P14.6 THE DIURNAL CYCLE OF PRECIPITATION IN RADAR OBSERVATIONS, MODEL FORECASTS AND MAPLE NOWCASTS

Similar documents
Data Assimilation. Alan O Neill Data Assimilation Research Centre University of Reading

Estimation of the particle concentration in hydraulic liquid by the in-line automatic particle counter based on the CMOS image sensor

12.3 REAL-TIME STORM-SCALE ENSEMBLE FORECAST EXPERIMENT ANALYSIS OF 2008 SPRING EXPERIMENT DATA

Incorporating Ensemble Covariance in the Gridpoint Statistical Interpolation Variational Minimization: A Mathematical Framework

DISTRIBUTION OF SUB AND SUPER HARMONIC SOLUTION OF MATHIEU EQUATION WITHIN STABLE ZONES

LAMEPS Limited area ensemble forecasting in Norway, using targeted EPS

Driving Cycle Construction of City Road for Hybrid Bus Based on Markov Process Deng Pan1, a, Fengchun Sun1,b*, Hongwen He1, c, Jiankun Peng1, d

Information processing via physical soft body. Supplementary Information

P1.1 VERIFICATION OF SURFACE LAYER OZONE FORECASTS IN THE NOAA/EPA AIR QUALITY FORECAST SYSTEM IN DIFFERENT REGIONS UNDER DIFFERENT SYNOPTIC SCENARIOS

Tests for the Ratio of Two Poisson Rates

Vorticity. curvature: shear: fluid elements moving in a straight line but at different speeds. t 1 t 2. ATM60, Shu-Hua Chen

APPROXIMATE INTEGRATION

New Expansion and Infinite Series

A signalling model of school grades: centralized versus decentralized examinations

Non-Linear & Logistic Regression

Chapter 4 Contravariance, Covariance, and Spacetime Diagrams

16A.3 A REAL-TIME STORM-SCALE ENSEMBLE FORECAST SYSTEM: 2009 SPRING EXPERIMENT

Fig. 1. Open-Loop and Closed-Loop Systems with Plant Variations

Properties of Integrals, Indefinite Integrals. Goals: Definition of the Definite Integral Integral Calculations using Antiderivatives

DECAMETER RADIO EMISSION OF THE SUN: RECENT OBSERVATIONS

1 Online Learning and Regret Minimization

A027 Uncertainties in Local Anisotropy Estimation from Multi-offset VSP Data

Student Activity 3: Single Factor ANOVA

Effects of peripheral drilling moment on delamination using special drill bits

1. a) Describe the principle characteristics and uses of the following types of PV cell: Single Crystal Silicon Poly Crystal Silicon

The steps of the hypothesis test

ADVANCEMENT OF THE CLOSELY COUPLED PROBES POTENTIAL DROP TECHNIQUE FOR NDE OF SURFACE CRACKS

Electron Correlation Methods

The impact of wind on air temperature distribution in Athens and in Santorini

A New Statistic Feature of the Short-Time Amplitude Spectrum Values for Human s Unvoiced Pronunciation

Emission of K -, L - and M - Auger Electrons from Cu Atoms. Abstract

Basic model for traffic interweave

7.2 The Definite Integral

Probabilistic Investigation of Sensitivities of Advanced Test- Analysis Model Correlation Methods

NUMERICAL INTEGRATION. The inverse process to differentiation in calculus is integration. Mathematically, integration is represented by.

Shear and torsion interaction of hollow core slabs

THE INTERVAL LATTICE BOLTZMANN METHOD FOR TRANSIENT HEAT TRANSFER IN A SILICON THIN FILM

Songklanakarin Journal of Science and Technology SJST R1 Thongchan. A Modified Hyperbolic Secant Distribution

CBE 291b - Computation And Optimization For Engineers

Monte Carlo method in solving numerical integration and differential equation

Lecture 3 Gaussian Probability Distribution

Predict Global Earth Temperature using Linier Regression

u( t) + K 2 ( ) = 1 t > 0 Analyzing Damped Oscillations Problem (Meador, example 2-18, pp 44-48): Determine the equation of the following graph.

