The Added Value of Surface Data to Radar-Derived Rainfall-Rate Estimation Using an Artificial Neural Network

Precipitation plays a crucial role to the global water and energy cycle that governs the weather, climate, and ecological systems. Thorough understanding and accurate forecasting of precipitation is essential to the affairs of humans. The Tropical Rainfall Measuring Mission (TRMM), launched in 1997, is a joint space mission between NASA and the Japan Aerospace Exploration Agency (JAXA) designed to monitor and study tropical rainfall. The Precipitation Radar (PR) on board the TRMM satellite is the first space borne instrument, capable of providing high-resolution vertical profile of precipitation on a global scale. TRMM-PR operates at a single frequency of Ku- (13.6 GHz) band. The microphysical retrieval algorithms for TRMM-PR rely on the surface-reference technique (SRT) to estimate path attenuation and correct the measured Ku-band reflectivity. With the attenuation-corrected reflectivities, a modified Hitschfeld-Bordan method [1] is then used to retrieve limited drop size distribution (DSD) information, and the rainfall rate [2]. One disadvantage of single-frequency space borne radar such as TRMM-PR is that it is not easy to retrieve the DSD parameters completely. Therefore, k-Z and Z-R relationships, with their inherent assumptions, are used to estimate rainfall rate which is not sufficient to capture the variability of precipitation and has large uncertainty. Global Precipitation Measurement (GPM) mission is poised to be the next generation observations from space after the TRMM mission. GPM is a science mission with integrated applications goals for advancing the knowledge of the global water/energy cycle variability as well as improving weather, climate, and hydrological prediction capabilities through more accurate and frequent measurements of global precipitation. The GPM core satellite will be equipped with a dual-frequency precipitation radar (DPR) operating at Ku- (13.6 GHz) and Ka- (35.5 GHz) band [3]. Taking two independent sets of observation, DPR on aboard the GPM is expected to improve our knowledge of precipitation processes relative to the single-frequency (Ku- band) radar used in TRMM by providing greater dynamic range, more detailed information on microphysics. Two parameters of DSDs can be retrieved from dual-frequency observations and better accuracies in rainfall estimation can be achieved. Theoretically, rainfall rate is a function of rain drop size distribution and rain drop terminal velocity, R=0.67π∗10−3∫ v(D)D3N(D)dD. The most critical component in rainfall rate estimation is the time-space variation of drop size distribution. Le and Chandrasekar (2014) [4] developed a hybrid method to retrieve drop size distribution parameters for GPM-DPR. The hybrid method is a profile-based optimization algorithm with the philosophy to combine the attributes of forward method and linear constraints of DSDs in rain. Two of the gamma distribution parameters [5], Do and Nw, at surface are optimized when the deviation between estimates and observations are minimized. The hybrid method can be used to estimate DSDs at each space and temporal resolution of GPM-DPR observation. In this paper, rainfall rate is calculated using DSDs retrieved through the hybrid method [4] based on assumptions of particle terminal velocity. Data collected by GPM-DPR is capability to cover ±65° latitude of the earth with every 2–4 hours. Thus, a global rainfall map can be generated. In polarimetric radar system, rainfall rate can be estimated through dual-polarized radar parameter such as Zdr [6][7]. Zdr is called differential reflectivity and it is a function of particle characteristics itself. Similar of Zdr to the dual-polarization radar retrieval, there is a parameter called dual-frequency ratio (DFR) that plays an important role in the dual-frequency radar retrievals. DFR is defined as the difference of the radar reflectivity at two frequencies in decibels which carries information of single particle characteristics. In this study, we investigate potential relation between rainfall rate, reflectivity at Ku- band and DFR using theoretical simulation [8] and curve fitting. Figure 1 shows a scattergram of rainfall rate estimation from R(Zh, DFR) versus true rainfall rate. Both Zku and DFR are in linear form in this equation. This provides a direct estimation of rainfall rate without prior knowledge of drop size distribution. Since both reflectivity and dual frequency ratio are intrinsic values, attenuation correction is needed before rain rate can be estimated through R(Zh, DFR) equation. The hybrid method [4] described in previous paragraph could be one of the algorithms to perform attenuation correction. The global rainfall map generated through R(Zh, DFR) approach will be cross-compared with DSD approach described in the previous paragraph, as well as other approaches such as attenuation based rainfall rate. The comprehensive analysis of the three techniques are presented.

