WORLD METEOROLOGICAL ORGANIZATION OPERATIONAL HYDROLOGY REPORT No. 42

METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

WMO-No. 813

Secretariat of the World Meteorological Organization Ð Geneva Ð Switzerland THE WORLD METEOROLOGICAL ORGANIZATION

The World Meteorological Organization (WMO), of which 187* States and Territories are Members, is a specialized agency of the United Nations. The purposes of the Organization are: (a) To facilitate worldwide cooperation in the establishment of networks of stations for the making of meteorological observations as well as hydrological and other geophysical observations related to meteorology, and to promote the establishment and maintenance of centres charged with the provision of meteorological and related services; (b) To promote the establishment and maintenance of systems for the rapid exchange of meteorological and related infor- mation; (c) To promote standardization of meteorological and related observations and to ensure the uniform publication of observations and statistics; (d) To further the application of meteorology to aviation, shipping, water problems, agriculture and other human activi- ties; (e) To promote activities in operational hydrology and to further close cooperation between Meteorological and Hydrological Services; and (f) To encourage research and training in meteorology and, as appropriate, in related fields and to assist in coordinating the international aspects of such research and training.

(Convention of the World Meteorological Organization, Article 2)

The Organization consists of the following:

The World Meteorological Congress, the supreme body of the Organization, brings together the delegates of Members once every four years to determine general policies for the fulfilment of the purposes of the Organization, to approve long- term plans, to authorize maximum expenditures for the following financial period, to adopt Technical Regulations relating to international meteorological and operational hydrological practice, to elect the President and Vice-Presidents of the Organization and members of the Executive Council and to appoint the Secretary-General;

The Executive Council, composed of 36 directors of national Meteorological or Hydrometeorological Services, meets at least once a year to review the activities of the Organization and to implement the programmes approved by Congress;

The six regional associations (Africa, Asia, South America, North and Central America, South-West Pacific and Europe), composed of Members, coordinate meteorological and related activities within their respective Regions;

The eight technical commissions, composed of experts designated by Members, study matters within their specific areas of competence (technical commissions have been established for basic systems, instruments and methods of observation, atmospheric sciences, aeronautical meteorology, agricultural meteorology, marine meteorology, hydrology, and climatology);

The Secretariat, headed by the Secretary-General, serves as the administrative, documentation and information centre of the Organization. It prepares, edits, produces and distributes the publications of the Organization, carries out the duties specified in the Convention and other Basic Documents and provides secretariat support to the work of the constituent bodies of WMO described above.

______* On 30 November 2003. WORLD METEOROLOGICAL ORGANIZATION OPERATIONAL HYDROLOGY REPORT No. 42

METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

WMO-No. 813

Secretariat of the World Meteorological Organization Ð Geneva Ð Switzerland 2003 Cover photo: Australian Bureau of Meteorology; WMO.

© 2003, World Meteorological Organization

ISBN 92-63-10813-7

NOTE

The designations employed and the presentation of material in this publication do not imply the expression of any opinion whatsoever on the part of the Secretariat of the World Meteorological Organization concerning the legal status of any country, territory, city or area, or of its authorities, or concerning the delimitation of its frontiers or boundaries. CONTENTS

Page Foreword ...... v Summary (English, French, Russian and Spanish) ...... vii

1. INTRODUCTION ...... 1

2. METEOROLOGICAL MEASURING AND OBSERVING SYSTEMS ...... 4 2.1 Conventional measuring networks ...... 5 2.1.1 Synoptic measuring networks ...... 5 2.1.2 Climatological measuring networks and hydrometeorological networks ...... 8 2.1.3 Error sources and measuring accuracy ...... 9 2.2 Remote-sensing procedures ...... 10 2.2.1 Radar-based precipitation measurements ...... 11 2.2.1.1 Measuring procedure ...... 11 2.2.1.2 Error sources and measuring accuracy ...... 12 2.2.2 Satellite data for the estimation of precipitation ...... 13 2.2.2.1 Precipitation intensity from cloud information with visible and/or infrared data ...... 13 2.2.2.2 Precipitation intensity from microwave data ...... 14 2.2.2.3 Measuring accuracy ...... 15 2.3 Combined measuring systems ...... 15

3. DATA DERIVED FROM METEOROLOGICAL MEASUREMENTS ...... 17 3.1 Evaporation ...... 17 3.1.1 Introduction ...... 17 3.1.2 Definitions ...... 17 3.1.3 Determination of evaporation on the basis of the water balance equation ...... 18 3.1.4 Determination of evaporation on the basis of the energy balance equation and the water vapour transport ...... 19 3.1.5 Estimation of evaporation on the basis of data from the meteorological standard measuring network ...... 20 3.1.5.1 Computation procedures for evaporation from water bodies ...... 20 3.1.5.2 Computation procedures for potential evaporation ...... 21 3.1.5.3 Computation procedures for actual evaporation from areas of land ...... 22 3.2 Heavy rainfall point statistics ...... 24 3.3 Regionalization in hydrometeorology ...... 25 3.3.1 Introduction ...... 25 3.3.2 Cartographic representation of regionalized precipitation data ...... 27 3.3.3 Cartographic representation of regionalized heavy rainfall amounts ...... 27 3.3.4 Cartographic representation of regionalized extreme values of precipitation supply ...... 30 3.4 Probable maximum precipitation (PMP) ...... 30 3.4.1 General ...... 30 3.4.2 Estimation of PMP using a physically-based evaluation of meteorological data ...... 31 3.5 Snowmelt ...... 31

4. ESTIMATION OF AREAL PRECIPITATION AND AREAL EVAPORATION ...... 34 4.1 Fundamentals ...... 34 4.2 Methods for the computation of areal values ...... 34 4.3 Error sources in areal computation ...... 35 4.4 Computation of areal precipitation ...... 35 4.5 Areal evaporation ...... 37

5. QUANTITATIVE PRECIPITATION FORECASTING ...... 39 5.1 Precipitation ...... 39 5.2 Quantitative precipitation forecasting procedures ...... 40 5.2.1 Synoptic-statistical procedures ...... 40 5.2.2 Stochastic procedures ...... 42 5.2.3 Dynamic models ...... 43 5.2.3.1 General characteristics and survey ...... 43 5.2.3.2 System of numerical weather prediction ...... 45 5.2.3.3 Precipitation forecasting with the aid of the routine models GME and LM ...... 46 5.2.3.4 Outlook ...... 51 IV CONTENTS

Page 5.2.4 Dynamic-statistical model interpretation ...... 51 5.2.5 Remote-sensing procedures ...... 52 5.3 Need for, and requirements of, QPF on behalf of hydrology and water resource management ...... 53 5.4 Evaluation of procedures and their further development ...... 54

6. WORLD WEATHER WATCH ...... 57 6.1 Description ...... 57 6.1.1 Global Observing System (GOS) ...... 57 6.1.2 Global Data-processing System (GDPS) ...... 58 6.1.3 Global Telecommunications System (GTS) ...... 59 6.1.4 Future developments ...... 59 6.2 Demands of hydrology and water resources management on the WWW ...... 60 6.3 Pilot projects ...... 60

7. HYDROLOGY AND WATER RESOURCES PROGRAMME ...... 61 7.1 Hydrology and Water Resources Programme ...... 61 7.2 The Hydrological Operational Multipurpose System (HOMS) ...... 61 7.3 World Hydrological Cycle Observing System ...... 61 7.3.1 Introduction ...... 61 7.3.2 Organization and objectives ...... 62 7.3.3 WHYCOS stations ...... 62 7.3.4 Regional components ...... 63

References ...... 64 FOREWORD

The World Meteorological Organization (WMO) coordinates global scientific activity to allow increasingly prompt and accurate weather information and other services for public, private and commercial use. Its activities contribute to the safety of life and property and provide support for developing international programmes related to global climate and other environmental issues, and to sustainable development. The World Weather Watch, the core of the WMO Programmes, combines observing systems, telecommunication facilities and data- processing centres — operated by Members — to make available the meteorological, climatological and hydrological information needed to provide efficient services in all countries. Through this Programme, WMO Members coordinate the presentation of observed data and processed information. The Commission for Hydrology has long had an interest in the application of the World Weather Watch to hydrology and in the application of climatological data and climate information in water resources projects. In 1993, a report was published in German on the subject of meteorological systems for hydrological applications. The Commission for Hydrology suggested that this report should also be published in English. The report has therefore been brought up to date in keeping with technical developments and translated into English for wider distribution. It is with great pleasure that I express the gratitude of WMO to the persons who contributed to the preparation of this report and gave advice and suggestions in finalizing it. These include Ms H. Bartels, Messrs H. Dunke, D. Frühwald, E. Heise and U. Wacker of the German Weather Service (Offenbach), Mr A. Herrmann and Mr F.J. Löpmeier from the University of Braunschweig, Messrs H.J. Liebscher, G. Strigel and H.G. Mendel from the Federal Institute of Hydrology (Koblenz), Messrs G. Haase, A. Hense, C. Simmer and Ms S. Theis of the Metorological Institute of the University of Bonn and Prof. H. Fleer of the Geographisches Institut Ruhr of the University of Bochum.

G.O.P. Obasi Secretary-General

SUMMARY

To ensure the successful implementation of hydrological services influencing factors. These always present the challenge of deriving, and to provide adequate support for research and planning purposes, from point data, estimates of areal values which are valid for all there is a need for hydrological and meteorological data on various cases other than very small catchment areas or special cases. timescales. This report consists of seven chapters covering a wide Chapter 1 presents in detail the use of meteorological data in range of issues on the relationship between meteorological systems hydrology and water resource management. Chapter 2 discusses and their applications for hydrological purposes. The objective of meteorological measuring and observing systems, while Chapter 3 this report is to present the need for meteorological systems for presents data derived from meteorological measurements. Chapter 4 hydrological purposes. The National Meteorological Services discusses the estimation of areal precipitation and areal evaporation, (NMSs) operate different types of measuring networks for the and Chapter 5 is devoted to quantitative precipitation forecasting. assessment of meteorological data. Depending on the type of Chapter 6 focuses on the World Weather Watch (WWW), which is a network, data are available for hydrological purposes in the form of global system for the collection, analysis, processing and lists, in databases or in publications. The direct measurement of dissemination of weather and environmental information and thus hydrologically relevant climate data (model input data) yields point represents the very core of meteorological services. Chapter 7 values. The extent to which these are representative of the concentrates on the WMO Hydrology and Water Resources neighbouring region depends on the spatial homogeneity of the Programme (HWRP).

RÉSUMÉ

Pour assurer l’efficacité des services hydrologiques et être en toujours difficile d’obtenir, à partir de données ponctuelles, des mesure de fournir l’appui nécessaire dans les domaines de la estimations de valeurs zonales qui soient valables pour tous les cas recherche et de la planification, il faut disposer de données autres que ceux qui concernent de très petits bassins hydrologiques et météorologiques à diverses échelles temporelles. hydrographiques ou les cas spéciaux. Le présent rapport contient sept chapitres traitant de toute une série Le chapitre 1 traite de façon détaillée de l’utilisation de de questions ayant trait aux relations entre les systèmes données météorologiques dans le domaine de l’hydrologie et de la météorologiques et leurs applications à des fins hydrologiques. gestion des ressources en eau. Le chapitre 2 concerne le mesurage L’objectif visé est d’exposer la nécessité de systèmes météorologique et les systèmes d’observation, et le chapitre 3 météorologiques permettant de répondre aux besoins hydrologiques. présente des données obtenues à partir de mesures météorologiques. Les Services météorologiques nationaux (SMN) exploitent Le chapitre 4 porte sur l’estimation des précipitations et de différents types de réseaux de mesure pour les paramètres l’évaporation pour une zone donnée, et le chapitre 5 est consacré à météorologiques. Selon le type de réseau, des données sont la prévision quantitative des précipitations. Le chapitre 6 met disponibles à des fins hydrologiques soit sous forme de liste soit l’accent sur la Veille météorologique mondiale (VMM), qui, en tant dans des bases de données ou des publications. La mesure directe que système mondial de collecte, d’analyse, de traitement et de de paramètres climatologiques utilisables dans le domaine diffusion d’informations météorologiques et environnementales, est hydrologique (comme données d’entrée de modèles) permet la véritable pierre angulaire des services météorologiques. Enfin, d’obtenir des valeurs ponctuelles, qui ne pourront être dans le chapitre 7 il est question essentiellement du Programme représentatives de leurs régions avoisinantes qu’en fonction de d’hydrologie et de mise en valeur des ressources en eau (PHRE) de l’homogénéité spatiale des facteurs déterminants. A cet égard, il est l’OMM. VIII CONTENTS

