Laser-radar-cloud-radar-based composite precipitation cloud recognition method

Abstract

The application discloses a composite precipitation cloud layer identification method based on a laser radar-cloud radar, and comprises the following steps: acquiring observation data of the laser radar and the cloud radar, and performing data matching on the observation data; processing signals of the laser radar in a 18-20km height interval to obtain preprocessed laser radar signals; performing laser radar cloud layer boundary identification based on a VED algorithm, and giving different laser radar cloud scores according to signal attenuation rates and signal gradients; removing ground clutter by using a threshold method based on spectrum width and Doppler velocity observation; performing cloud radar significant echo identification based on the threshold method, and determining cloud radar cloud score sizes according to different reflectivity factor signal strengths; determining light rain and heavy rain areas according to Doppler velocity sizes; obtaining cloud layer boundary information and drizzle areas through cloud layer identification result fusion, and completing cloud layer identification and precipitation area identification; and the application improves the accuracy of weather prediction.

Description

Technical Field

[0001] This invention belongs to the field of atmospheric remote sensing data inversion technology, and particularly relates to a method for identifying composite precipitation clouds based on lidar-cloud radar. Background Technology

[0002] Raindrops exhibit optical characteristics similar to clouds as they fall to the Earth's surface. Due to limited observational data, past cloud identification studies often excluded rainfall or simply ignored it. However, this approach frequently leads to significant errors in obtaining cloud microphysical parameters. For instance, in microwave radiometer (MWR) cloud water path inversion algorithms, the cloud base information, which is often lower than the actual cloud depth, tends to result in a higher content of cloud water paths derived from the MWR, because it includes information about liquid water paths. Therefore, accurately removing rainfall areas is crucial for precisely inverting cloud macro and micro parameters in cloud identification.

[0003] Currently, lidar is limited by signal attenuation and decreased long-distance detection accuracy caused by severe weather (such as rain, snow, and fog), making it difficult to stably acquire high-resolution three-dimensional cloud structures. Doppler radar suffers from insufficient resolution for detecting small particles (cloud droplets) due to its long wavelength and is susceptible to interference from turbulence and vertical wind shear, limiting the inversion of microphysical parameters. Furthermore, existing cloud macro- and micro parameter inversion technologies mostly focus on single technologies and lack collaborative identification methods that integrate multi-source data, thus hindering the breakthrough development of refined cloud monitoring and weather early warning.

[0004] In view of the above problems, this application utilizes the different scattering characteristics of visible light and millimeter waves in clouds and rain, and adopts a lidar-cloud radar composite precipitation identification method to obtain rich information on the vertical distribution of clouds and rain. This method has the characteristics of identifying liquid precipitation and improving cloud identification ability. Compared with single millimeter wave cloud radar cloud identification, this method has a more reliable cloud base identification capability and can detect high-level ice clouds and drizzle. Compared with single lidar cloud identification, this algorithm can effectively penetrate low-level clouds to achieve high-level cloud detection and can identify light rain and heavy rain. Summary of the Invention

[0005] To address the problems existing in the background technology, the present invention aims to propose a composite precipitation cloud identification method based on lidar-cloud radar, which can provide multi-dimensional feature parameters for accurate identification of precipitation patterns, solve technical problems such as confusion between drizzle and cloud droplets, lidar failure caused by heavy rain signal attenuation, and interference of rainfall with cloud microphysical parameter inversion, and provide highly reliable data support for meteorological observation, climate simulation and weather modification.

[0006] To address the aforementioned technical issues, this invention proposes a composite precipitation cloud identification method based on lidar and cloud radar. The method leverages the fact that drizzle regions, like clouds, exhibit significant echoes in the millimeter-wave echo profiles obtained by cloud radar, while lidar possesses certain penetrating capabilities. By comparing the cloud identification results from both methods and combining them with the Doppler velocity's ability to distinguish between heavy and light rain, more effective precipitation area identification can be achieved, while simultaneously obtaining reliable cloud information. Furthermore, by fusing lidar and millimeter-wave cloud radar, this invention constructs an altitude-layered cloud-rain dynamic identification system, enabling accurate differentiation between drizzle and cloud droplets, suppression of heavy rain signal attenuation, and reconstruction of cloud boundaries to eliminate rainfall interference.