High Temporal Resolution Rainfall Information Retrieval from Tipping-bucket Rain Gauge Measurements

This paper describes a recent development in rainfall estimation using satellite-flown and ground-based radars. The Tropical Rainfall Measuring Mission (TRMM) Precipitation Radar (PR), its algorithms and data processing are discussed. The ground validation algorithms and processing of the ground-based radar reflectivity data are explained. The estimates of attenuation-corrected radar reflectivity factor and rainfall rate are given at each resolution cell of the PR. The estimated near-surface rainfall rate and average rainfall rate at the altitudes of 2 km are calculated for each beam position. The TRMM PR profiling algorithm and processing of the PR reflectivity for rain distribution are explained. The TRMM rain products and their geophysical parameters are derived from the measurements from the satellite and ground-based radar. The derived geophysical parameters include vertical rain and hydrometeor profile, rain type, radar back-scatter cross-section, raindrop size distribution, rain gauge rain rates and 5-day and monthly average rain rates. For validation purposes the instantaneous and climatological comparison of the rain estimates from both the Precipitation Radar and ground-based radar at Melbourne, Florida, was carried out on the basis of rain type; i.e. convective/stratiform, vertical structure and rain maps. The error sources in rain profile retrieval from space-borne radar; i.e. the PR and ground-based radar with their algorithm limitations are discussed. A second set of data, this time for an area where no simultaneous ground data are available has also been analysed; the data were chosen for the three-dimensional rain distribution over some parts of India. The issues such as discrimination of rain from surface clutter, calibration accuracy and sensitivity of precipitation radar and discrimination of rain echo from noise are discussed.

The use of meteorological radar reflectivity Z to estimate rainfall rate R is approached using a different perspective from the classical "Z-R relation". Simultaneous rain measurements from different sensors are combined to construct a model that estimates the vertical air velocity by minimizing the error in reflectivity between the different sensors. This model is based on the fact that rain rate and reflectivity are both dependent on the integrals of rain drop size distribution (DSD) but only R depends on vertical air velocity. This study attempts to validate the vertical air velocity estimates and quantify their affects on the rainfall rate estimation. Disdrometer Flux Conservation Model (DFC) uses measurements from disdrometers and other sensors such as vertically pointing radar profilers and scanning radars. Disdrometers measure a drop size flux Φ(D), defined as the number of drops passing a horizontal surface per unit time, per unit area, per drop size. The flux is equal to the product of the drop size distribution near the ground NG(D) and drop velocity near the ground vG(D). The drop velocity is the difference between the droplet terminal velocity and the vertical component of the wind velocity, which varies with altitude. The estimates derived from the DFC model using two pair wise selected sensors are used to study the change of reflectivity and vertical air velocity with altitude. Sensitivity tests for the DFC model are also discussed and these outcomes are validated by comparison with independent profiler vertical velocity observations.

The effect of multiple scattering on the estimation of rainfall rates using incoherent backscattered power at two frequencies (16 and 34.8 GHz) is analyzed. Simulation of circularly polarized transmissions propagating through homogeneous rain is used to estimate the rainfall rates from echo power levels calculated with and without multiple scattering effects included. The results indicate that in homogeneous rain, the rainfall rate estimated using the multiple scattering calculation is always smaller than that using the conventional radar equation. The difference is significant for heavy rain cases. Since the multiple scattering process may exist in real rain situations, one must be careful in heavy rain cases when interpreting rainfall rates estimated using dual-wavelength algorithms. >

Vertically pointed wind profiling radars can be used to obtain measurements of the underlying drop size distribution (DSD) for a rain event by means of the Doppler velocity spectrum. Precipitation parameters such as rainfall rate, radar reflectivity factor, liquid water content, mass-weighted mean drop diameter, and median volume drop diameter can then be calculated from the retrieved DSD. The DSD retrieval process is complicated by the presence of atmospheric turbulence, vertical ambient air motion, selection of fall speed relationships, and velocity thresholding. In this note, error analysis is presented to quantify the effect of each of those factors on rainfall rate. The error analysis results are then applied to two precipitation events to better interpret the rainfall-rate retrievals. It was found that a large source of error in rain rate is due to unaccounted-for vertical air motion. For example, in stratiform rain with a rainfall rate of R = 10 mm h−1, a mesoscale downdraft of 0.6 m s−1 can result in a 34% underestimation of the estimated value of R. The fall speed relationship selection and source of air density information both caused negligible errors. Errors due to velocity thresholding become more important in the presence of significant contamination near 0 m s−1, such as ground clutter. If particles having an equivalent volume diameter of 0.8 mm and smaller are rejected, rainfall rate errors from −4% to −10% are possible, although these estimates depend on DSD and rainfall rate.

