B.W.Golding * Met Office, Exeter, UK

Similar documents
The Nowcasting Demonstration Project for London 2012

Heavier summer downpours with climate change revealed by weather forecast resolution model

Aurora Bell*, Alan Seed, Ross Bunn, Bureau of Meteorology, Melbourne, Australia

Strategic Radar Enhancement Project (SREP) Forecast Demonstration Project (FDP) The future is here and now

Assimilation of SEVIRI cloud-top parameters in the Met Office regional forecast model

Nowcasting for the London Olympics 2012 Brian Golding, Susan Ballard, Nigel Roberts & Ken Mylne Met Office, UK. Crown copyright Met Office

Figure 1: Tephigram for radiosonde launched from Bath at 1100 UTC on 15 June 2005 (IOP 1). The CAPE and CIN are shaded dark and light gray,

DATA FUSION NOWCASTING AND NWP

For the operational forecaster one important precondition for the diagnosis and prediction of

Introduction to Meteorology and Weather Forecasting

Convective Scale Ensemble for NWP

Convective-scale data assimilation at the UK Met Office

Observations needed for verification of additional forecast products

Application and verification of ECMWF products 2016

Operational convective scale NWP in the Met Office

Nesting and LBCs, Predictability and EPS

Huw W. Lewis *, Dawn L. Harrison and Malcolm Kitchen Met Office, United Kingdom

Generating probabilistic forecasts from convectionpermitting. Nigel Roberts

QPE and QPF in the Bureau of Meteorology

Heavy Rainfall and Flooding of 23 July 2009 By Richard H. Grumm And Ron Holmes National Weather Service Office State College, PA 16803

Heavy Rainfall Event of June 2013

i. Motivation of report

Understanding Weather and Climate Risk. Matthew Perry Sharing an Uncertain World Conference The Geological Society, 13 July 2017

MSG FOR NOWCASTING - EXPERIENCES OVER SOUTHERN AFRICA

TIFS DEVELOPMENTS INSPIRED BY THE B08 FDP. John Bally, A. J. Bannister, and D. Scurrah Bureau of Meteorology, Melbourne, Victoria, Australia

Utilising Radar and Satellite Based Nowcasting Tools for Aviation Purposes in South Africa. Erik Becker

On the use of radar rainfall estimates and nowcasts in an operational heavy rainfall warning service

Severe Weather Watches, Advisories & Warnings

Regional Flash Flood Guidance and Early Warning System

Improving real time observation and nowcasting RDT. E de Coning, M Gijben, B Maseko and L van Hemert Nowcasting and Very Short Range Forecasting

Predicting rainfall using ensemble forecasts

ECMWF products to represent, quantify and communicate forecast uncertainty

L alluvione di Firenze del 1966 : an ensemble-based re-forecasting study

Xinhua Liu National Meteorological Center (NMC) of China Meteorological Administration (CMA)

DETECTION AND FORECASTING - THE CZECH EXPERIENCE

0-6 hour Weather Forecast Guidance at The Weather Company. Steven Honey, Joseph Koval, Cathryn Meyer, Peter Neilley The Weather Company

Severe storms over the Mediterranean Sea: A satellite and model analysis

REVISION OF THE STATEMENT OF GUIDANCE FOR GLOBAL NUMERICAL WEATHER PREDICTION. (Submitted by Dr. J. Eyre)

Seamless Probabilistic Forecasts for Civil Protection: from week to minutes

Charles A. Doswell III, Harold E. Brooks, and Robert A. Maddox

Flood Forecasting with Radar

Chapter 3 Convective Dynamics 3.4. Bright Bands, Bow Echoes and Mesoscale Convective Complexes

LATE REQUEST FOR A SPECIAL PROJECT

Nimrod: A system for generating automated very short range forecasts

Aviation Hazards: Thunderstorms and Deep Convection

Observation Based Products Technical Report No. 13

Northeastern United States Snowstorm of 9 February 2017

Shaping future approaches to evaluating highimpact weather forecasts

Forecasting the "Beast from the East" and Storm Emma

Preliminary results. Leonardo Calvetti, Rafael Toshio, Flávio Deppe and Cesar Beneti. Technological Institute SIMEPAR, Curitiba, Paraná, Brazil

Reprint 797. Development of a Thunderstorm. P.W. Li

REQUIREMENTS FOR WEATHER RADAR DATA. Review of the current and likely future hydrological requirements for Weather Radar data