HTP?VT

Для успешного обеспечения гидрологического обслуживания воздействующих факторов. Всегда существует проблема и предоставления адекватной поддержки для проведения расчета по данным в точке оценок значений по площадям, исследований и планирования необходимо иметь которые являются достоверными для всех случаев, за гидрологические и метеорологические данные различных исключением территорий небольших водосборов или особых временных масштабов. Настоящий доклад состоит из семи случаев. глав, охватывающих широкий диапазон вопросов, В главе 1 приводится подробная информация по касающихся взаимосвязи между метеорологическими использованию метеорологических данных в гидрологии и для системами и их применениями для гидрологических целей. водохозяйственной деятельности. В главе 2 излагаются Задачей настоящего доклада является предоставление вопросы, касающиеся систем метеорологических измерений и информации о потребностях в метеорологических системах наблюдений, в то время как в главе 3 приводятся данные в для гидрологических целей. Национальные метеорологические отношении расчетных величин, полученных по службы (НМС) эксплуатируют различные типы сетей метеорологическим измерениям. В главе 4 излагаются вопросы наблюдений для оценки метеорологических данных. В оценки осадков по площадям и испарения по площадям, а зависимости от типа сети, данные для гидрологических целей глава 5 посвящена количественному прогнозу осадков. В главе имеются в виде перечней, баз данных или в виде публикаций. 6 внимание концентрируется на Всемирной службе погоды Непосредственные измерения актуальных с точки зрения (ВСП), которая является глобальной системой для сбора, гидрологии климатических данных (входные данные для анализа, обработки и распространения метеорологической и моделей) содержат точечные значения. Степень, в которой они другой информации об окружающей среде и является основой являются репрезентативными для территории, на которой они для метеорологического обслуживания. В главе 7 говорится о находятся, зависит от пространственной однородности Программе ВМО по гидрологии и водным ресурсам (ПГВР).

RESUMEN

El éxito de la ejecución de los servicios hidrológicos y del estimaciones de valores de áreas válidos en todos los casos, adecuado respaldo a tareas de investigación y planificación exige exceptuando pequeñas cuencas o casos especiales. contar con datos hidrológicos y meteorológicos a diferentes En el Capítulo 1 se da una descripción pormenorizada del escalas cronológicas. Este informe consta de siete capítulos en que empleo de datos meteorológicos en la hidrología y la gestión se abordan gran número de cuestiones que tienen que ver con la de los recursos hídricos. El Capítulo 2 presenta los sistemas de relación entre los sistemas meteorológicos y sus aplicaciones para medición y observación en meteorología. En el Capítulo 3 se fines hidrológicos. Los Servicios Meteorológicos Nacionales reflejan los datos derivados de esas mediciones. En el Capítulo (SMN) tienen en funcionamiento diferentes tipos de redes de 4 se abordan los tema de la estimación de la precipitación de medición destinadas a la evaluación de datos meteorológicos. área y de la evaporación de área. El Capítulo 5 está dedicado a Dependiendo del tipo de red, podría disponerse también de datos la predicción cuantitativa de la precipitación. El Capítulo 6 para fines hidrológicos, disponibles en listas, bases de datos o versa sobre la Vigilancia Meteorológica Mundial (VMM), que publicaciones. La medición directa de datos climáticos que tienen es un sistema mundial para la recogida, el análisis, el interés para la hidrología (datos para ingresar en los modelos) se procesamiento y la difusión de información sobre el tiempo y traduce en valores puntuales. La medida en que esos datos son el medio ambiente y, por consiguiente, representa el elemento representativos de la región circundante depende de la central de los servicios meteorológicos. El Capítulo 7 se centra homogeneidad espacial de los factores incidentes. Siempre se en el Programa de Hidrología y Recursos Hídricos (PHRH) de presenta la dificultad para derivar, a partir de datos puntuales, la OMM. CHAPTER 1

USE OF METEOROLOGICAL DATA IN HYDROLOGY AND WATER RESOURCE MANAGEMENT

To implement their tasks, Hydrological Services, as well as research In hydrology and water resource management, a large and planning institutes (consultancies), need, not only hydrological number of mathematical simulation models are used for the data, but also meteorological data based on different timescales. For planning and operation of water resource systems. While for planning purposes, the need is mainly for non-real-time planning purposes non-real-time values are needed, for operational meteorological data, usually in the form of climatological data sets. purposes real-time data are required. The simulation models used However, forecasts and warnings, as well as the operation and are precipitation-discharge models. In the case of models simulating control of water projects, require real-time data, usually in the form exclusively flood discharge, only precipitation data are included. In of synoptic data sets. areas in which a snow cover forms for a longer time period during Climatological data are made available by Meteorological winter, snow cover and temperature data are also used. In models Services in the form of point values for stations. These are then used for the simulation of low-flow discharge or discharge hydrographs to calculate various time-based means — such as daily, monthly or over a longer time period, evaporation has to be taken into account. annual averages, frequency or extreme-value distributions and areal In this case, the meteorological data required for the calculation of analyses — in the form of maps derived from point measurements. evaporation are needed. The same data are required for models However, hydrological and water resource management often need simulating groundwater recharge. All of the above-mentioned areal values for use in calculations involving runoff. Only in a few applications require areal data (Chapter 4). cases are such values directly assessed, e.g. radar-measured The models for the simulation of the water quality in lakes precipitation (Chapter 2). For this reason, a conversion of point and rivers require data on radiation, air temperature and atmospheric values into areal values is usually required (Chapter 4). humidity. These data are also used for simulating the temperature In developing hydrological models, less and less river behaviour of surface water. For models simulating ice formation on catchment basins are considered directly. It is preferable to lakes and rivers, data on atmospheric temperature are also needed. concentrate on deriving result representations in the form of grid Unlike in the above-mentioned cases, the applications refer only to points, the value assigned to the grid point being representative of a the immediate river course. The river course itself is usually located corresponding area within the grid. This method has the advantage in a valley having its own moisture regime. In most countries, the that, with the aid of the grid, any areal divisions can be produced. meteorological networks are located in places that are as The meteorological data should, therefore, be available as input data representative of a region as possible. For this reason, there are in the form of a grid point data set. hardly any meteorological stations in the immediate vicinity of a That requirement is met with meteorological data of river course. It is also for this reason that the data from the climate numerical prediction models. However, the use of hydrological stations of Meteorological Services are usually not simply models for planning also requires the derivation of climatological transferable to the river courses and their immediate surroundings. data sets determined over many years for grid fields. The The Hydrological Services, therefore, often have to perform their meteorological values of conventional measuring networks own meteorological measurements along the river course in order determined by derived point measurements must be converted into to obtain the required data. grid points in the light of orography, exposure, vegetation cover, For the design of structures such as dams, dikes, form of settlement, etc. reservoirs, sewage treatment plants and urban drainage systems, Hydrology relies on precipitation values provided by data on the occurrence probability of extreme events are required. Meteorological Services. Usually, such values are needed in the This is possible with the aid of extreme-value considerations form of areal values with different temporal resolutions. For the (Subchapter 3.2). For the computation of flood probabilities, apart plausibility control of measured hydrological data (e.g. for from discharge data, measurement data and estimations of extreme discharges), precipitation data are often required. Many Services rainfall are also used. Probable maximum flood (PMF) is often carry out a comparison of all the values of the area to test their computed from probable maximum precipitation (PMP) using discharge values. A balance along a river stretch covering all simulation models (Subchapter 3.4). measured discharges of the main stream, as well as those of its For certain designs (e.g. urban drainage structures), the tributaries, is established using mathematical models. If there are no required data can be obtained directly from extreme-value statistics measurement data of the inflow from tributaries, then they are such as storm statistics. For the computation of low-flow probabili- simulated through precipitation-discharge relationships. The same ties, except for precipitation depths, data on the duration of dry periods procedure is used to fill gaps in series of observations. (periods without rainfall or with abundant rainfall) are also needed. Surveys of water resources must be made for many water In the case of all the above-mentioned applications where resource plans and for master plans. In most cases, this is done by non-real-time data are concerned, a maximum of measured data can assessing directly areal runoffs, by areal precipitation amounts or be made available for the required computations. This does not by establishing complete water balances. Since a complete areal apply in the case of operational forecasting (e.g. flood forecasting). coverage of evaporation measurements is not possible to this day, In that case, mathematical models are used only with data available it is usually calculated from climatological data — e.g. radiation, at that time for forecasting. Only a small number of stations of the air temperature, air moisture, sunshine duration, wind, etc. meteorological observation networks (synoptic measuring network) (Chapter 3). are equipped to provide direct access to that data. This is true, in 2 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

Table 1.1 Hydrological time series, parameters, methods and procedures needed in water management planning activities

Hydrological and meteorological Water management sector Methods and procedures Elements Influencing factors

Water resource planning Master plans Discharge Means, minimum, maximum, frequency, duration, Precipitation mapping techniques Evaporation Groundwater Water management plans Discharge Means, minimum, maximum, frequency, duration, Water pollution groundwater budget Precipitation Dissolved matter Environment impact statements Dissolved matter Means, minimum, lowest 30-day mean Bed load Suspended load Public water supply (drinking water and Discharge Means, minimum, maximum, frequency, duration processed water for industry and domestic use Dissolved matter Surface water (storage reservoirs, river water) Water temperature Sub-surface water Groundwater level Permeability Means, minimum, maximum, frequency, duration, Infiltration Pore volume groundwater budget Precipitation Grain distribution Dissolved matter Springs Delivery Means, minimum, lowest 30-day mean Dissolved matter Power generation Water power River power stations Discharge Slope Means, minimum, maximum, extreme value, Suspended load statistics, frequency, duration, lowest 10-day mean Sediment transport Dissolved matter Storage reservoirs Discharge Means, frequency, duration, extreme value statistics, Suspended load complete water balance Sediment transport Evaporation Thermal power Discharge Means, minimum, frequency, duration extreme value Water temperature statistics, lowest 5-day mean Dissolved matter Air temperature Flood control Flood retention basin Discharge Maximum, extreme value statistics, forecasting Accumulated runoff Precipitation Storage reservoirs Discharge Means, maximum, extreme value statistics, Accumulated runoff forecasting Dams, dikes Water level Navigation Water level Means, minimum, maximum, extreme value Suspended load statistics, frequency, duration, lowest 30-day Sediment transport mean, forecasting techniques Dissolved matter Irrigation Precipitation Plant properties Frequency, duration, extreme value statistics, Evaporation Infiltration complete water balance Infiltration Capacity Soil moisture Field capacity Runoff Wilting point Drainage Sub-surface drainage Groundwater level Transmissivity Means, maximum, frequency, duration Soil moisture Field capacity Precipitation Wilting point Evaporation Urban drainage Precipitation Intensity, duration, maximum, extreme value statistics, PMP Water pollution control Dissolved matter Means, maximum, minimum, frequency, Discharge intensity, duration, extreme value statistics, Radiation lowest mean of several days Air temperature Precipitation Fishing Dissolved matter Means, minimum, maximum Suspended load Recreation, leisure, sports Dissolved matter Means, minimum, maximum Discharge frequency, duration CHAPTER 1 — USE OF METEOROLOGICAL DATA IN HYDROLOGY AND WATER RESOURCE MANAGEMENT 3 particular, for precipitation stations. As areal data are needed, Tables 1.1 and 1.2 list the various application areas of precipitation data measured by radar or derived from satellite hydrology and water resource management and their demands on measurements are desired (Subchapters 2.2 and 2.3). Moreover, meteorological data. Today, hydrological models are not yet quantitative precipitation forecasting (QPF) values are required as available with all time and space scales. Figure 1.1 shows the time input data for hydrological forecasts used for warnings issued to the and space scales of mathematical models. population and for the operation of water plants (Chapter 5).

Table 1.2 Application of meteorological data in hydrology

Field of application Hydrological element needed Meteorological element needed Type of meteorological input data Timescale Spacescale

Data processing, Runoff Precipitation d, m s, a plausibility check of hydrological data Water balance Runoff Precipitation y, m, d a (non-real-time) Evaporation Radiation d, m s, g Soil moisture Sunshine duration1 d, m s, g Groundwater Air temperature1 d, m s, g Air humidity1 d, m s, g Windspeed1 d, m s, g Simulation of time series Runoff Precipitation y, m, a (non-real-time) Groundwater Radiation d, m s, g Water temperature Air temperature h, d, m s, g Dissolved matter2 Air humidity Extreme value statistics of floods Runoff Precipitation y, m, d s and low flow (non-real-time) Water level min, max Forecasting (real-time) Runoff Precipitation h, d s, a Water level Snow cover, water d s, a Equivalent of snow, Snowmelt Radiation h, d s Soil moisture Air temperature h, d s s = Point values a = Areal values g = Grid values y = Annual values m = Monthly values d = Daily values h = Hourly values min = Minimum values max = Maximum values 1 Ð Meteorological elements needed for the calculation of evaporation. 2 Ð Averages of selected weather situations (dry and wet weather conditions).