[0007] The present invention adopts the following technical solution:

[0008] A method for identifying composite precipitation clouds based on lidar-cloud radar, comprising the following steps:

[0009] Acquire observation data from lidar and cloud radar, and perform data matching to ensure that the temporal and spatial resolutions of the lidar and cloud radar observation data are consistent.

[0010] The lidar signal in the altitude range of 18-20km is processed to obtain the preprocessed lidar signal;

[0011] LiDAR cloud boundary identification is performed based on the VED algorithm, and different LiDAR cloud scores are assigned according to the signal attenuation rate and signal gradient.

[0012] Based on spectral width and Doppler velocity observations, a thresholding method was used to remove surface clutter.

[0013] Significant echo identification of cloud radar is performed based on the threshold method, and the cloud radar cloud score is determined according to the signal intensity of different reflectivity factors.

[0014] The areas of light and heavy rain are determined based on the magnitude of the Doppler velocity.

[0015] By fusing cloud identification results, cloud boundary information and drizzle areas are obtained, and cloud identification and precipitation area identification are completed.

[0016] Furthermore, the method for assigning different lidar cloud scores based on signal attenuation rate and signal gradient is as follows:

[0017]

[0018]

[0019] The formula for calculating the signal attenuation rate is:

[0020]

[0021] Where F(z) is the signal attenuation rate; z is the observation distance; and P(z) is the signal of the lidar at a distance z from the instrument.

[0022] The maximum signal attenuation T and the minimum signal attenuation D are obtained by using the signal attenuation rate calculation formula.

[0023] Furthermore, based on spectral width and Doppler velocity observations, the thresholding method for removing surface clutter is as follows:

[0024] When the spectral width is less than 0.3 m / s, the absolute value of the Doppler velocity is greater than 0.2 m / s, and the cloud base height is less than 1.5 km, it is judged as surface clutter and is directly removed.

[0025] Furthermore, based on the threshold method for identifying significant echoes from cloud radar, the method for determining the cloud radar cloud score based on different reflectivity factor signal intensities is as follows:

[0026] When the reflectivity factor is greater than -35dBZ, the cloud radar score is assigned 50; when the reflectivity factor is greater than 0dBZ, the maximum cloud radar score is assigned 100; when the reflectivity factor is between -35dBZ and 0dBZ, the cloud radar score changes linearly.

[0027] Furthermore, the method for determining light rain and heavy rain areas based on the magnitude of Doppler velocity is as follows: when the Doppler velocity is <-3.5m / s, it is identified as light rain or heavy rain.

[0028] Furthermore, by fusing cloud identification results, cloud boundary information and drizzle areas are obtained, and the method for completing cloud identification and precipitation area identification is as follows:

[0029] When the first cloud base detected by the lidar is higher than the first cloud base detected by the cloud radar, the area between the two cloud bases is identified as drizzle.

[0030] Based on the cloud base information identified by the lidar, the lidar cloud score and the cloud radar cloud score are added together and averaged to obtain the total cloud score. When the total cloud score is greater than 60, the location is identified as having clouds.

[0031] Furthermore, the method for processing the lidar signal in the 18-20km altitude range to obtain the preprocessed lidar signal is as follows:

[0032] The observation signals of the lidar pole in the altitude range of 18-20km are regarded as observation results with only background signals. The mean value of background noise is calculated and then subtracted. After subtracting the mean value of background noise from the original lidar signal, a 3-point moving average method is used for smoothing to suppress random noise.

[0033] Furthermore, data matching is performed on the observation data to ensure that the temporal and spatial resolutions of the observation data from the lidar and cloud radar are consistent. In this case, the temporal and spatial resolutions are based on the observation data from the cloud radar.