1. COLD FRONT - CLOUD STRUCTURE IN SATELLITE IMAGES

P5.11 TACKLING THE CHALLENGE OF NOWCASTING ELEVATED CONVECTION

Appendix 1: UK climate projections

Seamless nowcasting. Open issues

Update on CoSPA Storm Forecasts

Mesoscale meteorological models. Claire L. Vincent, Caroline Draxl and Joakim R. Nielsen

JMA Contribution to SWFDDP in RAV. (Submitted by Yuki Honda and Masayuki Kyouda, Japan Meteorological Agency) Summary and purpose of document

1. Current atmospheric DA systems 2. Coupling surface/atmospheric DA 3. Trends & ideas

DEPARTMENT OF GEOSCIENCES SAN FRANCISCO STATE UNIVERSITY. Metr Fall 2012 Test #1 200 pts. Part I. Surface Chart Interpretation.

Weather Forecasting: Lecture 2

Assimilation of Doppler radar observations for high-resolution numerical weather prediction

SATELLITE MONITORING OF THE CONVECTIVE STORMS

Judit Kerényi. OMSZ - Hungarian Meteorological Service, Budapest, Hungary. H-1525 Budapest, P.O.Box 38, Hungary.

Model enhancement & delivery plans, RAI

AROME Nowcasting - tool based on a convective scale operational system

New applications using real-time observations and ECMWF model data


FLORA: FLood estimation and forecast in complex Orographic areas for Risk mitigation in the Alpine space

CHARACTERISATION OF STORM SEVERITY BY USE OF SELECTED CONVECTIVE CELL PARAMETERS DERIVED FROM SATELLITE DATA

USE OF SATELLITE INFORMATION IN THE HUNGARIAN NOWCASTING SYSTEM

Evaluating Parametrizations using CEOP

CONTRIBUTION OF ENSEMBLE FORECASTING APPROACHES TO FLASH FLOOD NOWCASTING AT GAUGED AND UNGAUGED CATCHMENTS

2 July 2013 Flash Flood Event

Minor Winter Flooding Event in northwestern Pennsylvania January 2017

ACTIVITY. Weather Radar Investigation. Additional Activities

MODEL TYPE (Adapted from COMET online NWP modules) 1. Introduction

JOINT WMO TECHNICAL PROGRESS REPORT ON THE GLOBAL DATA PROCESSING AND FORECASTING SYSTEM AND NUMERICAL WEATHER PREDICTION RESEARCH ACTIVITIES FOR 2006

Recent ECMWF Developments

JP1J.9 ASSIMILATION OF RADAR DATA IN THE MET OFFICE MESOSCALE AND CONVECTIVE SCALE FORECAST SYSTEMS

Inner core dynamics: Eyewall Replacement and hot towers

Linking climate change modelling to impacts studies: recent advances in downscaling techniques for hydrological extremes

Medium-range Ensemble Forecasts at the Met Office

Indiana County Flash Flood of 22 June 2017

Nowcasting of Severe Weather from Satellite Images (for Southern

Atmospheric Moisture, Precipitation, and Weather Systems

The effect of ice fall speed in the structure of surface precipitation

Meteorology Lecture 18

Mesoscale analysis of a comma cloud observed during FASTEX

Flood Risk Forecasts for England and Wales: Production and Communication

Extratropical cyclones a forecaster s perspective

Thunderstorm Downburst Prediction: An Integrated Remote Sensing Approach. Ken Pryor Center for Satellite Applications and Research (NOAA/NESDIS)

Introduction to African weather

Mid-West Heavy rains 18 April 2013

INVESTIGATION FOR A POSSIBLE INFLUENCE OF IOANNINA AND METSOVO LAKES (EPIRUS, NW GREECE), ON PRECIPITATION, DURING THE WARM PERIOD OF THE YEAR

AN OBSERVING SYSTEM EXPERIMENT OF MTSAT RAPID SCAN AMV USING JMA MESO-SCALE OPERATIONAL NWP SYSTEM

Regional Hazardous Weather Advisory Centres (RHWACs)

Probabilistic Weather Prediction

INTERPRETATION OF MSG IMAGES, PRODUCTS AND SAFNWC OUTPUTS FOR DUTY FORECASTERS

Transcription:

7.35 A NEW APPROACH TO NOWCASTING AT THE MET OFFICE B.W.Golding * Met Office, Exeter, UK 1. BACKGROUND * The Nimrod system was introduced into operational use in the Met Office in 1996 (Golding, 1998). It produces nowcasts of precipitation, cloud, visibility, temperature, wind, lightning and some related variables by combining an extrapolation forecast, based on a detailed analysis, with output from the Met Office mesoscale NWP model. The essence of the Nimrod approach is illustrated in figure 1, which shows the decline in information content of three curves: the upper, deep blue, line falls as the chaotic predictability limit is reached at progressively larger scales; the nowcast, light blue, line starts high due to use of a detailed analysis but falls more steeply due to the crude extrapolation technique; and the model, red, curve is poor to start with, due to the limited model resolution, but falls more slowly due to the representation of non-linear processes. The aim of Nimrod was to harness the best of the nowcast and NWP approaches. Initially, most Nimrod nowcasts were produced on a 5km grid. Subsequently, the Gandolf precipitation nowcast was added, using a 2km grid to provide updates every 15 minutes, and with the addition of a lifecycle model of severe convection. convection-resolving NWP model. Preliminary tests of this model in summer 2004 demonstrated the capability of a 1km grid configuration for nowcasting, and prompted early introduction of a 4km grid configuration. At the same time, progress in quantitative probabilistic forecasting, especially using ensembles, has demonstrated that for high impact weather forecasting, a measure of uncertainty should always accompany any forecast, however short the lead time. Figure 2 illustrates the importance of uncertainty in precipitation nowcasting. It uses the rules of thumb that a perfect non-linear model should be able to predict three lifecycles ahead, while a linear extrapolation model can only predict a half lifecycle at best. The largest precipitation systems, MCSs and fronts, have spatial scales of around 100km and development lifecycles of about 12 hours. Systems of this size may be predictable up to 36 hours ahead using models, but only 6 hours ahead using extrapolation. However, the most damaging precipitation systems are thunderstorms and their embedded maxima, with lifecycles of an hour or less, and correspondingly shorter predictability limits. Log Scale Nowcast (Extrapolation Forecast) Space Scale 1000km 100km MCS Front Extratropical Cyclone Limit ( Perfect Forecast ) 10km Thunderstorm Information Content NWP (Model Forecast) 1km Lifetime Predictability Nowcast Hail shaft 10mins 1 hr 12hrs 3 days 30mins 3 hrs 36hrs 9 days 5mins 30min 6hrs 36hrs 0.01 0.1 1 10 Forecast Length Log Scale (days) Figure 1: Representation of the loss of accuracy in forecasts Over the past few years, a new strategy for very short range forecasting over the UK has been developed, based on the application of a * Corresponding author address: Brian W. Golding, Met Office, FitzRoy Road, Exeter EX1 3PB, UK; e-mail: brian.golding@metoffice.gov.uk Figure 2: Representation of the time and space scales of precipitation systems, and their predictability In order to permit the smooth transition from current nowcasts, based on extrapolation, to a future system based on convective scale NWP, the role of the nowcasting system was reconsidered. A new system configured as a model post-processing system, has been designed to meet the identified key requirements:

All model output will be post-processed to ensure consistency Post-processing will be an integral part of the NWP suite Changes of model should be transparent, allowing integration of output from coarse resolution models at longer lead times. Output products will be re-gridded and downscaled to a standard grid, initially at 2km resolution, but subsequently to be decreased to 1km. Key variables will be updated to latest observations in a Rapid Update Cycle, running every hour or less. This will involve adjustment to an extrapolated analysis over the subsequent few hours of forecast. The option of including an adjustment to a manually imposed broad scale evolution will be included if required. A measure of uncertainty should be attached to every forecast product to represent, at least, the limited predictability of the finest scales. Sources of uncertainty from ensembles and other sources should be integrated and should be consistently specified at different lead times. Additional variables, diagnosed from the main model outputs, should be provided consistently. In summary, the new system, based on convective scale NWP, can be characterised as: A rapid update cycle for forecasting hazardous weather variables using high resolution observations merged with NWP forecasts. 2. OBSERVATION SOURCES The observation sources remain similar, but image processing of radar and satellite observations has progressed substantially since 1996, and has been separated into independent systems which generate the required input products. The quality of data obtained from both of these sources has improved in recent years. 2.1 Radar A major upgrade to the processing of radar data has been implemented in 2005, following several years development work. Instead of processing being split between the radar sites and Nimrod, the raw data are now communicated by all radar sites to the central Radarnet IV computer, which performs all processing in polar co-ordinates prior to product generation. This has enabled implementation of more advanced signal processing techniques and, in particular, has permitted more intelligent infilling of clutter contaminated data. Figure 3 shows the benefit of the improved processing on processed data from the Jersey radar, as seen in the frequency of detection of precipitation during January 2005. The left hand image shows the new technique and the right hand image is taken from Nimrod. The new technique shows much less evidence of clutter on the French coast and less evidence of sea clutter at around 50 km range. The accumulation image (not shown) confirms these observations. Additionally it shows accumulations in the sector between 0-60 Figure 3: Frequency of detection of precipitation from the Jersey radar, in January 2005, after processing using Radarnet IV (left) and Nimrod (right)