SUMMER KNOWHOW STUDY AND LEARNING CENTRE

Operations with Polynomials

Physics 201 Lab 3: Measurement of Earth s local gravitational field I Data Acquisition and Preliminary Analysis Dr. Timothy C. Black Summer I, 2018

Supplementary Information to The role of endogenous and exogenous mechanisms in the formation of R&D networks

ESCI 343 Atmospheric Dynamics II Lesson 14 Inertial/slantwise Instability

#6A&B Magnetic Field Mapping

A Signal-Level Fusion Model for Image-Based Change Detection in DARPA's Dynamic Database System

8 Laplace s Method and Local Limit Theorems

A Brief Review on Akkar, Sandikkaya and Bommer (ASB13) GMPE

Math 426: Probability Final Exam Practice

UNIT 1 FUNCTIONS AND THEIR INVERSES Lesson 1.4: Logarithmic Functions as Inverses Instruction

A new algorithm for generating Pythagorean triples 1

Review of basic calculus

Short-Term Electrical Load Forecasting Using a Fuzzy ARTMAP Neural Network

Keywords Synthetic Aperture Radar (SAR); Range Doppler Algorithm (RDA); Range Cell Migration (RCM); remote sensing; Radarsat1 Y 0

Online Short Term Load Forecasting by Fuzzy ARTMAP Neural Network

DETERMINATION OF MECHANICAL PROPERTIES OF NANOSTRUCTURES WITH COMPLEX CRYSTAL LATTICE USING MOMENT INTERACTION AT MICROSCALE

Introduction to ISAR imaging systems. Elaborazione Sistemi Radar delle immagini radar

Math 8 Winter 2015 Applications of Integration

THE HANKEL MATRIX METHOD FOR GAUSSIAN QUADRATURE IN 1 AND 2 DIMENSIONS

Acceptance Sampling by Attributes

Vyacheslav Telnin. Search for New Numbers.

Industrial Electrical Engineering and Automation

P4.18 Drizzle Detection for Maritime Stratocumulus Clouds by Combined Use of TRMM Microwave Imager and Visible/Infrared Scanner

Problem Set 3 Solutions

20 MATHEMATICS POLYNOMIALS

Improper Integrals, and Differential Equations

Chapter 9: Inferences based on Two samples: Confidence intervals and tests of hypotheses

MASSACHUSETTS INSTITUTE OF TECHNOLOGY Physics Department Statistical Physics I Spring Term Solutions to Problem Set #1

Design Data 1M. Highway Live Loads on Concrete Pipe

Quantum Physics II (8.05) Fall 2013 Assignment 2

Linear and Non-linear Feedback Control Strategies for a 4D Hyperchaotic System

THE DECORRELATION SCALE: METHODOLOGY AND APPLICATION FOR PRECIPITATION FORECASTS

Strategy: Use the Gibbs phase rule (Equation 5.3). How many components are present?

Lecture 20: Numerical Integration III

1.9 C 2 inner variations

APPROPRIATE ATTENUATION MODEL FOR CHIANG MAI, THAILAND FROM FIELD MEASUREMENT TO MODEL EQUATION

UNIVERSITY OF KWAZULU-NATAL EXAMINATIONS: JUNE 2007

Numerical Integration

Adam K. Baker * and Gary M. Lackmann North Carolina State University, Raleigh, North Carolina

Where did dynamic programming come from?

FEM ANALYSIS OF ROGOWSKI COILS COUPLED WITH BAR CONDUCTORS

Time Truncated Two Stage Group Sampling Plan For Various Distributions

LECTURE 14. Dr. Teresa D. Golden University of North Texas Department of Chemistry

Testing categorized bivariate normality with two-stage. polychoric correlation estimates

Measuring Electron Work Function in Metal

Review of Calculus, cont d

Advanced Calculus: MATH 410 Notes on Integrals and Integrability Professor David Levermore 17 October 2004

8.2 ESTIMATING DETENTION VOLUMES

P 3 (x) = f(0) + f (0)x + f (0) 2. x 2 + f (0) . In the problem set, you are asked to show, in general, the n th order term is a n = f (n) (0)

Reversals of Signal-Posterior Monotonicity for Any Bounded Prior

Jim Lambers MAT 169 Fall Semester Lecture 4 Notes

3.4 Numerical integration

A New Receiver for Chaotic Digital Transmissions: The Symbolic Matching Approach

MATH SS124 Sec 39 Concepts summary with examples

Quantum Physics I (8.04) Spring 2016 Assignment 8

How to simulate Turing machines by invertible one-dimensional cellular automata

Transcription:

P14.6 THE DIURNAL CYCLE OF PRECIPITATION IN RADAR OBSERVATIONS, MODEL FORECASTS AND MAPLE NOWCASTS Mdlin Surcel 1*, Mrc Berenguer 1, Isztr Zwdzki 1, Ming Xue 2,3 nd Fnyou Kong 3 (1) Deprtment of Atmospheric nd Ocenic Sciences, McGill University, Montrel, Cnd (2) School of Meteorology, University of Oklhom, Normn, Oklhom (3) Center for Anlysis nd Prediction of Storms, University of Oklhom, Normn, Oklhom 1. INTRODUCTION We hve seen in recent pper (Surcel et l., 29) tht the diurnl cycle of precipittion over the continentl US exhibits some sesonl vribility. This is consequence of the fct tht during the summer, precipittion is strongly forced by the diurnl cycle of solr heting nd it usully initites s smll scles, while during spring, precipittion occurs t lrger scles, being synopticlly forced. Surcel et l. (29) lso showed tht the bility of convection prmeterized, numericl wether prediction (NWP) model to depict the diurnl cycle of precipittion lso vries with seson showing more skill during spring. Here, we dd to the previous work, the evlution of precipittion forecsts from two other models. The differences in performnce between models re indictive of the importnce of horizontl resolution, convective prmeteriztion nd rdr dt ssimiltion. In ddition, we compre numericl wether forecsts to Lgrngin persistence, rdr-bsed nowcsts. 2. DATA 2.1 Verifiction dt The verifiction dt consists of mps of hourly rinfll ccumultions derived from the 2D US composite rdr reflectivity mosics t 2.5km ltitude (obtined from Ntionl Severe Storm Lbortory; Zhng et l., 25), by pplying stndrd Z-R reltionship, Z=3R 1.5, nd ccumulting instntneous rinrte mps with 1-minute temporl resolution. 2.2 Model forecsts The first set of forecsts ws generted by the (Globl Environmentl Multiscle; Milhot et l., 26) model nd is described by Surcel et l. (29).* The reminder of the forecsts comes from the two control runs (from now on referred to s CN nd C) of the CAPS (Center for Anlysis nd Prediction of Storms) Storm-Scle Ensemble Forecsting (SSEF) system, run s prt of n experimentl forecsting progrm from April to June 28 t NOAA!s Hzrdous Wether Testbed (HWT; Xue et l., 29). The 1-member, WRF-bsed system produced 3 hour-forecsts strting t UTC. Nine of the members ssimilted rdr observtions of reflectivity nd rdil velocity using * Corresponding uthor ddress: Mdlin Surcel, McGill University, Dept. of Atmospheric nd Ocenic Sciences, Montrel, Cnd; e-mil: mdlin.surcel@mil.mcgill.c. CAPS 3DVAR system. The two control runs nlyzed here do not hve SREF-bsed perturbtions nd hve identicl model configurtions, with the only difference being the ssimiltion of rdr observtions in CN nd not in C. The min differences between the model runs re summrized in Tble 1, nd more informtion on the forecsting systems cn be found in Milhot et l. (25) nd Xue et l. (29). Model SSEF Reference Horizontl resolution Initil conditions Cumulus prmeteriztion Tble 1. Model chrcteristics. Milhot et l., 26 Xue et l., 29 15 km 4 km Regionl dt ssimiltion system Kin- Fritsch/Kuo trnsient 2.3 Rdr-bsed nowcsts NAM 12UTC Rdr dt ssimiltion in CN none MAPLE (see complete description in Germnn nd Zwdzki, 22) is n extrpoltion-bsed technique for precipittion nowcsting. It uses the vritionl echo trcking method (VET; Lroche nd Zwdzki, 1995) to estimte the motion field of precipittion, nd semi-lgrngin bckwrds lgorithm to extrpolte reflectivity mps to generte the forecsts. Here, we hve run MAPLE using the NSSL 2.5-km reflectivity mps to generte, every hour, 8-hour forecsts with resolution of 2 km in spce nd 15 minutes in time. 2.4 Cse studies nd nlysis domin The nlysis focused on 25 dys from 16 April 28 to 6 June 28, when forecsts from ll model configurtions were vilble. The sptil domin (blck rectngle in Fig. 1) ws chosen to cover most of centrl nd estern United Sttes, running from 13 W to 78 W in longitude nd between 32 N nd 45 N in ltitude, in order to void the region round the Rocky Mountins where the qulity of rdr observtions is poor.

