same anomalous body. It is easier to operate than the methods in the introduction, since the various methods need to extract complex information, such as amplitude, spectrum, phase and so on.
(2) The data quality requirements are lower, even if the radar data contains various types of noise, so as to affect the data waveform, amplitude and frequency, it will not affect the applicability of this method.
(3) No prior information and other data are needed. This method does not require any a priori information, such as incident wave information and does not require other data, such as TDR (Time-Domain Reflectometry) data and so on.

Discussion
The method for permittivity estimation in this paper has the advantages compared with the method mentioned in the introduction: (1) It is easier to understand and operate. This method uses two sets of data with different offsets to estimate the permittivity. The main work is reading the arrival times of the same anomalous body. It is easier to operate than the methods in the introduction, since the various methods need to extract complex information, such as amplitude, spectrum, phase and so on. (2) The data quality requirements are lower, even if the radar data contains various types of noise, so as to affect the data waveform, amplitude and frequency, it will not affect the applicability of this method. (3) No prior information and other data are needed. This method does not require any a priori information, such as incident wave information and does not require other data, such as TDR (Time-Domain Reflectometry) data and so on. (4) It is more suitable for the LPR data collected by CE-3 mission. The original intention of this method is only for the parameter estimation of LPR data. The before methods depend on the low signal-to-noise ratio of the data and need to have all kinds of accurate information, while LPR data is collected on the moon, various types of noise affect the signal-to-noise ratio of the data and deep processing, including gain, filtering, background removal and so forth, has to be done which affected the nature of the data. The method in this paper only needs the information of the arrival time to perform parameter inversion.
However, it is undeniable that this method also has disadvantages compared to other methods. Two sets of data with different offsets are required. It is necessary to extract an accurate wavelet. Only the dielectric constant can be estimated and the conductivity is ignored. This method is only applied in the case of containing various types of scatterers.
Since this method is mainly aimed at the inversion of the lunar soil parameters, it is still a good predictor for the parameters of the lunar regolith.
There are also many factors which will affect the accuracy of the result when using this method for the parameter estimation: (1) Sampling rate. Horizontal and vertical sampling rate will determine the resolution of the radar profile, which will affect the accuracy. (2) Offset. The distances of transmitters and receivers will also affect the result. In general, the larger the offsets are, the more accurate the result will be. (3) The height of the radar. (4) Wavelet extraction. Selecting a good standard wavelet waveform will affect our reading, which will further affect the result. (5) Noise. The signal-to-noise ratio will also affect the accuracy of the data and affect the result of parameter estimation. (6) The heterogeneity of the medium. If the medium is strong non-uniformity, the accuracy of the result will be affected. (7) Personal error. When reading various types of data, personal error cannot be ignored.
An important direction of radar development is single-input multiple-output (SIMO) and multiple-input multiple-output (MIMO), therefore the data with different offsets can be obtained. It is also an important boost for this method which can estimate the position of the abnormal body and the dielectric constant of the medium. In order to avoid the damage to the regolith collector, the Chang'E-5 (CE-5) mission of China will install a MIMO radar on the lander which will help to detect the nature of the regolith and the position of the underlying abnormal rocks on the landing site. The above method will also be able to apply to the CE-5 radar data for the analysis of the stratigraphic structure on the landing site.

Conclusions
According to the geometric propagation law of electromagnetic waves, the propagation paths of two different electromagnetic waves are used to deduce the position of the anomalous body and the electrical parameters of the medium. The method is proved to feasible in theory. When focus on two sets of radar data whose offset are different, the dielectric constant of the medium is inferred by the two different reflection time of a same anomalous body. The process of the parameter estimation method is feasible. A simple model and a complex model of lunar regolith is simulated for the verification of the method. The estimated parameters are consistent with the models.
The LPR data collected by CE-3 mission whose CH-2 data are with different offsets give us the chance to estimate the dielectric constant of regolith on the landing site. The results of the Apollo program show that the dielectric constant is intrinsically linked to the density and the iron-titanium content of regolith. Combined with the estimated dielectric constant from LPR data, the density and the TiO 2 and FeO content are obtained.
Compared with the parameters estimated method of by Apollo's samples, the application of Lunar Penetrating Radar for parameter estimation has a larger range and lower cost. Compared with the remote sensing estimation, the LPR can be more accurate. The parameters of regolith estimated by LPR data help us predict the reserves of resources at the CE-3 landing site and even the entire Mare Imbrium.

Conflicts of Interest:
The authors declare no conflict of interest.  A   Table A1. Explanation of data processing.

Processing Explanation
i Data reading Data is divided into 9 sections. According to their storage format (*.psd, the standard storage format in aviation and spaceflight field), the data and location information are read one by one.
ii Data stitching Data is divided into 9 sections, which should be spliced into a group.
iii Traces amending Not only the differ of section but also the error of collection start time can make the Image longitudinal displacement. Based on the phase of a strong reflection event, we adjust the traces.
iv Traces selecting The rover patrolled with non-uniform motion, since the rover might stop at some points on the way to collect other scientific data such as APXS data, VNIS data and so on. However, LPR never stop acquisition, which resulted in repeated acquisition of multiple traces at same location. Stack and average the repeated traces. The first and second sections with low SNR (Signal to Noise Ratio) should be deleted.
v Time lag adjustment The arrival time of the radar echo was delayed by 28.203 ns corresponding to the start time for recording data.
vi Useless data deleting The first 150 ns data with great research value need to be aimed in, since the rest of data has low SNR.
vii Band-pass filter Eliminate noise by band-pass filter.
viii Background removal In order to highlight the abnormal information, subtract the average trace.
ix Automatic gain control AGC (Automatic Gain Control) has been to observe the bottom information.
x Re-positioning Add location information to the image.