Area (km2)

107

106 Macroscale models Macroscale 105

104

103 Mesoscale models

102 models Flood forecasting

1 10 Waterbalance models

100 Process-oriented 10-1 watershed models 10-1 100 101 102 103 104 t (d)

1d 1mon 1a 30a Figure 1.1 — Time and space scales of mathematical models. CHAPTER 2

METEOROLOGICAL MEASURING AND OBSERVING SYSTEMS

The National Meteorological Services (NMSs) operate different differentiations have been made. Thus, the mesoscale range types of measuring networks for the assessment of meteorological has been subdivided into the scales meso γ and meso α and the data. Depending on the type of measuring network, the data are small-scale range includes the scales meso γ, micro α, micro β and available for hydrological purposes immediately (real-time) or only micro γ. after a certain period has expired (non-real-time) in the form of lists, Requirements for the global measuring network are laid in a database or in publications. down and coordinated by the Commission for Basic Systems Depending on the meteorological objectives set and on the (CBS); requirements for the regional measuring network by the six description of phenomena, synchronous observation (synoptical Regional Associations (RAs) and Antarctica (WMO, 1981b); and method) — or the observation at the same position of the sun requirements for the national measuring network by the Weather (climatological method) — is predominant globally. However, Services of the various Member States (WMO, 1994a). following the progressive automation of measuring equipment, Accordingly, each station of the global measuring network normally these differences are becoming partly indistinct. Physical meets the requirements both of the regional and the national descriptions of meteorological phenomena fall within different measuring network. It should be added that many global categories of data. Figure 2.1 shows a schematic representation of meteorological phenomena can be encountered on the planetary as the characteristic length and timescales for different important forms of motion and processes in the atmosphere (Fortak, 1971). Planetry The shading indicates the inherent energetics. The small circles Forms of motion waves in the atmosphere Cyclone 106 represent in each case the characteristic speed of 10 m sec-1 to be 10 waves observed on the convective, meso, and partly large scale. The field 105 1d Cloud delimited at the top right approximately outlines the processes Cumulomnibus cluster 10 convection covered explicitly by circulation models. The relationships between 104 size of area, meteorological phenomenon, prediction model and 1h Small-scale turbulence convectionCumulus prediction period must be taken separately for extratropical and 103 10 tropical systems from Table 2.1. In this table, convective and small 102 scales have been combined. The result is almost unavoidably a 1 min Gravity waves subdivision into global, regional and national measuring and Characteristic timescale (sec) 10 observing systems, which are distinguishable according to their Mesoscale Sound waves observation and measurement data, network density, accuracy and 1 frequency of observations (Tables 2.2 and 2.3). In this connection, Small scale Convective scale Large scale attention should be given to the fact that areal extents to be included 1cm 10 1m 10 102 1km 10 102 103 104 in numerical prediction models should be larger, the longer the Characteristic length scale scheduled prediction period. At the same time, it is clear that in the Figure 2.1 — Characteristic length and timescales for forms of motion modelling field of numerical weather prediction (NWP), general and processes in the atmosphere (Fortak, 1971).

Table 2.1 Scales of meteorological phenomena and prediction models (WMO, 1987)

Extratropical systems and relevant prediction models in the 1990s Tropical systems and relevant prediction models in the 1990s

Scale System Prediction Prediction Scale System Prediction Prediction model period model period Planetary scale Rossby waves General Long-range Planetary scale Monsoon Global Long-range (> 5 000 km) circulation Medium-range (> 5 000 km) Hadley call Statistical Medium-range model Walker call techniques Global ITCZ Large scale Anti-cyclones/ Global Medium-range Large scale Monsoon Global Medium-range (1 000Ð5 000 km) extratropical Hemispheric Short-range (1 000Ð5 000 km) depression Fine-mesh Short-range cyclones Easterly wave Limited area MesoscaleÐ Squall lines Global Medium-range MesoscaleÐ Tropical Fine-mesh Short-range (100Ð1 000 km) Fronts Fine-mesh Short-range (100Ð1 000 km) cyclone Limited area Very-short-range Polar lows Limited area Very-short-range Squall lines (moveable area) (vorticity (nested) Cloud cluster maxima) Small scale Thunderstorms Boundary-layer Short-range Small scale Thunderstorms Boundary-layer Very-short-range (< 100 km) Sea breeze and mesoscale Very-short-range (< 100 km) Sea breeze and mesoscale Nowcasting Cumulus models and nowcasting models CHAPTER 2 — METEOROLOGICAL MEASURING AND OBSERVING SYSTEMS 5

Table 2.2 Global observation and measurement data (WMO, 1981b)

Basic set of global observation data required and to be met by the GOS by the late 1990s* (both in situ observations and remotely-sensed data)

Horizontal resolution Vertical resolution Observational error (rms) Frequency of observation Upper-air 50 km (A) 10 layers in troposphere 0.5Ð1¡C in troposphere 2Ð4 per day temperature 5 layers in stratosphere 1Ð2¡C in stratosphere Upper-air wind 250 km 10 layers in troposphere 1Ð2 m sÐ1 in troposphere 2Ð4 per day vector (V) 5 layers in stratosphere 2Ð3 m sÐ1 in stratosphere Upper-air relative 250 km 4 layers 10 per cent 2Ð4 per day humidity (RH) Sea surface 250 km 0.5¡C with systematic Instantaneous difference measurements wrong observating systems averaged over 3 days eliminated on 3 days average Surface pressure (P) 250 km +/Ð 1 hPa 4 per day temperature (T, Td) +/Ð 0.5¡C temperature wind vector (V) +/Ð 1Ð2 m sÐ1 State of surface ** ** ** ** Satellite imagery (B) At leas 3 km horizontal At least 3 layers-low, To be determind; will be 8 per day resolution of imagery middle, high and cloud function of latitude for top height geostationary satellites

(A) Tropics: 500 km resolution sufficient temperature. (B) Satellite imagery included here because of its use in computing vertical motion and divergence fields, as well as for determination of synoptic distribution of water vapour, precipitable water and cloudiness. * This table defines a basic set of the global observational data requirements which generally can be met by the GOS and, therefore, should be used in the design and imple- mentation of the GOS. ** Includes precipitation, soil moisture, soil temperature, emissivity, albedo, snow and ice coverage. Resolution, accuracy and frequency not yet determined; information required from other commissions.

Table 2.3 Regional measuring networks. Requirments for horizontal spacing and frequency of reporting from the regional network (WMO, 1981b)

Density Minimum for sparsely Type of observation Adequate populated and oceanic Frequency areas Land surface 250 km 300 km 8 per day, at the main and intermediate standard times

Oceanic surface 250 km 500 km 4 per day, at the main standard times

Surface-based 250 km 1 000 km 2 to 4 per day, at the main standard times with priority upper-air to 0000 and 1200 UTC 1 to 2 per day in the tropics at 0000 and/or 1200 UTC well as the large scale, just as regional meteorological phenomena (b) Synoptic measuring networks; develop on the large scale and mesoscale, while dynamic feedbacks (c) Other synoptic measuring networks (ship, buoy and aircraft between the phenomena of the various scales are taking place. stations); Moreover, as the meteorological values show a different variability (d) Weather radar networks; in space and time, there are further differences within the individual (e) Satellite observation. measuring networks as a result of the temporal spacing of Since the weather processes in the free atmosphere observations and the areal distribution of the stations. develop on a considerably larger scale than those in the ground- The increasingly applied remote-sensing techniques close atmosphere, considerably fewer measuring stations are represent a good addition to the point measurements in conventional required for the aerological measuring network than for the synop- measuring networks, because of their direct areal assessment. tic measuring network. Thus, about 1 000 aerologic stations and about 10 000 synoptic stations are in operation globally. Within 2.1 CONVENTIONAL MEASURING NETWORKS the global real-time data exchange, however, at present not all 2.1.1 Synoptic measuring networks these stations reach the respective regional centres (Figures 2.2 The actual situation of atmospheric conditions required for weather and 2.3). forecasting is usually carried out globally at the same time, according At the aerological stations, the vertical profile (up to about to universal time coordinated (UTC), with the assistance of: 30 km) of the following meteorological data is determined and fed (a) Aerological-synoptic measuring networks; into a global telecommunication system in coded form: 6 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

180 W 160 W 140 W 120 W 100 W 80 W 60 W 40 W 20 W 0 20 E 40 E 60 E 80 E 100 E 120 E 140 E 160 E 180 E

80 N 80 N

60 N 60 N

40 N 40 N

20 N 20 N

0 0

20 S 20 S

40 S 40 S

60 S 60 S

80 S 80 S

180 W 160 W 140 W 120 W 100 W 80 W 60 W 40 W 20 W 0 20 E 40 E 60 E 80 E 100 E 120 E 140 E 160 E 180 E Figure 2.2 — Survey of globally-agreed synoptic ground stations: Deutscher Wetterdienst (DWD); Global Precipitation Climatology Centre (GPCC). Observing stations which, according to WMO, are to transmit SYNOP messages (total number 6 924) (Source: Weather Reporting, Volume A, WMO-No. 9).

180 W 160 W 140 W 120 W 100 W 80 W 60 W 40 W 20 W 0 20 E 40 E 60 E 80 E 100 E 120 E 140 E 160 E 180 E

80 N 80 N

60 N 60 N

40 N 40 N

20 N 20 N

0 0

20 S 20 S

40 S 40 S

60 S 60 S

80 S 80 S

180 W 160 W 140 W 120 W 100 W 80 W 60 W 40 W 20 W 0 20 E 40 E 60 E 80 E 100 E 120 E 140 E 160 E 180 E

Figure 2.3 — Survey of synoptic stations participating in global data exchange including precipitation: Deutscher Wetterdienst (DWD); Global Precipitation Climatology Centre (GPCC). Observing stations which have transmitted SYNOP precipitation messages in June 1987 (4 343 out of a total of 6 924) (Source: Weather Reporting, Volume A, WMO-No. 9).

(a) Atmospheric pressure; The general requirements to be met by a meteorological (b) Air temperature and atmospheric humidity; station, be it on land, on water or in the atmosphere, as well as the (c) Wind direction and speed. requirements to be met by the measurement of various The measurements and observations made at synoptic meteorological elements are laid down in the following WMO ground stations are likewise exchanged through internationally- documents: agreed codes and include global, regional and national key groups (a) Technical Regulations (WMO-No. 49); for various meteorological data as exemplified by Germany. They (b) Guide on the Global Observing System (WMO-No. 488); are listed in Table 2.4. (c) Manual on the Global Observing System (WMO-No. 544); Table 2.4 — Global, regional and national key groups with additional national climate data; FM 12-SYNOP land station (Source: Deutscher Wetterdienst) Paragraph 0 (global code groups) Paragraph 1 MiMiMjMj YYGGiw IIiii iRixhVV Nddff (OOfff) 1snTTT 2snTdTdTd 3POPOPOPO 4PPPP 5appp 6RRRtR 7wwW1W2 8NhCLCMCH Air temperature in 1/10 T.3845 Symbolic figure Sign of the data in 1/10 hPa Dew point temperature Symbolic figure in 1/10 hpa Pressure at station level Symbolic figure in 1/10 hpa Pressure at station level Symbolic figure tendency Characteristic of 3 hours pressure tendency in 1/10 hPa Amount of 3 hours pressure Symbolic figure T.3590 Amount of precipitation (1 = 6, 2 12 hours) obsevation before time of 6-hour interval T.4667 T.551 Symbolic figures Present weather Past weather Symbolic figure (= AAAX) Kind of information Part of information T.1855 Data of the month (UTC) the nearest whole hour (UTC) to Actual time of observation, wind speed Indicator for source and units of Block number Station number Indicator for 6RRRt T.0513 C Total cloud cover with T.0509 T.0515 Low clouds Middle high clouds High clouds ufc T.4377(7wwwW T1600 operation Indicator for type of station lowest cloud seen of the Height above surface surface Horizontal visibility at Total cloud cover Wind direction in tens of degrees T.3845 Wind speed Symbolic figure Sign of data L or C N 1 W 2 T.1880 ) R ° C T.2582 VuB T.1819 T2700 1