Visible Autosat IV IR Nimrod Figure 4: Example of improved cloud top detection. Upper and lower left images are the visible and infra-red MSG images respectively. Lower right is the Nimrod cloud top height field based on Meteosat-7 visible and infra-red images. Upper right is the Autosat IV cloud top height field based on Meteosat-8 (MSG) multi-channel images degrees which are more consistent with surrounding accumulations. This is because Nimrod uses infilling with a higher elevation scan in this sector whereas Radarnet IV uses the lowest scan. 2.2 Satellite Satellite data processing has been revolutionised by the operational introduction of Meteosat Second Generation (MSG). The basic technique for cloud detection continues to be the application of a set of one-sided threshold filters. However, in place of the two Nimrod tests on the visible and infra-red images, the new MSG processing system, on Autosat IV, uses the following six channels: Vis0.6, Vis0.8, IR3.9, IR8.7, IR10.8, IR12.0. The principal benefit of the additional channels is to give reliable cloud discrimination throughout both the night and day. A completely new approach to cloud top height detection has been implemented based on a variational formulation using the NWP model forecast profile brightness temperatures, satellite brightness temperatures in six bands, and also taking into account atmospheric stability. The bands used are IR6.2, IR7.3, IR8.7, IR10.8, IR12.0, IR13.4. Figure 4 shows an example of differences in the processed cloud top height between the two systems. Note especially removal of the spurious fog to the south of Ireland from the Nimrod product. Advantages of the upgraded satellite processing include improved fog detection and spurious precipitation echo removal, especially at night. 3. CONVECTIVE SCALE UK NWP MODEL The post-processing system is designed as part of a strategy to implement a convection resolving model on a 1km grid towards the end of the decade. This model will be nested within the North Atlantic European area model recently implemented with a 12km grid, either through an intermediate nested grid, or using variable resolution. Its primary aim will be the forecasting of high impact weather, especially thunderstorms (Roberts, 2005). In the interim, good performance in convective events that occurred in summer 2004 has prompted implementation of a 4km grid configuration. This model has now been implemented in operational trial mode for summer 2005, covering the whole UK, as shown in figure 5, and forms the basis for the first release of the post-processing system. The inner area marked in the figure is the post-

processing area, which will be the limit of availability of end products. it placed the highest accumulations too far north-east, relative to the radar observations. The 1km grid model was much more accurate in the location of the heaviest rain and captured some of the main details of the distribution of the storms, including the meander in the track to the north of Boscastle (figure 7). More details of the development and performance of the convective scale model are given by Roberts, Lean, Wilson, Clark, Bornemann, Dixon, Li, Swarbrick, Stiller and Ballard other symposium papers. Fog represents a very different type of high impact weather which benefits from the application of high resolution models. Figure 8 shows the representation of orographic influence on radiation fog for models with 12km, 4km and 1km grids. The 1km model shows the orographic control very clearly. The 4km model captures the main orographically forced features, while the 12km model has only a poor representation. Figure 5: Domain of the UK 4km NWP model and post-processing sub-domain The devastating flood event in Boscastle, Cornwall on 16 th August 2004, provided a demanding test of the capability of the new model. Figure 6 summarises the precipitation accumulations forecast and observed for the sixhour duration of the storm. The 12km model, using entirely parametrized convection, failed to predict large accumulations or to locate preferred locations for severe precipitation. The 4km grid model, using largely resolved convection, successfully predicted a narrow band of precipitation along the north Cornish coast, though 4. DOWNSCALING In order to insulate the final products from changes in the resolution of the NWP model, the output is first downscaled to a standard grid. This will also permit allowance to be made for use of smoothed orography fields in the model, for artificial smoothing introduced by the numerical integration scheme, and for different resolutions at different lead times. Allowance can be made both for land height differences and for land type differences in this process. Initially the standard grid will be a 2km grid on the UK National Grid projection. It is planned to move to a 1km grid in due course. Initially, Figure 6: Rainfall accumulations for south west England, 1200-1800UTC 16/8/2004. Operational 12km model (left), Experimental 4km model (centre) and radar observations (right)