The selection of appropriate function forms linking the radar-based polarimetric variables and quantitative precipitation estimates (QPEs) is critical for the estimation of precipitation accurately. In this study, several functional forms using nonlinear optimization and artificial neural networks are evaluated. The polarimetric variables including radar reflectivity factor (ZH), differential reflectivity factor (ZDR), and specific differential propagation phase (KDP) values are obtained from an X-band multiparameter radar (XMPR) recently installed in a Yogyakarta region, Indonesia. The region experiences highly variable rainfall with mostly frontal rainfall events confined to December to March. Observed precipitation data available from a sparse rain network in this region is used for comparative evaluation of the rainfall rates. Rain gauge data for three years (2016-2018) are used for the assessment. Results from the analysis suggest that improved estimates of rainfall at finer temporal resolutions of 10 minutes can be obtained using functional forms developed with the optimization of polarimetric thresholds when adequate observed rainfall data and polarimetric variables are available. Furthermore, the use of the multiple linear, stepwise regression model and neural networks approach can measure the sensitivity of the functional forms used for rainfall rate estimates. Local models based on specific temporal windows (e.g., month or a season) can help improve the functional forms and therefore rainfall rate estimates.

Semi-empirical characteristics of modified lognormal DSD inputs using rain rate distributions for radio links over the African continent

A technique for model initialization using three-dimensional radar reflectivity data has been developed and applied within the NOAA 13-km Rapid Refresh (RAP) and 3-km High-Resolution Rapid Refresh (HRRR) regional forecast systems. This technique enabled the first assimilation of radar reflectivity data for operational NOAA forecast models, critical especially for more accurate short-range prediction of convective storms. For the RAP, the technique uses a diabatic digital filter initialization (DFI) procedure originally deployed to control initial inertial gravity wave noise. Within the forward-model integration portion of diabatic DFI, temperature tendencies obtained from the model cloud/precipitation processes are replaced by specified latent heating–based temperature tendencies derived from the three-dimensional radar reflectivity data, where available. To further refine initial conditions for the convection-allowing HRRR model, a similar procedure is used in the HRRR, but without DFI. Both of these procedures, together called the “Radar-LHI” (latent heating initialization) technique, have been essential for initialization of ongoing precipitation systems, especially convective systems, within all NOAA operational versions of the 13-km RAP and 3-km HRRR models extending through the latest implementation upgrade at NCEP in 2020. Application of the latent heat–derived temperature tendency induces a vertical circulation with low-level convergence and upper-level divergence in precipitation systems. Retrospective tests of the Radar-LHI technique show significant improvement in short-range (0–6 h) precipitation system forecasts, as revealed by reflectivity verification scores. Results presented document the impact on HRRR reflectivity forecasts of the radar reflectivity initialization technique applied to the RAP alone, HRRR alone, and both the RAP and HRRR. Significance Statement The large forecast uncertainty of convective situations, even at short lead times, coupled with the hazardous weather they produce, makes convective storm prediction one of the most significant short-range forecast challenges confronting the operational numerical weather prediction community. Prediction of heavy precipitation events also requires accurate initialization of precipitation systems. An innovative assimilation technique using radar reflectivity data to initialize NOAA operational weather prediction models is described. This technique, which uses latent heating specified from radar reflectivity (and can accommodate lightning data and other convection/precipitation indicators), was first implemented in 2009 at NOAA/NCEP and continues to be used in 2022 in the NCEP-operational RAP and HRRR models, making it a backbone of the NOAA rapidly updated numerical weather prediction capability.

The presence of undetected mixed-phase precipitation or superimposed intense clutter can cause serious errors in the estimation of rainfall rate and other parameters of precipitation occurring in the radar scattering volume. To reduce or avoid these errors it is necessary to distinguish between the rain echo, and that due to other types of precipitation, and between precipitation radar echoes and ground clutter. Multiparameter radar measurements may be exploited for this discrimination. In particular, it is demonstrated that dual-linear polarization measurements may play a major role in this process. Sample radar data are employed to illustrate several different tests to classify radar data: the results refer to comparisons of dual-polarized echoes due to precipitation (rain or mixed-phase event) and land with echoes from land alone. This is illustrated by example through the application of a series of tests on a clutter-contaminated dual-polarized dataset obtained during the May Polarization Experiments (MAYPOLE) 1984 field program in Colorado.