Paragraph 2 (maritime global-, regional code groups) Paragraph 3 222Dsvs OsnTwTwTw 2PwPwHwHw 333 1snTxTxTx 2snTnTnTn 3EsnTgTg 4E’sss (5j1j2j3j4) (6RRRtR)8NsChshs 9SpSpspsp hpT.0700 exchange for maritime data global Symbolic figure group ship Displacement of the Ships average speed Symbolic figure T.3845 Sign of data in 1/10 temperture Sea surface Symbolic figure in seconds Period of wind waves waves in units of 5 m Height of the wind in 1/10 above the ground temperature 5 cm T.0975 Minimum of Symbolic figure cover measurable ice with snow or State of the ground Total depth of snow Symoblic figure T.2061 tion Indicator T.2061 informa- Supplemtary information supplementary Specification relating to Symbolic figure Amount of precipitation observation 6 hours before time of Symbolic figure T.2700 Amount of C-cloud T.1676 T.0500 Genus of cloud cloud layer t.0668 Height of base Symbolic figure RA VI Special phenomena for in 1/10 T.0901 Minimum temperature Symbolic figure cover measurable ice with snow or State of the ground T.3845 Sign of data in 1/10 Maximum temperature Symbolic figure T.3845 Sign of data exchange) (data for regional Symbolic figure group Symbolic figure T.3845 Sign of data ° ° ° ° C C C C T.3889 T.2700 T.4451

Paragraph 4 (national code groups) Paragraph 5 444 N’C’H’H’Ct 555 OTbTbShSh 1R1R1R1sR 2snTmTmTm 3PoPoPoPo 4SSSS 5s’s’s’tR 7RRRzR 911ff 912ff PIC INp BOT hesnTTT for data national use Symbolic figure group Amount of C-cloud Genus of cloud of C-clouds surface Altitude of the upper Top of cloud the ground in 0 height of 5 cm above Air temperature at a of observation hour preceding the time min.) during the last Duration of sunshine (in Symoblic figure the time of observation the last hour preceding which has fallen during Amount of precipitation Indicator of precipitation Symbolic figure Sign of data the day of observation of the day preceeding temperature (in 1/10 Daily mean of air Symbolic figure (without thousands) height in 1/10 hPa Air pressure in barometer Symoblic figure day of observation the day preceeding radiation in joule cm T.3880 Daily sun of the global Symbolic figure snow Depth of newly fallen observation 6 hours before time of Symbolic figure Amount of precipitation oservation hours before time of Time of 12 intervals Symbolic figure hour before observation (> 21 kn) during the last Maximum wind speed Code number hour before observation (> 21 kn) during the last average of wind speed Maximum 10 minute cover of mountains Code word for cloud mountains Cloud cover of ground temperature of the Code word for Depth of measurement T.2845 Sign of data ground in 1/10 Temperature of the Symbolic figure group Symbolic figure ° ° T.3590 T.0029 C C – ° 2 C) of

Paragraph 5 (additional climatological data) 555 80000 OsnTxkTxkTxk 1snTnkTnkTnk 2snTgTgTg 3Es’s’s’ 4E’sss 5sTgsTgUrUr 6RRRR 7wtwtwtWR 8fnTfTfTf 9NLVLVLL data for national use Symbolic figure group for 07, 14 21 UTC observation follow,observation time of measurements and climatological data of supplementary T.3845 Symbolic figure group: Symbolic figure Sign of data measurement until 21 UTC (day of before day of measurement) 1/10 T.3845 Maximum temperature in Symbolic figure Sign of data (day of measurement) measurement) until 21 UTC (day before day of T.3845 Minimum of temperature, Symbolic figure Sign of data until 07 UTC T.0901before day of measurement) 1/10 cm above the ground in Minimum of temperature 5 Symbolic figure cover snow or measurable ice State of ground without T.70 07 UTC time of observation Depth of newly fallen snow, Symbolic figure 21 UTC 07, 14, time of observation or measurable ice cover State of ground with snow T.3889 Total depth of snow Symbolic figure measurement 07 UTC ground, time of minimum thermometer at Depth of snow above the 21 UTC of the air in %, 07, 14, Registered relative humidity Symbolic figure 1/10 mm, 07, 14, 21 UTC Amount of precipitation in T.88 Symbolic figure of observation T.85 Present weather during time of precipitation T.21Indicator to group 6/, form Symbolic figure temperature) Sign of data (wet-bulb 1/10 Wet-bulb temperature in Symbolic figure Total cloud cover at point L at the point LT.4377surface Horizontal visibility at (number) Point of observation ° ° ° C period: 21 UTC (day C period: 21 UTC (day C T.3889 T.70 T62

90000 OVAVAVBVB 1VCVCfkfk 2snTmTmTm 7R24R24R24R24 912fxkfxk 94SSS 99sss 9RRRR 9’RRR measurement) 07 UTC (day of measuremment follow at day before of climatological data of the supplementary Symbolic figure group: T.82 Symbolic figure the day before Precipitation fallen during during the day before T.84 Precipitation deposited Symbolic figure day before phenomena during the Other weather during the day before Maximum wind speed T.3845 Symbolic figure Sign of data the day before temperature (in 1/10 Daily mean of air Symbolic figure measurement)) before) + 07 UTC (day of mm (14 + 21 UTC (day 24-hour period in 1/10 precipitation during the Total amount of Code number (00 21 kn of the day before average of wind speed > Maximum 10 minute Code number T.3889 the day before in 1/10 h Duration of sunshine Symbolic figure number Depth of snow cover code Symbolic figure before) in 1/10 mm equivalent of snow (day Total amount of water Symbolic figure average in 1/10 mm snow depth of 1 cm on Water-equivalent of a – 24 UTC) in kn T.83 T.84 T.20 ° C) of 8 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

(d) Guide to Climatological Practices (WMO-No. 100); Associations are preparing a survey under the World Climate (e) Guide to Meteorological Instruments and Methods of Data Information Referral Service (INFOCLIMA) (WMO, Observation (WMO-No. 8); 1985a). Joined to the global data exchange are the monthly (f) Manual on Meteorological Observing in Transport Aircraft CLIMAT messages supplying recent monthly data of (WMO-No. 197). atmospheric pressure, air temperature and humidity, precipitation depth and sunshine duration as well as the related long-term 2.1.2 Climatological measuring networks and monthly reference values (WMO, 1971). Globally, there is a total hydrometeorological networks of 2 200 stations from which at the utmost 1 500 messages are As a result of the reaction of ground-near atmospheric layers to regularly received by the meteorological regional centres incoming and outgoing radiation conditions, these stations — (Figures 2.4 and 2.5). established to supplement the global synoptic measuring network A large number of raingauge stations are operated in — often have measuring dates at the same position of the sun or are national hydrometeorological or hydrological networks. equipped with recording measuring instruments. In this connection, There is little information on the national special climate and precipitation stations should be mentioned as well as measuring systems available to WMO Member countries. The special measuring networks important for hydrological problems, number of precipitation stations (one daily measurement of the such as: precipitation depth) pursuant to information from the Hydrological (a) Global radiation (direct and diffuse radiation); Information Referral System (INFOHYDRO) worldwide is (b) Soil moisture; 150 000, that of precipitation recording stations (continuous (c) Wind direction and speed; recording of precipitation events) about 41 000, and that of (d) Precipitation (recorder); evaporation stations 10 600. (e) Snow cover depth; The measuring systems of the Deutscher Wetterdienst (f) Water equivalent of snow cover; have been compiled in Table 2.5 to give an idea of the (g) Sunshine duration; respective network density, the observation frequencies (h) Phenological observations. and the data access time. As for various hydrological fields of The general requirements to be met by the climatological application, the areal distribution of the real-time stations is often observation network as well as the requirements for location, data too coarse-meshed; special services are being established and measuring accuracy, operation and maintenance of climate permitting a quasi real-time retrieval of data sets, e.g. water stations are described in the Guide to Climatological Practices resources management reporting service, warnings in the case of (WMO-No. 100), in the Technical Regulations (WMO-No. 49) and snowmelt, etc. in the Guide to Hydrological Practices (WMO-No. 168). Other special measuring systems include, for instance, Recommendations for the establishment and operation of weather radar stations, atmospherics detection stations (lightning agrometeorological stations can be taken from the Guide to counter), rocket stations, ozone-sounding stations, background air Agricultural Meteorological Practices (WMO-No. 134). pollution stations, planetary boundary-layer stations and tide- At present, no survey exists stating how many climate gauge stations. The establishment and operation of stations are stations are in operation globally. However, the Regional described in WMO, 1978; 1989 and 1996.

180 W 160 W 140 W 120 W 100 W 80 W 60 W 40 W 20 W 0 20 E 40 E 60 E 80 E 100 E 120 E 140 E 160 E 180 E

80 N 80 N

60 N 60 N

40 N 40 N

20 N 20 N

0 0

20 S 20 S

40 S 40 S

60 S 60 S

80 S 80 S

180 W 160 W 140 W 120 W 100 W 80 W 60 W 40 W 20 W 0 20 E 40 E 60 E 80 E 100 E 120 E 140 E 160 E 180 E Figure 2.4 — Survey of globally-agreed CLIMAT stations: Deutscher Wetterdienst (DWD), Global Precipitation Climatology Centre (GPCC). Observing stations (total 2 201) which are to transmit monthly climatological means from ground stations (Source: Weather Reporting, Volume A, WMO-No. 9). CHAPTER 2 — METEOROLOGICAL MEASURING AND OBSERVING SYSTEMS 9

180 W 160 W 140 W 120 W 100 W 80 W 60 W 40 W 20 W 0 20 E 40 E 60 E 80 E 100 E 120 E 140 E 160 E 180 E

80 N 80 N

60 N 60 N

40 N 40 N

20 N 20 N

0 0

20 S 20 S

40 S 40 S

60 S 60 S

80 S 80 S

180 W 160 W 140 W 120 W 100 W 80 W 60 W 40 W 20 W 0 20 E 40 E 60 E 80 E 100 E 120 E 140 E 160 E 180 E Figure 2.5 — Survey of CLIMAT stations participating in global data exchange in 1987: Deutscher Wetterdienst (DWD); Global Precipitation Climatology Centre (GPCC). Observing stations which in 1967 regularly transmitted monthly CLIMAT messages (total 1 017 stations out of 2 201 offical — CLIMAT observing stations) (Source: Weather Reporting, Volume A, WMO-No. 9).

2.1.3 Error sources and measuring accuracy arise in computing a global water balance. Apart from a large The measuring accuracy and reporting interval for meteorological number of random errors occurring in precipitation measurements, data required for hydrological purposes is given in Table 2.6. the operation of measuring instruments installed above ground Since the number of requirements for precipitation causes, in particular, the following errors: measurements in the operational and in the research areas of (a) Wind field deformation; meteorology, in surveying climate changes, in hydrology and in (b) Suspended water losses; ecology is continuously increasing, and since precipitation (c) Evaporation losses; represents the only input into the global water balance, problems (d) Splash water caused by rebounding raindrops.