Figure 7 Observed (left) and forecast (right) rainfall rates for Cornwall and Devon, 1600UTC 16/8/2004. Boscastle is on the north coast near the western limit of radar range adjustments for orographic height difference will be applied to pressure, wind, temperature, visibility and rainfall. It is also planned to use the appropriate tile values for surface quantities from the model surface exchanges scheme where appropriate. 5. RAPID UPDATE The rapid update cycle uses observations to generate an analysis of key variables and to adjust the next few hours of forecast time accordingly. The procedures used for the adjustment are different for each variable, and are largely based on existing Nimrod techniques. They are increasingly including a probabilistic element. A typical procedure includes an hourly analysis based on the latest satellite and radar images, combined with surface data, lightning fixes and in situ observations. The analysis is compared with the model forecast for that time, and either the analysis or the anomaly may be extrapolated forward to T+6 hours. The result is merged with the corresponding downscaled model forecasts. This procedure is carried out for precipitation, lightning, cloud, visibility, humidity, temperature, wind & pressure. Results are typically available within about 15 minutes of nominal data time. A brief outline of the techniques used for each variable follows. Visibility in m Figure 8: Visibility forecast for part of southern England, 0800UTC 18/11/2002. Operational 12km model (left), 4km model (centre) and 1km model (right)

5.1 Precipitation The analysis is generated using a corrected radar composite as preferred data source where the radars have good visibility. Elsewhere, the analysis is generated using a variational analysis technique which blends a very short period extrapolation forecast, satellite-derived rain (calibrated against radar), lightning fixes (which impose a minimum rain rate), and surface in situ weather reports. The forecast is generated using the STEPS procedure, consisting of a scale decomposition, extrapolation of each scale with smaller scales progressively replaced by auto-correlated noise, and merging with the model. It is planned to implement an ensemble of realisations as the basis for providing probability forecasts. More detailed presentations on STEPS are provided by Pierce, Bowler and Seed in other symposium papers. 5.2 Cloud The analysis is performed on 29 horizontal levels, mainly concentrated near the ground, using satellite imagery and surface observations to adjust the forecast first guess as shown in figure 9. The extrapolation forecast is generated by advecting precipitating cloud with vectors computed in the precipitation nowcast, and nonprecipitating cloud with the appropriate model layer wind. The result is merged with the model cloud forecast. Surface visual & instrumental observations define lower cloud layers Satellite images define highest cloud layer Model forecast & surface visual cloud type observations define intermediate cloud layers Figure 9: Schematic diagram showing the use of observations in the cloud analysis 5.3 Visibility The visibility nowcast is largely performed using the conserved variables: liquid water temperature and total water. The analysis uses satellite imagery and surface observations and is carried out in separate domains according to whether the satellite imagery detects possible fog or not, as shown in figure 10. The forecast is created by merging an extrapolation with the model forecast. Low cloud from the cloud forecast is also added where cloud base is below the fine scale orography. The end products are derived from the conserved variables taking account of the model predicted aerosol content. x x x x xx x x x Figure 10: Schematic of the visibility analysis. Observations within the red ring, where the satellite indicates possible fog, are analysed separately from those outside it 5.4 Wind & pressure The wind and surface pressure analyses are created by analysing anomalies from the model forecast, indicated by surface in situ observations. These anomalies are advected with the 10m wind. 6. UNCERTAINTY Observations Satellite area of fog/low cloud A key component of the planned development of the post-processing system is that different estimates of uncertainty should be integrated. Currently, different variables use different approaches: Precipitation uses the STEPS ensemble to characterise the uncertainty due to the advection vector and to stochastic evolution at small scales. More detailed information on the STEPS ensemble is given by Seed, Bowler and Pierce in other symposium papers. Visibility uses a Gaussian distribution of humidity that is part of the cloud fraction formulation Precipitation type uses the fractional melting of snow between the wet bulb freezing level