Fig. 1. Anlysis domin. The red contours represent the coverge of the 2.5 km CAPPI mps, while the blck rectngle extending between 13W nd 78W in longitude nd 32N nd 45N in ltitude corresponds to the domin on which ll sttistics re computed. The precipittion pttern observed in this figure is typicl for spring 28. 3. THE DIURNAL CYCLE DURING SPRING 28 FROM PRECIPITATION FORECASTS Figure 5 shows the diurnl cycle of verge hourly rinrte during spring 28 on Hovmöller digrm, for observtions,, C, CN nd MAPLE. For the model forecsts we hve considered forecst hours -23, nd MAPLE ws initilized t UTC, 8 UTC nd 16 UTC. The models depict the observed propgting signl (digonl rinfll bnd running from 13W t 19 UTC to 85W t UTC; Fig. 2), but they show more vribility in propgting pths (the model-simulted bnds re wider thn observed; Figs. 2b-d). This could be due to the fct tht precipittion forecsts suffer of positionl nd timing errors, nd hence the model-simulted streks re offset with respect to the observed. The spin-up time is visible in Figs. 5b nd c in the cse of nd C, cusing n bsence of mximum t UTC long longitude 13W. The spin-up time lso cuses the discontinuity t UTC in Figs. 5b nd c, underlining the difference in forecst qulity between led times nd 23. Thnks to the rdr dt ssimiltion, CN does not suffer of these issues. : Rdr b C c CN d MAPLE e 18: 12: time UTC [h] 6: : 18: 12: 6: :.5.48.44.4.36.32.28.24.2.16.12.8.4. vg R [mm/h] Fig. 2. Averge diurnl cycle of precipittion intensity for the period 16 April 28 to 6 June 28 for rdr observtions () (b), C (c), CN (d) nd MAPLE (e). Units re mmh -1. We hve lso nlyzed the bility of MAPLE to reproduce the diurnl cycle of precipittion (Fig. 5e). As explined by Surcel et l. (29), nd illustrted in Fig. 2, there re two mechnisms of rinfll occurrence over the Continentl US: initition, long the foothills of the Rockies (longitude 13W) nd propgtion, in the longitudinl rnge 13W to 9W. As MAPLE is bsed on Lgrngin persistence (i.e., it tkes the observed rinfll mps t initiliztion time nd extrpoltes them ccording to the estimted motion field, leving in this wy the intensity of the precipittion field constnt long the forecst), it is ble to depict the prt of the diurnl cycle tht is merely explined by stedy propgtion of the systems (which Germnn et l., 26, explined s combintion of steering level winds nd pprent motion resulting from systemtic growth nd decy), but not the prt ssocited with initition nd decy. This is evident in Fig. 2e. Being initilized t UTC, time of precipittion initition, MAPLE nowcsts hve little skill during hours -7 UTC. On the other hnd, MAPLE does better job thn the NWP models in the 8-12 UTC time window, when the min mechnism of precipittion occurrence is propgtion. Therefore, MAPLE!s performnce is highly dependent on initiliztion time, nd initilizing it t times when precipittion systems re well orgnized (for exmple every eight hours strting t 2 UTC, not shown), would improve its depiction of the diurnl cycle. Figures 3 nd 4 show, respectively, the power spectr of the Hovmöller diurnl cycles of verge rinrte in Fig. 5 nd the phse of the 24-hour hrmonic s function of longitude.