Table 2.5 Frequency of observation, density of stations and availability of data of the different network of stations (Deutscher Wetterdienst) Measuring networks Observation frequency Station density Data acces (number of stations) Aerological stations 9 Six-hourly measurement 1:40 000 km2 Real-time according to UTC Synoptic stations 176 One-hourly measurement 1:2 000 km2 Real-time according to UTC partly continuous Climatological stations 595 Daily time 00, 06, 12, 18 partly 1:600 km2 Non-real-time continuous partly real-time Precipitation stations 3 500 Daily time 07:30 partly 1:100 km2 Non-real-time continuous partly real-time Special measuring networks Global radiation 59 Continuous 1:6 000 km2 Non-real-time, real-time Wind 60 Continuous 1:6 000 km2 Non-real-time, partly real-time Rainfall recorder 460 Continuous 1:780 km2 Non-real-time Water equivalent 511 Three times a week at a fixed time 1:700 km2 Non-real-time partly real-time Sunshine duration 164 Continuous 1:2 200 km2 Real-time partly non-real-time Phenological observing 2 020 Daily during vegetation period 1:180 km2 Non-real-time network partly real-time 10 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

Table 2.6 The first two errors are very important as they concern the Measuring accuracy and observation frequency of loss of volume. As these physical processes depend on instrumental meteorological values for hydrological purposes (WMO, 1998c) parameters, on the one hand, and on climate factors, on the other, Reporting interval the magnitude of the systematic error in precipitation measurements Element Precision for hydrological varies considerably depending on the type of raingauge used, the forecasting purposes region and the season of the year (WMO, 1982b). About 150 Precipitation: ± 2 mm below 40 mm Six hours(2) different types of instruments are in use worldwide. The result is (1) amount and form ± 5 per cent above 40 mm that jumps occur in the isohyets at the national borders of Snow depth ± 2 cm below 20 cm Daily neighbouring countries which use different types of raingauges ± 10 per cent above 20 cm (Figure 2.6). At present, different methods for the correction of Water equivalent ± 2 mm below 20 mm Daily errors are being applied. Investigations show that such corrections of snow ± 10 per cent above 20 mm may be used for the long-term mean, but that they cannot be Air temperature ± 0.1¡C Six hours generally used for individual events, in particular not for heavy Wet-bulb ± 0.1¡C Six hours rainfall (Subchapter 3.2). temperature It is recommended to apply corrections to the data used in Net radiation ± 0.4 JmÐ2 dÐ1 budget studies and analyses, especially in high geographical below 8 M JmÐ2 Daily latitudes. It is not necessary to correct heavy rainfall data. The Ð2 Ð1 ± 5 per cent above 8 M Jm d correction of data and the method used should be indicated if Pan evaporation ± 0.5 mm Daily corrected data or products are distributed. Surface temperature: ± 1¡C Daily snow 2.2 REMOTE-SENSING PROCEDURES Temperature ± 1¡C Daily Remote-sensing procedures — satellite and radar techniques which profiles: snow have become operational in the recent past — represent an addition Wind: speed ± 10 per cent Six hours to the already very old techniques of precipitation determination by direction ± 10¡ means of weight and volume measurements. These remote-sensing Sunshine duration ± 1 hour Daily procedures do not permit, it is true, direct precipitation Relative humidity ± 1 per cent Six hours measurement, but offer the advantages of areal coverage and real- ______time transmission. Satellite techniques at present are mainly drawn (1) In some locations, it will be necessary to distinguish the form of the from the cloud-top temperatures (passive procedure). In using radar- precipitation (liquid or solid). based techniques, conclusions are drawn from the backscattering (2) The reporting interval in flash flood basins is often required to be two measured (active procedure). Considering their physics, both hours or less; in other locations, daily values may suffice. measuring methods are particularly suited for the interpretation of

9 10 11 40 41 42 43 15 16 55 20 21 22 23 56

47 57 24 27 48 49 38 39 30 31 32 33 58 44 34 35 36 46 28 29 7 54

2 1 5 53 8 6 4 50 45 51 17 37 12 52 18 13 59 19 14 Figure 2.6 — Map of basic and evaluation stations, with numbers 1Ð52 in rings and 53Ð59 in squares, respectively; one point means one or more nearby stations. Indicated are correction factors for the systematic error in precipitation measurement (WMO, 1982b). CHAPTER 2 — METEOROLOGICAL MEASURING AND OBSERVING SYSTEMS 11 the areal phenomena of precipitation (WMO, 1982a, 1985b). Both 5 cm) compared to the diameter of the scattering particles (so-called methods are already in operational use, satellite techniques being hydrometeors), then the relationship between backscattered energy increasingly used for precipitation estimations over areas for which (meaning measured received energy), emitted energy (including no data would otherwise be available. The parameters for antenna parameters) and the cross-section of backscattering of the atmospheric conditions derived from satellite data are, however, less particles can be represented in a simplified form by the so-called accurate than the in situ measurements of conventional networks. Rayleigh approximation. This area of backscattering covers the The determination of precipitation over large areas with individual particles with their diameter raised to the power of 6. the aid of remote-sensing procedures was first used only for That term is usually designated as radar reflectivity factor Z. The application in minimum-term forecasts (up to a few hours after the radar equation consequently links the incoming energy (PE) with time of assessment). Radar-based techniques are operational for the instrumental constant (C), the goal distance (r) and the hydrological uses, inter alia, in Europe, Japan and the United States reflectivity factor (Z) as follows: (WMO, 1988a, 1988b), particularly for flood forecasting. PE = C K2 r2 Z (2.1) As already mentioned, precipitation measurements based on radar and satellite imagery are not to be considered as an The factor K2 originates from the Rayleigh approximation alternative to conventional ground-based measurements but, rather, and contains the complex refraction index. Consequently, it is as complementary. Figure 2.7 gives a survey of comparison different for the different states of aggregation (water: K2= 0.93; dry networks which are taken as a basis for the Global Precipitation snow: K2 = 0.208). The reflectivity factor Z is computed via the Climatology Project of the World Climate Research Programme radar equation from the measured incoming energy (PK). (WCRP) for the calibration of satellite-based data and consist of The requested precipitation intensity (R) (mm h-1) is precipitation measuring networks and combined radar and linked via the so-called Z/R relationship with the reflectivity (Z) precipitation measuring networks (WMO, 1988a). (mm6 m-3) which has been determined. The general form of the Z/R relationship reads as follows: 2.2.1 Radar-based precipitation measurements 2.2.1.1 MEASURING PROCEDURE Z = ARB (2.2) The determination of precipitation by means of a weather radar set is based on the location of precipitation areas with the aid of A and B are a function of the drop spectrum and may vary electromagnetic waves, analogous to the location of precipitation from one precipitation to another. In practice, mean Z/R areas from aircraft or ships, on the one hand, and the measurement relationships depending on the respective precipitation type of that fraction of transmitting energy backscattered from the obtained from long-term drop spectra measurements are used. In the precipitation area, on the other hand. In physics, the MIE theory middle latitudes, however, it usually suffices to distinguish between gives a general description of the scattering of electromagnetic two main precipitation types, e.g. precipitation occurring as a result waves on spherical particles in space. If, as in the case of a radar- of warm and cold air advection. A computer determines the based measurement, the wavelength is long enough (usually about instantaneous precipitation intensity via the Z/R relationship fed in, N °

9 N70

° 8 7 3 N50 ° 13 4 10 16 N30 1 15 ° 5 20 12 21 18 19

S10 11 ° 2 14 17 6 S10 °

PRIMARY SITES 6 South Brazil SECONDARY SITES

S30 1 Thailand 7 Canada 11 Indonesia 16 Puerto Rico ° 2 N. Australia 8 United Kingdom 12 Thailand 17 South Brazil 3 Japan 9 Sweden 13 South Japan 18 Ivory Coast 4 Florida 10 Israel 14 N. Australia 19 Kenya 5 Caribbean 21 Kwajalein 15 Honduras 20 India S50 ° 70 90°E 120°E 150°E 180°W 150°W 120°W90°W60°W30°W300° °E60°E90°E

Figure 2.7 — Reference measuring networks for the calibration of satellite-based estimations (WMO, 1988a). 12 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES it extrapolates that relationship for the scanning time interval and error. Dual-polarization radars measure the reflectivity depo- finally obtains the resulting hourly or event precipitation depths for larization ratio, which is useful for indicating the presence of the respective area. In this case, the resolution is limited by the areal hail. Dual-wavelength radars, also designed for hail detec- element over which the integration is made, usually 1 kilometre, 1 tion, usually use an S-band and an X-band radar together degree. At high rain rates (> 70 mm h-1), differential phase (i.e. 10 cm and 3 cm wavelengths, respectively); techniques (Joe, 1996) are more accurate than a Z/R approach (d) The error occurring as a result of the attenuation of the radar (Chandrasekar et al., 1990). It is insensitive to hail in a rain-hail wave on its passage through areas of heavy rainfall is practi- mix, independent of system calibration, independent of rain cally negligible in temperate latitudes and for the attenuation, beam blockage and beam filling since these factors do wavelengths (e.g. C-band) used; not affect the differential phase shifts. To improve the result (e) More serious are the cases in which the radar beam cuts the obtained by radar, some good ground stations are often compared melting zone at a low or a decreased 0¡C limit. Such bright- with the related areal elements in the radar-based result in order to band effects lead, in the melting zone range, to considerable derive correction factors. The data output can be adjusted to the data echo increases. Smith (1986) devised an analytic technique user’s requirements via the peripheral instrumental equipment of the for reducing errors due to the bright-band, which proved computer. Representation on a coloured monitor, isohyetal plots, very promising in tests. In routine operation, however, it was digital printouts, storage on magnetic tape and supply of data with found to be susceptible to large errors, when significant vari- the aid of remote data transmission are possible. Various ations in the freezing level occurred over the area covered by possibilities of hydrological applications are described in Collinge a radar. Rosenfeld, Amitai and Wolff (1995) derived a and Kirby (1987). The contribution of radar data to hydrological number of radar echo classification criteria which may be forecasting is demonstrated in Collier (1996). employed to recognize the occurrence of bright-band effects. Further work is needed to assess the operational reliability of 2.2.1.2 ERROR SOURCES AND MEASURING ACCURACY this technique (Collier, 1996); The main error sources in radar precipitation measurements are the (f) Disturbing and unpleasant are the ground returns which are various states of aggregation and the respective sizes of the not immediately distinguished by the radar from precipita- precipitation particles (Figure 2.8): tion echoes and are, therefore, likewise converted into (a) If the precipitation particles are very large in the radar precipitation intensities. The so-called clutter data sets which volume, then the Rayleigh approximation no longer applies; are stored in the computer can help for fixed target echoes (b) If a Z/R relationship other than that adjusted to the type of (mountains and buildings). However, this procedure is no precipitation is used, then the error rate in the conversion of help in the case of temporary ground returns resulting from the echo power into precipitation intensity can be up to 50 abnormal propagation conditions (extreme ranges). For per cent; ground clutter, the frequency spectrum shows a peak at zero (c) If the state of aggregation of hydrometers in the measuring frequency and becomes flat for frequencies higher than 20- volume with regard to time and/or with regard to space is 30 Hz. By discriminating between theses frequency spectra partly different from the one assumed (e.g. sleet or snow using a high-bandpass filter and estimates of the mean power storages), then the error through K2 must be added to the Z/R of the rain echo, ground echoes may be separated from rain

CONTINENTAL NUCLEI ICE NUCLEI MARITIME NUCLEI WATER VAPOUR WATER VAPOUR WATER VAPOUR

Nucleation Nucleation Nucleation condensation condensation condensation

CIRRUS SECONDARY SEEDING ICE PARTICLES

NARROW CLOUD SPECTRA ICE CRYSTALS BROAD CLOUD SPECTRA

Slow broadening Vapour by coalescence Break-up Heterogeneous deposition Heterogeneous Coalescence freezing freezing

FROZEN DROPS FROZEN DROPS SNOW CRYSTALS DRIZZLE ICE PELLETS

Chumping

SECONDARY Riming Riming SECONDARY ICE PARTICLES Chumping ICE PARTICLES SNOW FLAKES RIMED CRYSTALS RIMED FLAKES RIMED FLAKES SNOW PELLETS GRAUPELS Riming

Partial Partial Continued GRAUPELS melting melting Continued coalescence coalescence BRIGHT BAND BRIGHT BAND Wet and dry riming with Heterogeneous drops and Heterogeneous freezing crystals Melting Melting freezing