[0034] The beneficial technical effects of this invention are as follows:

[0035] (1) This invention proposes a fusion scheme for radar data from different sources and with different resolutions, which makes full use of the advantages of the two radars. By combining Doppler velocity with the gradient of lidar signal, it distinguishes between drizzle and light rain / heavy rain, eliminates the interference of rainfall on cloud identification, and reconstructs the interference-free three-dimensional boundary structure of the cloud.

[0036] (2) This invention significantly improves the accuracy of cloud microphysical parameter inversion. It integrates millimeter-wave reflectivity layered weighting with lidar VED gradient model to suppress the influence of heavy rain signal attenuation and improve the accuracy of cloud water path and other parameters. It effectively solves the problem of detection blind zone of single sensor in weak precipitation signal identification and strong attenuation environment, improves the accuracy of meteorological forecasting, and provides data support for aviation safety assessment, maritime operation decision-making and artificial weather modification. Attached Figure Description

[0037] Figure 1 A schematic diagram of a lidar signal provided in an embodiment of the present invention;

[0038] Figure 2 A schematic diagram of cloud radar reflectivity factor provided in an embodiment of the present invention;

[0039] Figure 3 This is a schematic diagram of cloud and rainfall identification provided in an embodiment of the present invention. Detailed Implementation

[0040] The method for identifying composite precipitation clouds based on lidar-cloud radar provided by the present invention will be further described clearly and completely below with reference to the accompanying drawings:

[0041] This embodiment provides a method for identifying composite precipitation clouds based on lidar-cloud radar, including the following steps:

[0042] S1: Acquire observation data from lidar and cloud radar, and perform data matching to ensure that the temporal and spatial resolutions of the lidar and cloud radar observation data are consistent; the temporal and spatial resolutions are based on the cloud radar observation data.

[0043] S2: Treat the observation signal at the far distance of the lidar as the observation result with only background signal, calculate the mean value of background noise, subtract the mean value of background noise from the original signal and smooth it to obtain the preprocessed lidar signal.

[0044] Specifically, the signal from the lidar at an altitude of 18-20km (this distance range is at the end of the laser signal range, and the signal is relatively clean) is extracted, the mean background noise is calculated, and then subtracted. It should be noted that the subtracted mean background noise comes from the background noise of sunlight, and the remaining noise consists of random noise from the instrument, signals from atmospheric molecules, and signals from clouds and aerosols. After subtracting the mean background noise from the original lidar signal, a 3-point moving average method is used for smoothing to suppress random noise.

[0045] The formula for calculating the mean background noise is as follows:

[0046] P noise =Mean(P(z)) 20>z>18

[0047] Among them, P noise Represents the background noise from sunlight; z is the observation distance, which can be in kilometers; P(z) is the signal of the lidar at a distance of z kilometers from the instrument; Mean() represents taking the average value.

[0048] S3: LiDAR cloud boundary identification based on VED algorithm. This method, proposed by Zhao et al. (2014), aims to solve the problem of increasing signal noise with distance in traditional LiDAR. The specific process includes: First, suppressing noise influence through semi-discrete processing (SDP) technology. This technology is based on noise estimation (e.g., the standard deviation of background noise multiplied by an empirical coefficient K=3) and combined with finite moving average (using three-point average when vertical resolution > 0.03km, otherwise averaging within a 0.06km window) to preserve the sharp change characteristics of the signal (e.g., cloud boundaries), generating the processed signal PD(z); Second, reorganizing the signal values ​​using the value distribution equalization (VDE) method to equalize their distribution (by calculating the mapping scaling factor PE(i) and generating a new dataset PN(z)), thereby reducing the signal strength attenuation with distance and highlighting the signal deviation of the cloud / aerosol layer; Finally, identifying the particle layer boundary based on the deviation between the equalized signal and the baseline B(z) (with the minimum layer depth threshold set to 45m);

[0049] Then, different LiDAR cloud scores are assigned based on the signal attenuation rate and signal gradient. It should be noted that at low altitudes, the identified feature layer is more likely to be aerosols rather than clouds; therefore, the cloud base height limit is increased to help eliminate the influence of aerosols.