.25 Rdr b C c CN d MAPLE e 4. 4.36.2 4.8 5.33 6. freq [h -1 ].15 6.86 8. scle [h] 1..1 9.6.8 12..6.5 48. -15-1-95-9 -85-8 -15-1-95-9 -85-8 -15-1-95-9 -85-8 -15-1-95-9 -85-8 -15-1-95-9 -85-8 Fig. 3. Normlized power spectr corresponding to the diurnl cycles of verge hourly rinrte of Fig. 2. Indeed, the models, especilly nd C show weker diurnl signl thn observed (there is power ssocite with the 12- nd 8-hour hrmonics, Figs. 3b nd c). CN reproduces best the observed power spectrum, showing tht rdr dt ssimiltion hs stbilizing effect on the forecsted systems, s it corrects for positionl errors t initiliztion time. The propgtion speed, determined by the grdul shift of the phse of the 24-hour hrmonic with longitude (Fig. 4) is best depicted by nd worst by C. MAPLE forecsts cpture the propgtion of the precipittion systems in the 93W-9W longitudinl rnge, but with systemtic 2-hour dely. phse [deg] 36 18-18 -36 freq:.4h -1 (scle: 24. h) -15-1 -95-9 -85-8 18 12 6 18 12 6 pek time UTC [h] OU model cn MAPLE 1-8h OU model c NSSL mosic Fig. 4. Phse of the 24-hour hrmonic s function of longitude corresponding to the verge Hovmöller digrms of Fig. 2 for rinfll ccumultions. Different lines correspond to rdr observtions (thick blck line), models (CN in green, C in mgent, in red), nd MAPLE (blue). The dotted lines indicte tht the hrmonic does not explin 1% of the observed vrince. 4. OBJECTIVE EVALUATION OF PRECIPITATION FORECASTS We hve investigted the diurnl vribility of forecst skill for models nd MAPLE, in terms of the Criticl Success Index (CSI, Wilks, 1995) nd correltion (Fig. 5) between forecsts nd observtions. Here, the correltion coefficients were computed in 16. 24. logrithmic units nd without subtrcting the men (s in Germnn nd Zwdzki, 22). correltion CSI 1..5.1 5 1 15 2 UTC [h] 1..8.6.4.2. 5 1 15 2 UTC [h] b.4.2. CN C MAPLE OU cn OU c MAPLE Fig. 5. Overll performnce of the models s function of time of the dy. () correltion of the equivlent reflectivity fields (in dbz); (b) Criticl Success Index for the intensities over.2 mm/h. CN performs better thn C throughout the forecst, showing tht rdr dt ssimiltion t initiliztion time hs beneficil impct on the verge qulity of the forecsts. On the other hnd, it ppers tht incresing horizontl resolution beyond 15 km does not hve ny effect on forecst qulity, s is better, or t most equl to C t ll times. However,