RAIN SLEET GRAUPELS HAIL SNOW RAINSNOW GRAUPELS RAIN SLEET RAIN (WARM) GRAINS SNOW PELLETS (WARM) SMALL HAIL Figure 2.8 — Aggregate states of hydrometeors and microphysical processes in precipitation formation (WMO, 1988a). CHAPTER 2 — METEOROLOGICAL MEASURING AND OBSERVING SYSTEMS 13

echoes (Collier, 1996). A compilation of standard clutter (a) Precipitation as a result of large-scale dynamic processes filters is presented by Meetschen (1999). However, clutter Promising methods for the assessment of precipitation inten- information can also be used for calibration, attenuation sities from fronts or large-scale cloud fields are based on the correction and determination of the refraction index. fact that, for precipitation activity, the air must be subjected The potential errors (a) to (d) above can, for the most part, to a large-scale lifting process which is triggered off by be corrected by calibration procedures with the aid of ground-based convergence in low atmospheric layers and divergences in measurements, while for errors (e) and (f), correction procedures the upper atmosphere. The divergence in the upper atmos- are more complicated and still being developed. phere usually becomes apparent from the horizontal Users again and again ask to what degree the precipitation extension of the cloud surface in time, the change in the area measurements by radar are accurate. It is difficult to answer that being proportional to the vertical speed in an upward direc- question as there is no real reference value. Comparison of the value tion. Estimation of the precipitation intensity results from an measured by radar with that measured by a ground-based raingauge evaluation of the development cycle of large-scale cloud always implies, in addition to potential errors in the ground-based structures (life history method). In this context, apart from value (Subchapter 2.1.3), the problem that a receiving area of the form of the clouds and its change in a horizontal direc- 200 cm2 is compared with a volume of about 0.5 km3. Such tion, information on the synoptically-conditioned air mass comparisons were routinely made in Great Britain with the result character must be included (Neil and Spagnol, 1987). Most that, for calibrated radar values, the quotients of the hourly methods with visible or infrared data require an interactive precipitation depths were for more than a year within factor 2 range. subjective treatment of the satellite images and are therefore A more important comparison in hydrological practice is hardly suited for an automatic derivation of the precipitation the comparison of areal precipitation depths computed from point depth under operational conditions. Since satellite images measurements with those derived from radar measurements. The areal taken at frequent time intervals are needed, data from the precipitation depths routinely computed in Germany use the catch- geostationary satellites may be taken for the application of ment basin procedure (Subchapter 4.4) — i.e. areal precipitation the “life-history” methods (Figure 2.9); depths for hydrologically-fixed areas of basic areas (size: about 100 (b) Precipitation as a result of convective processes km2) are computed over catchment basins up to entire river basins. Most methods for the assessment of precipitation from satel- Radar-based precipitations for individual basic areas supply negative lite data yield a relatively good application for convective as well as positive differences during the summer months — between clouds. Ð30 and +35 per cent — as compared to the computed areal precipi- For convective precipitation events, the basic assumption tation depths. If the reference area is enlarged, i.e. if several basic areas is that precipitation probability is larger the colder the temperature are combined to form a catchment basin of the size of 1 500 km2, then at the cloud tops. The cloud surface temperature is determined from the deviation is reduced to a negative difference of 5 per cent the radiation values in the atmospheric window channel (infrared (Deutscher Wetterdienst, 1986). channel). At present, in many countries, radar networks are being In the United States, a procedure has been developed established. The main objectives of such networks — quantitative to estimate the intensity of precipitation from the cloud height measurements, qualitative images — are, however, very different. But via its surface temperature and from its vertical and horizontal it may be assumed that the assessment and determination of precipi- changes. tation will be further improved and applied with the aid of radar This method supplies good results for large, convective techniques within the next few years. Globally, at present, about 600 cloud clusters in tropical air masses and is generally used in the radar stations are in operation (with an increasing tendency). United States for the forecasting of so-called “flash floods”, i.e. flood events setting in very suddenly (Scofield and Oliver, 1977). 2.2.2 Satellite data for the estimation of precipitation It is more difficult for central Europe to use this In principle, two methods for the assessment of precipitation depths method because the very rigorous boundary conditions are from satellite data must be distinguished: fulfilled only in a very few cases. It has nevertheless been possible (a) Derivation of the precipitation intensity from cloud informa- to estimate approximately in individual cases the precipitation tion with visible and/or infrared data; and depth for southern Germany by means of that method (Krüger, (b) Derivation of the precipitation intensity from microwave 1983). However, estimations of precipitation depths from satellite data. data at present still have an error rate of 500 per cent in individual In the visible wavelength range, information about cloud cases. thickness, structure and composition can be taken from satellite The European Space Operations Centre (ESOC) at images; in the infrared range, the cloud height is derived indirectly Darmstadt regularly derives the so-called “precipitation index” via the radiation temperature of the top layer of the cloud. The two from the European Organization for the Exploitation of wavelength ranges can normally be used separately, but a Meteorological Satellites (EUMETSAT) series of meteorological combination of both (bispectral method) leads to better results. geostationary satellites (METEOSAT) data. It is mainly Microwave data techniques use the characteristics of based on the height of clouds calibrated by ground-based absorption/emission of raindrops and the spreading of ice particles. measurements via regression equations. The index is a mean over five days each and is derived for areas from 32 × 32 pixels. The 2.2.2.1 PRECIPITATION INTENSITY FROM CLOUD INFORMATION evaluation area for the precipitation index is situated between 40 WITH VISIBLE AND/OR INFRARED DATA degrees north and 40 degrees south, and between 50 degrees west Precipitation intensity from cloud information with visible and 50 degrees east (Turpeinen, Abidi and Belhaouane, 1986). A and/or infrared data includes: similar method is applied by the National Oceanic and 14 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

Atmospheric Administation (NOAA) for the estimation of low and the signal of the rain cloud can barely be distin- precipitation from geostationary operational environmental guished from its surroundings. This is the main disadvantage satellite (GOES) data. of the low-frequency approach. At higher frequencies, absorption/emission and scattering by precipitation particles 2.2.2.2 PRECIPITATION INTENSITY FROM MICROWAVE DATA increase. This leads to a strong non-linear signal dependence Precipitation intensity from microwave data includes: on the precipitation intensity and to saturation at decreasing (a) Passive high-frequency microwaves precipitation intensity with increasing frequency. This non- At frequencies above 60 GHz, the interaction intensity linear relation is a serious problem, because rain cells are between radiation and precipitation has its maximum. typically smaller than the resolution of satellite radiometers Raining clouds are almost opaque; absorption scattering by (e.g. 30 to 50 km for SSM/Imager). Compared to the both liquid and frozen precipitation-size particles is impor- methods described above, the low-frequency approach is tant. Accordingly, raining clouds appear cold compared to based on a more direct relation between signal and precipita- the environment. The radiance depression is dependent on tion intensity. This method is, however, restricted to oceanic the amount of scatterers in the upper part of the cloud. Since areas (Simmer, 1996); ice particles play a key role in precipitation generation, the (c) Rain radar precipitation intensity can be estimated from the observed Active microwave techniques are designed to be used by the radiance depression. The microwave reflection and emission equipment of a satellite with a precipitation radar of water surfaces is highly polarized in contrast to emission (Subchapter 2.2.1) in the Tropical Rainfall Measuring and scattering by atmospheric constituents. As a result, in Mission (TRMM) together with various microwave radiome- addition to the radiance depression, the polarization reduc- ters, as well as a radiometer operating in the visible and tion serves as another piece of information over ocean areas. infrared band. In TRMM, the vertical distribution of precipi- The main advantage of the high-frequency method is its rela- tation will be measured over the tropics in a band between tive independence of the signal of the surface below the ±35 degrees in latitude. Such information will greatly cloud. Thus, the signal can be observed over both land and enhance the understanding of the interactions between the ocean surfaces (Simmer, 1996); sea, air and land masses which produce changes in global (b) Passive mid- to low-frequency microwaves rainfall and climate. TRMM observations will also help Below 40 GHz, raining clouds become increasingly transpar- improve modelling of tropical rainfall processes and their ent for microwave radiation; below 10 GHz, emission influence on global circulation leading to better predictions dominates and the emitted radiation is almost proportional to of rainfall and its variability at various timescales. Although the rain water content in the atmospheric column. If the this, too, promises further progress in precipitation estima- height of the water column and its profile structure is known, tion from satellite data, it will be possible to meet only part then the precipitation intensity can be retrieved. As above, of the global requirements for precipitation data. additional information is provided by polarization reduction. In the microwave spectral region, the radiances are Over land, the polarization of the background radiation is directly related to the hydrometeors in the satellite field of view.

210° 240° 270° 300° 330° 0° 30° 60° 90° 120° 150° 180° 90° 90°

60° 60°

30° 30°

N N GOES-W GOES-E METEOSAT GOES GMS 0° + + + + + 0°

Latitude S USA USA EUROPE USSR JAPAN S

-30° -30°

-60° -60°

-90° -90° -150° -90° -30° 30° 90° 150° W 0° E Longitude Figure 2.9 — Planned global geostationary weather satellite system (ESA, 1981). CHAPTER 2 — METEOROLOGICAL MEASURING AND OBSERVING SYSTEMS 15

Contrary to the visible and infrared spectral region, the whole state covered by the various measuring systems of the conventional of the atmosphere described by the vertical profiles of temperature, measuring network, the ground-based radar measurements and water vapour, cloud and rain liquid and ice profiles, and the shape of satellite data. Thus, in Europe a comprehensive set of data is the particles determine the outgoing radiances at the top of the available through the dense measuring networks of conventional atmosphere. So any parameter retrieval is at least partially an stations, ground-based weather radar and METEOSAT satellites. It inversion of the radiative transfer equation. Concerning rain, is possible to use the more accurate system for the calibration of the basically three paths have been followed to construct rain retrieval next higher system, i.e. point measurements can be used for the algorithms (complete inversion, statistical inversion and indexing) determination of radar-based precipitation depths, which, in turn, (Simmer, 1996). can be used for the derivation of satellite estimations. In some countries, combined systems are already 2.2.2.3 MEASURING ACCURACY operational, e.g. the Forecasting Rain Optimized using New As the methods presently available for the estimation of Techniques of Interactively Enhanced Radar and Satellite precipitation from satellite data cover a large range of the (FRONTIERS) system in Great Britain (CEC, 1988). In this electromagnetic spectrum, it is of particular importance to collect system, visible and infrared images from METEOSAT are investigations on measuring accuracy and possibilities of combined with data from the weather radar network and data application. This has been done in Table 2.7 on the basis of the of the conventional measuring network. Figure 2.10 gives an literature published (EUMETSAT, 1988). Two essential points are example of this combined measuring system. However, it was found revealed by this table and should be noted: that, particularly when heavy rainfall systems occurred, the (a) The majority of the errors indicated are referred to an inte- forecaster was unable to step through the system procedures rapidly gral area of more than 1 000 km2, although some of the enough. Hence, a new fully automatic system, Nimrod, was methods are also suited for areas of a size of up to about 50 proposed (Collier, 1991). This system, in which radar and satellite km2. The major part of the investigations deals with the data are blended with mesoscale model data, has recently been determination of convective rainfall and permits no conclu- implemented (Golding, 1995). Further details of the analysis sion as to frontal rainfall; procedures being developed are described by Kitchen, Brown and (b) It should further be taken into account that the ground-based Davis (1994). measurements of the precipitation depth used for comparison A graphical representation of the presumed accuracy of are not performed with uniform accuracy (Subchapter 2.1.3), precipitation measurements from radar and satellite systems as a that the point measurements must be transferred to the area function of their respective reference areas is given in Figure 2.11. and that some instruments are equipped only for the record- First, it is striking that the two measuring systems are quite ing of heavy rainfall. complementary. The upper limit of area sizes for the assessment of quantitative radar data is about 10 000 km2 and the lower limit of 2.3 COMBINED MEASURING SYSTEMS the area sizes for the use of satellite data is about 1 000 km2. It can Over data-rich areas of the Earth, the total timescale of the further be noted that the differences of error rates from calibrated precipitation measurement (from minutes to years) and the space radar-based precipitation measurements within and outside the scale (from one point to an area of about 500 000 km2) can be melting zone range are, on average, from about 20 to 30 per cent.

Table 2.7 List of measuring accuracies of precipitation estimations from satellite data as a function of reference area and period of integration (Collier, 1996) Technique Areas over which Period of Approximate Sample references describing estimates are assessed integration percentage accuracy techniques (rainfall types) (km2) (h) (%) Cloud indexing 105 24 122 Follonsbee and Oliver (1975); Adler and (with cloud model) 1041/2 41 Negri (1988) (convective/stratiform) Area average 105 Inst- 30 × 24 15 Atlas and Bell (1992) (convective) Life history 104 1 85 Griffith et al. (1978) (convective); 105 24 55 Arkin and Meisner (1987) (convective); 1041/2 50 Wylie (1979) (convective) 6 × 1031/2 65 Stout et al. (1979) (convective) 105 30 × 24 10 Xie and Arkin (1995) (convective)

Bi-spectral 1051/2Ð2 49 Lovejoy and Austin (1979a,b); convective/frontal) Passive microwave 103 24 70 Lovejoy and Austin (1980) (convective/frontal) (Wilheit et al. (1973); Spencer et al. (1983) 105 30 × 24 10 Xie and Arkin (1995) (convective) Active 103 12 20 (when combined The accuracy of this technique is unknown with bi-spectral but Lovejoy (1981) suggests the figures given technique) may be possible (see also Atlas and Bell (1992)) technique 103 30 × 24 monthly 10 Simpson et al. (1988) (convective)

Summary of the performance of satellite estimation techniques (from Collier, 1985 partly based upon Lovejoy and Austin 1979, 1980) 16 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

Figure 2.10 — European composite from METEOSAT and radar-based data (COST 73) (CEC, 1988).