[0050] The formula for calculating the signal attenuation rate is:

[0051]

[0052] Where F(z) is the signal attenuation rate, z is the observation distance (in kilometers), and P(z) is the signal of the lidar at a distance z kilometers from the instrument.

[0053] The maximum signal attenuation T and minimum signal attenuation D are obtained using the signal attenuation rate calculation formula. The rules for assigning LiDAR cloud scores after obtaining the feature layer are as follows:

[0054] <1.5km T>12 and D<-18 80 <1.5km T>12 or D<-18 60 <1.5km other 30 1.5km-5km T>6 and D<-15 90 1.5km-5km T>4 and D<-9 70 1.5km-5km T>4 or D<-9 50 >5km T>4 and D<-9 90 >5km T>1.5 and D<-9 70 >5km T>1.5 or D<-9 50

[0055] S4: Based on spectral width and Doppler velocity observations, a threshold method is used to remove surface clutter; specifically, when the spectral width is less than 0.3 m / s, the absolute value of the Doppler velocity is greater than 0.2 m / s, and the cloud base height is less than 1.5 km, it is judged as surface clutter and is directly removed;

[0056] S5: Significant echoes from cloud radar are identified based on a threshold method. The cloud radar score is determined according to the different intensities of the reflectivity factor (dBZ). Specifically, when the reflectivity factor is greater than -35dBZ, a cloud radar score of 50 is assigned (indicating that the signal strength has reached the lower threshold for cloud detection, with medium confidence); when the reflectivity factor is greater than 0dBZ, the maximum cloud radar score of 100 is assigned (indicating that the signal strength is relatively strong, with high confidence); when the reflectivity factor is between -35dBZ and 0dBZ, the cloud radar score changes linearly (for example, calculated using the formula cloud radar score = 50 + 50 × [(Z + 35) / 35], where Z is the reflectivity factor value).

[0057] The principle behind this assignment is based on the typical cloud detection characteristics of millimeter-wave cloud radar (MMCR) in the Atmospheric Radiation Measurement (ARM) program: the reflectivity factor is typically within a dynamic range of -50 dBZ to 20 dBZ, and the lower limit threshold for cloud signal detection is often set at around -40 dBZ to -30 dBZ (for example, Clothiaux et al., 2000 reported that MMCR can accurately detect cloud signals from -50 dBZ to -20 dBZ, but in actual cloud masking algorithms, -40 dBZ is often used as the weak cloud threshold to distinguish noise). Here, -35 dBZ is selected as the starting threshold to reflect the edge signal of weak or thin clouds (below this value, it is mostly noise or non-cloud signals); 0 dBZ corresponds to stronger cloud echoes, often associated with thick clouds or clouds with high water content, and has a higher confidence in cloud presence. This linearly gradual assignment method provides a continuous cloud radar cloud score assessment, simulating the positive correlation between signal strength and cloud optical thickness or confidence, avoiding the limitations of binary classification.

[0058] S6: Determine the areas of light rain and heavy rain based on the magnitude of Doppler velocity; specifically, when the Doppler velocity is <-3.5 m / s, it is identified as light rain or heavy rain. It should be noted that the determination of this threshold <-3.5 m / s is based on the practical experience of millimeter-wave cloud radar in typical precipitation detection. In this case, negative values ​​of Doppler velocity indicate that the particles are moving downwards (falling velocity), and positive values ​​indicate that they are moving upwards. This notation convention is common in cloud radar data processing (refer to Chandra et al., 2014). The principle is that the terminal falling velocity of cloud particles (such as cloud droplets or ice crystals) is usually small (<-3.5 m / s is used as an empirical threshold to identify particle layers with significant falling velocities), indicating that there may be light rain (falling velocity range of about 3.5-5 m / s, corresponding to smaller raindrops) or heavy rain (falling velocity range >5 m / s, corresponding to larger raindrops or heavy precipitation).