there re some observtions to be mde. First, the two models re not identicl, nd therefore the differences between them my come not only from their horizontl resolutions, but lso from the qulity of their physicl representtion. For exmple, it seems tht C hs longer spin up time thn (roughly, 2 hours compred to 6), which cn be due to the fct tht the initil conditions for C re simply provided by the NAM nlysis, while for, they re provided by 3D- VAR, regionl dt ssimiltion system. Then, C is expected to perform better thn during the summer, when precipittion is orgnized t the mesoscle, therefore demnding high enough horizontl resolution to llow the explicit tretment of convection. Unfortuntely, s the SSEF system is not run on regulr bsis, it ws not possible to evlute it during summer period. MAPLE nowcsts show perfect CSI nd correltion t initil time, but which quickly decy with led-time. This decrese is steeper thn observed by previous studies (Germnn nd Zwdzki, 22), probbly becuse our dt set is not strictly composed of lrge scle, firly predictble systems, but it includes convective cses in which MAPLE hs little skill. In ddition, the time it tkes for the models to outperform MAPLE is lower in our cse (2 hours for CN, 3-4 hours for nd C) thn found by Lin et l. (25; 6 hours). This is lso due in prt to our dt set, nd to the better qulity of the models (e.g. CN ws run t higher resolution nd hd rdr dt ssimilted). For the model forecsts, other thn lredy mentioned, CSI nd correltion do not exhibit ny significnt dependence on the time of dy. Therefore, it seems tht the diurnl cycle of precipittion is not the min cuse for the poor model performnce. It is worth mentioning two problems ssocited with this type of objective evlution. First, s lredy mentioned, our dt set is composed of precipittion cses tht exhibit significnt growth nd decy. For these cses, MAPLE hs little skill, being bsed on Lgrngin persistence. On the other hnd, in verifiction ginst rdr derived rinfll mps, MAPLE nowcsts hve n dvntge over model forecsts, s they re generted from rdr observtions. These two fctors counterct ech other. 5. CONCLUSIONS In this pper, we hve compred the bility of three NWP models to depict the diurnl cycle of precipittion over the continentl US during spring 28. We found tht the nlyzed NWP models could reproduce the propgting signl, showing however more vribility in propgtion pths thn observed. CN performed best, especilly t the beginning of the forecst time, showing the importnce of rdr dt ssimiltion. Unfortuntely, the little difference in performnce between C nd did not show ny beneficil effect of incresing horizontl resolution beyond 15 km. We hve lso investigted the performnce of MAPLE s function of the time of dy. As lso suggested by previous studies (Berenguer et l., 28), the diurnl cycle of precipittion strongly influenced MAPLE!s performnce, such tht the qulity of the nowcsts ws highly dependent on the initiliztion time. Also, on verge, for ll precipittion cses during the study period, MAPLE nowcsts initilized t UTC outperformed CN only during the first two forecst hours, suggesting the potentil vlue of rdr dt ssimilting NWP models for very-short term forecsting. ACKNOWLEDENTS This work ws mde possible by the support of Environment Cnd to the J. S. Mrshll Rdr Observtory. REFERENCES Berenguer, M., B. Turner, nd I. Zwdzki, 28: The effect of the diurnl cycle of precipittion in rdr-bsed short-term forecsts. 5th Europen Conference on Rdr in Meteorology nd Hydrology, Helsinki, Finlnd, CD-ROM P11.21. Germnn, U. nd I. Zwdzki, 22: Scle-dependence of the predictbility of precipittion from continentl rdr imges. Prt I: Description of the methodology. Monthly Wether Review, 13, 2859-2873. Lroche, S. nd I. Zwdzki, 1995: Retrievls of Horizontl Winds from Single-Doppler Cler-Air Dt by Methods of Cross-Correltion nd Vritionl Anlysis. Journl of Atmospheric nd Ocenic Technology, 12, 721-738. Lroche, S., P. Guthier, J. St-Jmes nd J. Morneu, 1999: Implementtion of 3D vritionl dt ssimiltion system t the Cndin Meteorologicl Centre. Prt II: The regionl nlysis. Atmosphere- Ocen, 37, 281-37. Lin, C., S. Vsic, A. Kilmbi, B. Turner, nd I. Zwdzki, 25: Precipittion forecst skill of numericl wether prediction models nd rdr nowcsts. Geophysicl Reserch Letters, 32, L1481. Milhot, J., S. Belir, L. Lefivre, B. Bilodeu, M. Desggne, C. Girrd, A. Glzer, A. M. Leduc, A. Methot, A. Ptoine, A. Plnte, A. Rhill, T. Robinson, D. Tlbot, A. Trembly, P. Villncourt, A. Zdr, nd A. Qddouri, 26: The 15-km version of the Cndin regionl forecst system. Atmosphere-Ocen, 44, 133-149. Surcel, M., M. Berenguer, nd I. Zwdzki, 29: The diurnl cycle of precipittion from Continentl rdr imges nd Numericl Wether Prediction Models. Prt I: Methodology nd sesonl comprison. Monthly Wether Review (in review). Turner, B. J., I. Zwdzki, nd U. Germnn, 24: Predictbility of precipittion from continentl rdr imges. Prt III: Opertionl nowcsting implementtion (MAPLE). Journl of Applied Meteorology, 43, 231-248. Wilks, D. S., 1995: Sttisticl Methods in Atmospheric Sciences. 467.

Xue, M., F. Kong, K. W. Thoms, J. Go, Y. Wng, K. Brewster, K. K. Droegemeier, J. Kin, S. Weiss, D. Bright, M. Coniglio, nd J. Du, 28: CAPS reltime storm-scle ensemble nd high-resolution forecsts s prt of the NOAA Hzrdous Wether Testbed 28 Spring Experiment. 24th Conference on Severe Storms, Americn Meteorologicl Society, 12.2. Zhng, J., K. Howrd, nd J. J. Gourley, 25: Constructing three-dimensionl multiple-rdr reflectivity mosics: Exmples of convective storms nd strtiform rin echoes. Journl of Atmospheric nd Ocenic Technology, 22, 3-42.