In other areas of the world, limitations in the assessment Likely limit of of precipitation events are due to the absence of one or several of satellite data 100 these measuring systems. In this context, precipitation Cloud indexing measurements with a time resolution of less than one day should be 80 Life history mentioned. In particular, over tropical oceans, no ground-based passive microwave 60 measurements are available. In such areas, only satellite-based In bright-band Bi-spectral estimations can be used to obtain derivations of daily or monthly 40 precipitation values. No bright-band Limit of quntitative 20 radar data from Active single site microwave 0 Percentage error of hourly measurements error Percentage 101 102 103 104 105 106 Area of integration (km2)

Ground-based Satellite techniques radar echniques

Figure 2.11 — Areal coverage of radar- and satellite-based systems for the determination of areal precipitation depths and respective measur- ing accuracy (EUMETSAT, 1988). CHAPTER 3

DATA DERIVED FROM METEOROLOGICAL MEASUREMENTS

3.1 EVAPORATION necessitates the correction of the systematic error of the 3.1.1 Introduction precipitation measurement, because a smaller amount of Evaporation is the transition of water from the soil or ground precipitation is measured by the measuring instruments than reaches surface, from the vegetation cover and from open areas of water into the surface of the ground at the site under observation. The amount the atmosphere. The evaporation losses of the Earth's surface occur of correction depends on the type and exposure of the instrument as a consequence of various processes in which, apart from and on the climatic conditions at the site. The evident annual run of meteorological conditions, the specific characteristics of the the corrected values is important for the water balance. different surfaces play a part. At the Deutscher Wetterdienst, precipitation correction is Evaporation is a component not only of the energy routinely carried out according to the Richter (1995) procedure for balance equation (Equation 3.1) but also of the water balance hydrological and water management problems. equation (Equation 3.2). This means that the connection between Evaporation represents a quantity of loss in the water the heat balance and the water balance of the Earth’s surface is balance. On average, 75 per cent of the precipitation water made through evaporation. With a high energy supply, the amount evaporates from the Earth’s land masses. In Central Europe, this of evaporation is limited by the water supply available for the figure is 60 to 80 per cent (62 per cent in Germany). Even in the evaporation process, and vice versa: if a high water supply is light of considerable regional and seasonal differences, these figures available, then the maximum amount of water that can evaporate is show the extraordinary significance of evaporation in the restricted by the amount of energy available. hydrological cycle. The energy balance equation (in W m-2) is: In the Hydrological Atlas of Germany (BMU, 2000) the water balance components for the climatological normal value Rn + G + H + LE = 0 (3.1) period of 1961Ð1990 and further additional information on with: hydrography, soil, land use, etc. relevant to hydrological and water Rn: radiation balance; management problems are charted. In DVWK (1996) a summary of G: Soil heat flux, or for bodies of water; change in heat the procedures used in Germany to measure and compute content; evaporation with special consideration given to the requirements H: Sensible heat flux; arising from hydrological and water management practice is LE: Latent (evaporation) heat flux with: presented. L: Evaporation heat in W m-2 mm-1; The following descriptions of methods of measuring and E: Depth of evaporated water in mm. computing evaporation concentrate on the procedures used in All four components can be shown as positive or negative Germany. values; the evaporation heat flux is usually negative. Further balance components such as, for example, heat flux from the biomass, 3.1.2 Definitions energy transformation due to chemical processes are, for The total evaporation consists of the following elements: hydrological considerations, generally so small to be negligible. (a) Evaporation (physical processes): However, under certain weather conditions, the advective energy Evaporation from the Earth’s surface not covered with vege- input reaches a relevant order of magnitude in local energy tation (bare soil, snow and ice areas), from precipitation balances. retained on plants, buildings, roads, etc. (interception evapo- Depending on the time of year and weather situation, a ration) and from open water surfaces; considerable part of the radiation energy reaching the surface of the (b) Transpiration (plant-physiological processes): Earth, up to 100 per cent of the net radiation, can be expended on Evaporation from plants. evaporation. The evaporation process on the Earth’s surface is, The term evapotranspiration, i.e. the sum of evaporation therefore, an important energy supplier for the atmospheric and transpiration, combines the partial processes of evaporation processes and it is necessary to take this into account in numerical normally taking place simultaneously on a certain area of the Earth's forecasting models and climate models. surface. In hydrology and water resources management, the areal The water balance equation (in mm) is: evaporation — the average value, for example — over the area of a catchment basin with its different kinds of land use, is of special P + E + Q + ∆W = 0 (3.2) interest (see Subchapter 3.5). with: In the scientific literature on evaporation modelling, the P: Precipitation (including correction of the systematical terms potential evaporation and actual evaporation are frequently measurement error of the instrument); used. However, the terms are not always used clearly or E: Evaporation; consistently. In this report, the following definitions are used (DIN Q: Runoff (surface and groundwater runoff); 4049, 1994; DVWK, 1996): ∆W: Change in water supply in the period under consideration. (a) Potential evapotranspiration: evaporation from an area of Note concerning the water balance equation: The use of natural vegetation with given meteorological conditions and measured data for the amount of precipitation in Equation 3.2 an unlimited supply of water in the soil. On account of this 18 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

prerequisite, the potential evaporation is a fictitious value as the advective component of the heat balance. Therefore, computed from meteorological data that evaluates sites, differences in the temporal course (daily pattern, annual pattern) as regions and periods of time according to their evaporation- well as in the amount of evaporation losses can be noted between climatic conditions. the various types of pans. Pan evaporation is used as a relative value For an unlimited supply of soil moisture, the (maximum) for estimating the evaporation from natural waters or the potential amounts of evaporation from plants more or less deviate from evaporation adapted to the characteristics of the instruments and the potential evapotranspiration. The extent and signs of the devia- evaporation climate. With raft evaporation pans, there is an tions depend on the kinds of plants and their phenological stage approximation between the heat balance of the pan water and that of of development. In order to differentiate exactly the quantities the lake. These evaporation values are used for the derivation and used in evaporation modelling, it seems appropriate to introduce the parameterization of evaporation formulas for the evaporation the term maximum evapotranspiration, a parameter that is char- from bodies of water (Richter, 1969). acteristic of plants, as an additional term alongside potential evapotranspiration, the meteorological parameter. Lysimeter (non-weighing and weighing of the most different (b) Actual evapotranspiration: evaporation from an area of construction types) (DVWK, 1980; WMO, 1994a) natural vegetation with given meteorological conditions These instruments measure the actual or, in the case of water- without the above restriction of an adequate supply of water saturated soil, the maximal evapotranspiration (including to the plants, i.e. with the actual soil water supply. interception evaporation) of the plant crop and the soil substrate. In addition, the precipitation should be measured in a precipitation 3.1.3 Determination of evaporation on the basis of the gauge. water balance equation Non-weighing lysimeters (percolation lysimeters) are Because the evaporation process depends on the specific used as small instruments (grass-covered) with comparatively characteristics of the individual kinds of surface, a direct simple measuring techniques as relative instruments. Another measurement of evaporation from a heterogeneously used area is important field of application for non-weighing instruments is the not possible. Only an estimation of the areal evaporation average of use of large lysimeters for water balance investigations of forest many years can be obtained fairly easily from the difference sites for which, due to the depth of the root zone, the installation of between the areal precipitation and runoff, if the storage element a weighing lysimeter poses a problem. The evaporation is ∆W in the water balance equation (Equation 3.2) becomes determined from measured values of the amount of precipitation negligible and if there are reliable runoff data from the catchment and the percolation as a long-term mean value assuming a negligible area (Subchapter 3.5). All direct measuring procedures determine change in water supply ∆W in the soil (see Equation 3.2). the amount of evaporation from individual, defined kinds of The installation, running and maintenance costs of surfaces and are exact only for these. weighing lysimeters are very high and are therefore only used for Listed below are some of the most common measuring research purposes, e.g. for the derivation and parameterization of instruments and procedures that are used for solving hydrological computation procedures for evaporation for hydrological practice. and water resources problems or for the verification and validation Modern precision scales allow a high-time resolution including data of evaporation models. on the daily pattern of the evaporation. The vegetation around the lysimeter should corresponds Atmometers/atmographs (Piche atmometer, Czeratzki disc) to that of the lysimeter over the largest possible area. This requires, The atmometers/atmographs use filter paper or porous ceramic discs for example, a homogeneous agricultural or forestry cultivation. as transducers. Therefore, a realistic link to evaporation processes The most important causes of errors in lysimeter on natural surfaces does not exist. The evaporation losses from these measurements lie in the accumulation of seepage water on the floor surfaces, which are constantly kept humid, correspond neither to the of the lysimeter, the vertical flow of water along the wall of the evaporation from natural surfaces nor to the potential evaporation. lysimeter, disturbance of the soil structure and growth disturbances Because of their relatively simple measuring techniques, such of the plant crop. These problems can be controlled by taking great instruments are used to record relative values of evaporation. When care in designing, installing and maintaining the lysimeters (Lützke, using the required correction procedures it should be taken into 1965; Olbrisch, 1975). account that these are only valid for the site and climate conditions At stations with weighing lysimeters and large for which they have been derived. comprehensive lysimeter measurements, observation programmes are normally carried out to record meteorological data, soil Evaporation pan (Class A pan, land and raft evaporation pan, characteristics and plant crop characteristics (phenological phases, GGI-3000) crop density, cultivation methods, yields, etc.). These The evaporation pans mostly used worldwide also on a network complementary observation programmes are of prime importance scale differ in construction (size, material) and method of for the generalization of lysimeter measurements as the evaporation installation (standing on the ground surface, embedded in the soil data are only valid for given meteorological, soil and plant or set up on an anchored raft on a lake). The evaporation losses are conditions. determined by measuring the changes in water level or in weight, or by measuring the quantity of water refilled, taking precipitation Determination of evaporation with hydrographs of the soil water into consideration (separate measurement with a precipitation content gauge) (DVWK, 1990; WMO, 1994a). This is a method generally used for scientific investigations that The construction and installation of the pan influence can be carried out at changing sites (unlike weighing lysimeters). considerably the heat balance of the body of water in the pan as well Tensiometers, neutron probes or time-domain-reflectometry CHAPTER 3 — DATA DERIVED FROM METEOROLOGICAL MEASUREMENTS 19