[0059] S7: Based on cloud scores, cloud identification results are fused to obtain cloud boundary information and drizzle areas, and cloud identification and precipitation area identification are completed. Specifically, when the first cloud base identified by the lidar (here, the first cloud base is the first layer identified; the algorithm processes it line by line, and several cloud layers may be identified in each profile; the first cloud base refers to the cloud layer closest to the instrument) is higher than the first cloud base identified by the cloud radar, the area between the two cloud bases is identified as a drizzle area. Combined with the precipitation identification in step 6, the precipitation area identification in the observation direction is completed.

[0060] Based on the cloud base information identified by the lidar, the lidar cloud score and the cloud radar cloud score are added together and the average value is obtained to get the total cloud score. When the total cloud score is greater than 60, the location is identified as having clouds.

[0061] As an example, in this embodiment, observation data from lidar and cloud radar are acquired. The lidar has a temporal resolution of 1 minute and a spatial resolution of 15 meters; the cloud radar has a temporal resolution of 1 minute and a spatial resolution of 30 meters; the observation period is from 18:00 to 24:00 on April 5, 2024. Data matching involves resampling the cloud radar data to a 1-minute resolution; interpolating the lidar spatial resolution to 30 meters to match the cloud radar resolution, resulting in joint observation data with a unified spatiotemporal resolution; and extracting lidar data... For signals reaching altitudes of 18-20 km, the mean background noise was calculated and subtracted. After subtracting the mean background noise from the original lidar signal, a three-point moving average method was used for smoothing to suppress random noise. For the pre-processed lidar signal, the signal attenuation rate F(z) was calculated, and feature layers were extracted. In a certain profile, the cloud base height was approximately 4.1 km, T = 7.2, and D = -9.2. Based on the cloud fraction assignment standard, a cloud fraction of 70 was assigned. Increased reflectivity was observed near 0.2 km, with an absolute Doppler velocity of 0.5. m / s, spectral width 0.2m / s; meets the surface clutter judgment criteria, the reflectivity at this height is assigned the lowest background value (-35dBZ) to remove clutter influence; within the significant echo height range of cloud radar, the cloud radar reflectivity factor distribution is as follows: reflectivity at 2.8km is approximately -14dBZ; reflectivity at 4.5km is approximately -24dBZ; cloud fraction is determined according to different intensities of reflectivity factor (dBZ): cloud fraction is 80 at -14dBZ; cloud fraction is 66 at -24dBZ; after 22:00, the identification threshold is high. Within the specified range, the Doppler velocity observation value is approximately -5.2 m / s, meeting the precipitation identification criteria; therefore, precipitation is determined to exist in this area. Cloud layer identification results are fused with drizzle determination: the first cloud base identified by the lidar is located at 4.1 km, while the cloud base identified by the cloud radar is 0.0 km; therefore, the cloud base height range is determined to be a drizzle area based on the difference. During the fusion process, cloud scores are superimposed: lidar cloud score 70; cloud radar cloud score 60; fused cloud score = 65; because the fused cloud score > 60, clouds are determined to exist above 3.2 km.

[0062] It should be noted that during observation, only one-dimensional vertical observation results can be obtained at each time point, therefore Figures 1-3 It is composed of a series of outlines pieced together; Figure 1 The data is from lidar observations between 18:00 and 24:00. Cloud base signals at 5km can be clearly observed from 18:00 to 24:00. Lidar signals are not sensitive to weak rainfall. Figure 2 The signal is from the cloud radar observation between 18:00 and 24:00. The signal indicates that rainfall occurred between 20:00 and 24:00. The cloud radar signal cannot distinguish between clouds and rainfall. Figure 3By using two instruments for cloud identification, the cloud base can be identified even during rainfall. The following information was obtained: Rain began at 20:30 as drizzle; at 21:30, it turned into light to heavy rain; at 23:00, it turned back into drizzle; during this period, the cloud base was approximately 4.1 km.

[0063] In summary, this invention provides a data fusion method for lidar (1064nm) and millimeter-wave radar (Ka-band), and overcomes the limitations of a single sensor by inverting cloud boundaries based on the fused data; it considers the influence of background noise and can remove background noise from signal characteristics at extremely long distances (18-20Km) to improve accuracy; and it achieves accurate differentiation between drizzle and cloud droplets by dynamically assigning cloud fractions based on Doppler velocity threshold, lidar signal gradient, and attenuation rate.