An analysis of drop size distributions (DSDs) measured in four very different precipitation regimes is presented and is compared with polarimetric radar measurements. The DSDs are measured by a 2D video disdrometer, which is designed to measure drop size, shape, and fall speed with unprecedented accuracy. The observations indicate that significant DSD variability exists not only from one event to the next, but also within each system. Also, despite having vastly different storm structures and maximum rain rates, large raindrops with diameters greater than 5 mm occurred with each system. By comparing the occurrence of large drops with rainfall intensity, the authors find that the largest median diameters are not always associated with the heaviest rainfall, but are sometimes located either in advance of convective cores or, occasionally, in stratiform regions where rainfall rates are relatively low. Disdrometer and polarimetric radar measurements of radar reflectivity Z, differential reflectivity ZDR, specific differential phase KDP, and R(Z) and R(KDP) rain-rate estimators are compared in detail. Overall agreement is good, but it is found that both R(Z) and R(KDP) underestimate rain rate when the DSD is dominated by small drops and overestimate rain rate when the DSD is dominated by large drops. The results indicate that a classification of different rain types (associated with different DSDs) should be an essential part of polarimetric rainfall estimation. Furthermore, observations suggest that ZDR is a key parameter for making such a distinction. Last, the authors compute and compare maximum and average of gamma shape, slope, and intercept parameters for all four precipitation events. Potential measurement errors with the 2D video disdrometer are also discussed.

<p>Taiwan is located at the northwest side of Pacific Ocean and also in the Circum-Pacific Seismic Belt, as a result, suffers from frequently typhoons and earthquakes. The plate collision creates steep mountains account for 70% area of Taiwan. Averaged annual rainfall is 3600 mm in Taiwan, whenever typhoon or weather front brings heavy rainfall addition with geological instability thus increase the landslide occurrence. Rainfall gauge stations are sparse in the mountainous region, and the interpolated rainfall are usually underestimated. The error sources include 1.) variation of raindrop size distribution which is rarely known and varies in time and space (James, 1979); 2.) radar beam attenuation, the rainfall estimation will be underestimated as distance from radar to gauges increase ( Joss, 1998); 3.) beam blockage by mountains, when the beam encounters terrain blocking, it will cause signal interference, which is known as ground clutter (Li and Chen, 2002). Therefore, it is necessary to overcome those issues while try to predict accurate rainfall via radar reflectivity in the mountain regions.</p><p>In this research, we use radar reflectivity combines ground rainfall gauges to compensate the forecasting rainfall. The first method uses the known radar echo intensity (Z) and the rainfall of ground stations (R) to calculate the A and b coefficients by genetic algorithm with the exponential relationship, Z=A*R<sup>b</sup>, proposed by Marshall and Palmer. However, the results are unreasonable, the value of A is varying in 0.01-1000, mostly under 1, and the value of b is varying in 0.1-30. Hence, we decide to use another method. First, we assume that for a short distance (ex: 30 Km), the raindrop size correction factor is constant without attenuation and beam blockage. Second, we estimate the correction factor with the attenuation pattern with distance. Third, the beam blockage from mountains is then considered, and it also takes the first two corrections in consideration. The approach we used is artificial neural network (ANN) to compensate the estimated rainfall from real time radar reflectivity.</p><p>The purpose of this study is to estimate the accurate rainfall of potential landslide area hours ahead typhoon or weather front reaches, we use the historical route of radar echo to infer the path of movement of next hour. If the estimated rainfall exceeds landslide thresholds, the alarm system will be activated. With these efforts the estimated rainfall in the mountain region is improved 70% from tryout experiments. It is found that is correction for radar reflectivity is not an universal transformation, it is depended on the nature of water concentration and also the drop size within the weather front.</p><p> </p><p>Key words: Radar Echo, Artificial Neural Network, Early Warning</p>

A phenomenological relationship between vertical air motion and disdrometer derived A-b coefficients

Tropical and subtropical sites are known as areas of intense rainfall, where both satellite and terrestrial links suffer from severe levels of signal degradation. In this study, Butare (2°36'S, 29°44'E) and Durban (29°52'S, 30°58'E), with tropical and subtropical characteristics respectively, are compared at different measurement periods. Rainfall microstructural parameters such as rainfall rate, rainfall Drop Size Distribution (DSD) and radar reflectivity, are examined. It is found that at 99.99% rain availability (or R <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.01</sub> ), these parameters are slightly higher at Butare than Durban. However, specific attenuation computations tend to be higher at Durban compared to Butare, between 40 GHz and 1 THz.

The differential reflectivity (ZDR) measures the mean shape of hydrometeors and provides an estimate of the mean size of raindrops Observations of ZDR for rain may be combined with the conventional radar reflectivity factor (Z) and fitted to any two-parameter raindrop size distribution and this information used to derive more accurate rainfall rates. In such work the precise shape of raindrops is a critical parameter. Recently available data suggest that large raindrops are more oblate than previously believed. These new shapes support the idea that ZDR values above 3.5 dB can be attributed to rain. Average values of ZDR as a function of Z obtained in heavy rain by the Chilbolton radar agree very closely with those predicted using the new shapes. Statistics are also presented of the natural variability of raindrop spectra in heavy rain. Analytic expressions are proposed for computing rainfall rate from Z and ZDR.