and the ground to indicate the probability of snow at the ground Lightning is characterised by the flash rate, expressed as a return period Looking to the future, implementation of a short range regional ensemble will provide estimates of synoptic scale uncertainty, which should map on to the advection vector uncertainty in STEPS, as well as providing a measure of large scale development uncertainty. A different approach to estimating the impact of small scale position and advection uncertainty has been taken in processing output from the 1km model. Figure 11 shows examples of products generated from a single 1km forecast of the Boscastle storm using different assumptions about such uncertainties, and different averaging areas. Further results on this approach are presented by Roberts and Lean in another symposium paper. Further work is planned to draw the (a) approaches together into a single representation of uncertainty that can be used for all variables and lead times. 7. DIAGNOSTIC PROCESSING In order to support critical weather services, the Nimrod system has been extended to compute several diagnostic quantities for which observations are either not available, or are too sparse to be spatially analysed and extrapolated. Several algorithms are applied to diagnose precipitation type, and convective severe weather using the nowcast precipitation together with NWP wind, temperature and humidity profiles. The outputs include: precipitation type, snow probability, hail probability & size, freezing precipitation, ice (b) Maximum accumulations Extreme accumulations (a) (c) 10 20 30 40 50 60 70 mm (d) (b) 40 60 80 100 120 140 160 mm Probability > 50 mm Maximum sub-area accumulations 0.1 4 8 12 16% 10 20 30 40 mm Figure 11: Example rainfall accumulation products for 1200 1800 UTC 16/8/2004 generated from a 1km NWP forecast initialised at 0000UTC: (a) maximum 1km value per 24km square; (b) maximum 1km value per 24km square if the storm was stationary; (c) probability of exceeding 50mm; (d) maximum 1km value per flood warning area

accretion, wind gust and indicators of the likelihood of severe weather including tornadoes. It is planned to extend the range of these algorithms to include damage predictors. The largest diagnostic component deals with hydrological land surface properties and uses an off-line version of the MOSES (Met Office Surface Exchange Scheme) from the Unified Model as its basis. It has been enhanced by the addition of the Probability Distributed Model, which allows for heterogeneous soil properties in computing run-off, and the River Flow Model, a grid-to-grid routing model (Smith et al, 2005). Together these models take analysed precipitation and cloud, together with surface wind, humidity and temperature, and diagnose the soil moisture profile, evaporation, run-off and river flow for the river network of the British Isles. Figure 12 shows an example of output from the river flow model during the flooding in Carlisle in January 2005, processed to show nearness to overtopping relative to the natural channel capacity (information on flood embankments has not yet been included). It is planned to extend this capability to predict other hydrologically related risk indicators. 8. SUMMARY A new nowcasting system design is being developed for use with output from convective scale NWP models. The model outputs are adjusted to a standard resolution and then updated to match latest observations using a rapid update cycle incorporating satellite and radar images, and extrapolation forecast techniques. A suite of diagnostic tools is used to produce generic weather impact forecasts for use in public weather services. 9. ACKNOWLEDGEMENTS The material presented in this paper has been provided by a large number of colleagues in the Met Office, including N.Roberts, H.Lean, P.Clark, C.Pierce, A.Cooper, R.Smith, B.Wright, M.Kitchen and S.Watkin. Mapping information is Crown Copyright 2005, reproduced under licence number 100008418 Figure 12: River flow diagnosed by the MOSES-PDM-RFM model suite during the Carlisle flood of 8/1/2005. Colours indicate flow relative to the natural channel capacity: blue < 85%, yellow 85-95%, orange 95-100% & red >100%

10. REFERENCES Golding,B.W., 1998, Nimrod: A System for generating automated very short range forecasts, Meteorol. Appl., 5, 1-16. Roberts, N., 2005: An investigation of the ability of a storm scale configuration of the Met Office NWP model to predict flood-producing rainfall, Met Office FR Technical Report 455. Available from http://www.metoffice.gov.uk/research/nwp/publicati ons/papers/technical_reports/fr.html Smith, R.N.B., Blyth, E.M., Finch, J.W., Goodchild, S., Hall, R.L. & Madry, S., 2005: Soil state and surface hydrology diagnosis based on MOSES in the Met Office Nimrod nowcasting system. Submitted to Meteorol. Appl.