[0064] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for identifying composite precipitation clouds based on lidar-cloud radar, characterized in that, The method includes the following steps: Acquire observation data from lidar and cloud radar, and perform data matching to ensure that the temporal and spatial resolutions of the lidar and cloud radar observation data are consistent. The lidar signal in the altitude range of 18-20km is processed to obtain the preprocessed lidar signal; LiDAR cloud boundary identification is performed based on the VED algorithm, and different LiDAR cloud scores are assigned according to the signal attenuation rate and signal gradient. Based on spectral width and Doppler velocity observations, a thresholding method was used to remove surface clutter. Significant echo identification of cloud radar is performed based on the threshold method, and the cloud radar cloud score is determined according to the signal intensity of different reflectivity factors. The areas of light and heavy rain are determined based on the magnitude of the Doppler velocity. By fusing cloud layer recognition results, cloud boundary information and drizzle areas are obtained, and cloud layer recognition and precipitation area recognition are completed. The method for obtaining cloud boundary information and drizzle areas by fusing cloud identification results, and completing cloud identification and precipitation area identification, is as follows: When the first cloud base detected by the lidar is higher than the first cloud base detected by the cloud radar, the area between the two cloud bases is identified as drizzle. Based on the cloud base information identified by the lidar, the lidar cloud score and the cloud radar cloud score are added together and averaged to obtain the total cloud score. When the total cloud score is greater than 60, the location is identified as having clouds.

2. The method for identifying composite precipitation clouds based on lidar-cloud radar according to claim 1, characterized in that: The method for assigning different lidar cloud scores based on signal attenuation rate and signal gradient is as follows: The formula for calculating the signal attenuation rate is: in, z is the signal attenuation rate; z is the observation distance; The signal from the lidar at a distance z from the instrument; The maximum signal attenuation T and the minimum signal attenuation D are obtained by using the signal attenuation rate calculation formula.

3. The method for identifying composite precipitation clouds based on lidar-cloud radar according to claim 1, characterized in that: Based on spectral width and Doppler velocity observations, the thresholding method for removing surface clutter is as follows: When the spectral width is less than 0.3 m / s, the absolute value of the Doppler velocity is greater than 0.2 m / s, and the cloud base height is less than 1.5 km, it is judged as surface clutter and is directly removed.

4. The method for identifying composite precipitation clouds based on lidar-cloud radar according to claim 1, characterized in that: The method for identifying significant echoes from cloud radar based on a threshold method, and determining the cloud radar cloud score based on different reflectivity factors and signal intensities, is as follows: When the reflectivity factor is greater than -35 dBZ, the cloud radar score is assigned 50; when the reflectivity factor is greater than 0 dBZ, the maximum cloud radar score is assigned 100; when the reflectivity factor is between -35 dBZ and 0 dBZ, the cloud radar score changes linearly.

5. The method for identifying composite precipitation clouds based on lidar-cloud radar according to claim 1, characterized in that: The method for determining light rain and heavy rain areas based on the magnitude of Doppler velocity is as follows: when the Doppler velocity is less than -3.5 m / s, it is identified as light rain or heavy rain.

6. The method for identifying composite precipitation clouds based on lidar-cloud radar according to claim 1, characterized in that: The method for processing the lidar signal in the 18-20km altitude range to obtain the preprocessed lidar signal is as follows: The observation signals of the lidar in the altitude range of 18-20km are regarded as observation results with only background signals. The mean value of background noise is calculated and then subtracted. After subtracting the mean value of background noise from the original lidar signal, a 3-point moving average method is used for smoothing to suppress random noise.

7. The method for identifying composite precipitation clouds based on lidar-cloud radar according to claim 1, characterized in that: When performing data matching on the observation data to ensure that the temporal and spatial resolutions of the observation data from the lidar and cloud radar are consistent, the temporal and spatial resolutions are based on the observation data from the cloud radar.

Images