(TDR) probes are used to measure the vertical profiles of the soil water vapour (Equation 3.5) or heat and of impulse in the water content. On the basis of the water content profiles, the corresponding formulation: ρ ∂ Ð ∂ upward-moving (evaporation) water flow in the soil is separated E = Ð KE q/ z (3.5) from the downward-moving (percolation) (Malicki et al., 1992). This method also gives evaporation values for individual sites. For Here the evaporation E is given as water vapour flow a generalization of the results, additional comprehensive (per area and time); according to the measuring of the specific observations of the meteorological, soil and vegetation conditions humidity q (temporal mean value) at two altitudes the differential are required. quotient ∂q/∂z is replaced by the difference quotient. What makes the evaluation of this equation problematic is the determination of 3.1.4 Determination of evaporation on the basis of the the diffusion coefficient like KE (Equation 3.5), which depends on energy balance equation and the water vapour subsoil characteristics, on altitude z, on wind velocity and on the transport stability of the atmospheric stratification. The influence of the The methods described below for the determination of evaporation ground requires a sufficiently large, homogeneous area as a suit- from meteorological data are mainly used for research purposes able measurement site. The results of extensive experimental tests because they require a great deal of technical measurements and into the impact of subsoil roughness and into atmospheric stratifi- place heavy demands on the instruments and the processing of the cation have been presented in the last decades (Monteith, 1976; measured values. The results are used, inter alia, for parameteri- Foken, 1990). zations of the evaporation processes on the basis of meteorological data from the standard measuring network. Energy balance method Evaporation is calculated as the remainder of the energy balance Turbulence correlation method (eddy-flux, eddy-correlation) (Equation 3.1) after all the other components have been determined (Swinbank, 1951) by measurement or computation. The evaporation is determined from the turbulent vertical There are special measuring instruments (e.g. balance water vapour flow in the layer of air near the ground. The method meter, pyranometer, etc.) and computation methods (e.g. Ångström requires synchronous measurements of the air density, vertical wind equation for global radiation) for the determination of the radiation velocity and specific humidity of the air at only one altitude. The balance Rn or its long-wave or short-wave components. Because of considerable efforts required for measurement can be reduced by the empirically-derived factors, the use of such computation the evapotron first presented by Dyer and Maher (1965). methods has to be limited to the conditions of the radiation climate Nevertheless, the eddy-flux method is not one of the standard for which they have been specially developed. DVWK 1996 methods for measuring evaporation. Roth (1989) recommends using contains a list of the radiation formulae suitable for special this method from a low-flying aeroplane to determine areal means application in evaporation models in Germany. for selected hours. The soil heat flux G is measured with the help of heat flux plates or is computed from measured values, possibly available on a Gradient methods network scale, of the soil temperature at various depths with the The exchange procedures according to Sverdrup (1936) and help of the heat conductivity equation. For certain problematic Thornthwaite and Holzmann (1942) require measurements of the positions and site conditions the soil heat flux can be considered vertical profiles of meteorological data. negligible. Some equations estimate the soil heat flux as a According to Sverdrup, the energy balance equation percentage of the radiation balance values (see Hoyningen-Huene, (Equation 3.1) is modified by way of the so-called Bowen ratio β 1983a; DVWK, 1996). (Bowen ratio method): For bodies of water, the change in heat content, i.e. the β γ = H/LE = (T2 Ð T1)/(e2 Ð e1) (3.3) temporal change in the amount of heat stored in the body of water, comparable to the soil heat flux of the ground, is determined by with γ = 0.65 hPa K-1 (psychrometer constant). measuring the vertical distribution of the water temperatures. In This ratio is derived from the vertical temperature and order to obtain reliable data, the annual course of the thermic humidity exchange with equal diffusion coefficients and requires stratification has to be recorded for deeper water. Owing to the high measured values of the air temperature T and of the vapour pressure heat capacity of water and the generally large mass of water in e in two altitudes (index 1 and 2) over the evaporating area. The lakes, significant amounts of energy are stored in the body of water lower measuring altitude should be as low over the plant crop as and then released during the course of the year’s seasonal cycle. As possible. For areas of water, the water surface temperature and/or a result of this, evaporation losses from the water surface can differ the saturation vapour pressure is used with index 1. considerably from the evaporation from the surrounding land areas The well-known Sverdrup formula for the evaporation depending on the depth of the lake and on the meteorological heat flux (Equation 3.4) is derived from Equation 3.1 together with conditions at the site. Equation 3.3: The two methods described above are available for the LE = (Rn + G)/(1 + γ x (∆T/∆e)) (3.4) separation of the evaporation heat flux from the sensible heat flux: (a) Direct measurement of the sensible heat flux with the help of ∆T and ∆e here are the air temperature and vapour pressure the turbulence correlation method (eddy-flux); differences of both altitudes derived from the above-described (b) Separation of both heat fluxes with the help of the Bowen gradient measurements. ratio β (Sverdrup method). The procedure according to Thornthwaite and Holzman is Other heat fluxes involved in the energy balance (e.g. heat based on exchange equations for the vertical turbulent transport of storage in plant crop, heat exchange on the bottom of bodies of 20 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES water) do not usually have to be taken into consideration. However, f(v) = a + b vc (3.8) lateral energy advections which on land are recorded only with a great deal of technical measurements can reach a significant order or f(v) = m v (3.9) of magnitude (Werner, 1987). The energy advection in a body of water (e.g. dammed reservoirs with deep water outlets) is, in The parameters a, b, c or m (mass transport coefficient) are derived comparison, easily and accurately determined by measuring the empirically; v usually refers to a height of 2 m that is effective in amounts of inflowing and outflowing water and the corresponding the evaporation process, so that a reduction in the height of the water temperatures. measurement of the data available from weather stations has to be made. Use of remote-sensing data The parameters of the wind function found in the litera- The radiation balance Rn and the temperature T0 of the Earth’s ture understandably differ considerably one from the other since surface are determined by satellite measurements. With these data, they are strongly influenced by the climatic and physiographic the evaporation heat flux can be established on the basis of the conditions under which they were derived. By using locally suit- energy balance equation (Equation 3.1), whereby the soil heat flux able parameters for the wind function, Equation 3.7 gives G is, as described above, normally estimated from the radiation sufficiently accurate results for the daily values of the amount of balance and the sensible heat flux H is determined from the evaporation. measured values of the land-surface temperature T0 and the If an adjusted wind function and/or wind data of a suitable temperature T of the air near the ground (site measurements) weather station are not available, evaporation formulae can be used according to H = cp/ra(T0 Ð T). Here, cp is the specific heat of the that do not take account of the influence of the wind and produce air and ra is the so-called bulk surface resistance of the land empirically a relation between the evaporation losses of the body of surface from the Penman-Monteith formula (see Subchapter water and the vapour pressure gradients at the water surface 3.1.5.3). With the use of standard values for ra the conditional (es(Two) Ð e) (Equation 3.10). Furthermore, evaporation formulae equation for the sensible heat flux (Löpmeier, 1991) is reduced to have been empirically developed which include the global radiation H = a(T0 Ð T). The coefficient a is empirically determined on the G as well as the vapour pressure gradients, and which can be basis of satellite measurements (Jackson, Reginato and Idso, 1977; regarded as a simplified version of the combination procedure Seguin and Itier, 1983; Seguin, 1989). described below (Richter, 1977): According to the above formulations, the energy balance equation (Equation 3.1) assumes the following form for the Ew = a' (es(Two) Ð e) + b' (3.10) computation of the evaporation heat flux LE: Ew = a'' (es(Two) Ð e) + b'' G + c'' (3.11) Rn + b Rn + a (T0 Ð T) + LE = 0 (3.6) In the practical application of computing Equations 3.7 3.1.5 Estimation of evaporation on the basis of data from to 3.11 as well as of the combination procedure described below, the meteorological standard measuring network the problem is often the unavailability of measured data for the Practical applications usually require procedures that can estimate water surface temperature Two. Richter (1977) developed a rapidly and reliably the evaporation from land and/or water without procedure with which the water surface temperature of lakes can special measurements and measuring networks. In the following be computed from measured values of air temperature. In the paragraphs, not only are theoretical procedures described, but also annual course of these two values, a phase lag can be observed those with empirical and semi-empirical equations that essentially which is dependent on the water volume of the lake and rely on data from the meteorological standard measuring network parameterized by the average depth of water. and on non-meteorological data available, for instance, from tabular With a corresponding equation, the mean vertical water compilations and maps. Thus, long time series for the water balance temperature in the lake can be calculated and this is required for the component evaporation, that are of interest for many hydrological computation of the change in heat content in the energy balance and water resources problems, can also be established on the basis procedure. of meteorological standard data. Energy balance and combination procedure 3.1.5.1 COMPUTATION PROCEDURES FOR EVAPORATION FROM The computation of the evaporation from the energy balance WATER BODIES equation with the help of the Sverdrup procedure has already been The aerodynamic or Dalton procedure is an empirical-statistical described in Subchapter 3.1.4. The application of this method is procedure that in the form of the Equation 3.7 has found wide indicated whenever there are no suitable empirical equations acceptance, even internationally, for the estimation of evaporation according to Equations 3.7 to 3.11 due to evaporation climate and losses Ew from areas of water (WMO, 1966b; Schrödter, 1985): site conditions. These conditions include, for example, a significant horizon superelevation that restricts the transport of atmospheric Ew = f(v) (es(Two) Ð e) (3.6) humidity or energy advection in the body of water (thermal pollution). The energy balance method is also recommended when a Here es(Two) is the saturation vapour pressure, with Two being the high degree of accuracy and a time resolution of the evaporation temperature of the water surface and e the vapour pressure of the values are required. Measured values of the water surface air, measured at a suitable weather station. temperature are not needed for this procedure because the balance For the wind function, f(v), the following equations are equation is iteratively solved using an initial value of the surface usual: temperature. CHAPTER 3 — DATA DERIVED FROM METEOROLOGICAL MEASUREMENTS 21

The demands made of meteorological input data are Penman’s combination equation (Equation 3.12) greater than with the aerodynamic procedure. However, the energy discussed in the previous section is a procedure for the computation balance method is also useful for practical purposes if simplified of potential evaporation; it is founded on theory and simplified by empirical equations can be used for the computation of the empirical equations, and thus is also suitable for practical individual balance components. The adaptation of such empirical application. The Penman equation reads as follows when related to components to the actual site conditions has to be checked for the wet, low grass cover: case of application. γ γ γ ETP = (s/s + ) (Rn + G)/L + ( /s + ) f(v) (eS(TL) Ð e) (3.14) Penman procedure The combination equation (Equation 3.12) derived from the energy Rn is the radiation balance of the grass cover and G is the mostly balance and the mass transport — as described by Penman, 1948 — negligible soil heat flux. Dividing these by the latent heat of with energy balance term ER and ventilation humidity term EA evaporation L results in the evaporation depth in millimetres. The represents, on the basis of its purely physical approach, the temperature of the evaporating surface and, thus, also its saturation theoretically most mature evaporation formula: vapour pressure can only be determined by complicated procedures (e.g. remote sensing) and are not usually known. For γ γ γ EPENMAN = (s/s + ) ER + ( /s + ) EA (3.12) the computation of radiation values dependent on temperature or on the vapour pressure gradients in terms of ventilation, the where s represents the gradient of the curve of the saturation vapour temperature TL from the meteorological standard network and the γ pressure and thus a temperature function, and is the psychrometer saturation deficit of the near-ground air (eS(TL) Ð e) is used in constant. approximation. With the simplifications and estimations introduced by When using the Penman equation, the adaptation of its Penman, e.g. concerning the aerodynamic roughness of the empirical components to the site conditions should be checked. surfaces observed of water or short-cut grass, and with the Various procedures and equations are given in the literature for the application of the Dalton equation (compare Equation 3.7), there computation of the radiation balance components in ER and for the results from the original theoretical equation a formula which wind function f(v). In order to use the Penman equation makes do with generally available meteorological measurement worldwide, Doorenbros and Pruitt (1977) developed a correction data. factor which takes into account the different climatic conditions. For the computation of evaporation from an open area of Bultot and Dupriez (1985) generalized the Penman equation with water, the Penman equation (Equation 3.12) assumes the following the help of a transfer factor which depends on the type of form: vegetation (i.e. albedo). Jaworski (1985) introduced the roughness length of the crop into the Penman equation and tested it up to a E = (s /s + γ) Rn /L + (γ/s + γ) f(v) (e (T ) Ð e) (3.13) w w S wo time period of one hour. RnW/L is the radiation balance of the water surface, expressed as Apart from the comparatively exacting versions the equivalent of evaporation, and (eS(Two) Ð e) is the vapour (Equations 3.13 and 3.14) of the combination equation, pressure gradient at the water surface. a number of simpler estimation procedures for potential Equation 3.13 requires that heat storage and energy evaporation are used worldwide in water balance models. The advection do not take place in the water body under consideration. computation procedures in use for potential evaporation ETP can This requirement is not satisfied in natural waters since they store be divided into: considerable amounts of energy in the warming-up phase in spring (a) Simple, empirical-statistical procedures which have mostly and summer which are then released again into the atmosphere been derived on the basis of measured values of evaporation during the cooling phase in autumn and winter. (Subchapter 3.1.3) and one or several of the meteorological The Penman formula is therefore not suitable for a direct standard quantities. This type of procedure includes, for calculation of the evaporation losses from natural waters. Its main example, the evaporation equations of Thornthwaite (1948), field of application is the use as potential evaporation, described in Blaney and Criddle (1950), Haude (1952; 1958) and Turc the following section. (1961) as well as those of Turc and Ivanov presented in Wendling, Müller and Schwede (1984); 3.1.5.2 Computation procedures for potential evaporation (b) Physical semi-empirical procedures on the basis of Penman’s Potential evaporation is widely applied in hydrological and water combination equation with the inclusion of empirical formu- resources practice as well as in agricultural consultation practice lations for quantities not available in the meteorological (e.g. sprinkler irrigation control). It is an initial quantity in models standard measurement programme (see, for example, for the computation of the actual evaporation or in models for areal Makkink (1957) and Priestley and Taylor (1972) as well as water balances. In addition, together with precipitation data, it is Turc and Wendling in Wendling, Schellin and Thoma used to characterize the hydroclimate of areas and/or periods of (1991)). time. The climatic water balance — the difference between the Detailed descriptions of these procedures are given in amounts of precipitation and potential evaporation — provides WMO (1966b) as well as in Schrödter (1985) and DVWK (1996) information on the climatically-dependent excess or deficits in the taking into account further developments of some of the above water regime situation and in their regional distribution. The actual procedures. evaporation or the balance from the amounts of precipitation and The water balance models and models for actual actual evaporation is determined by these hydroclimatic conditions evaporation used in Germany mainly apply a Turc or Haude and, furthermore, by the land use and soil type. formulation for the model input potential evaporation. For both 22 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES procedures, so-called crop coefficients, that record the influence of Hoyningen-Heune, Braden and Löpmeier, 1986). The following the plant crop and its phenological development within a year on equation then becomes: the maximal evaporation, have been developed. In order to take into account the soil evaporation of areas only partly covered with ETA = R(W, ETP) ETP (3.16) agricultural crops, Löpmeier (1987) divides the factors in the Haude equation into a soil and a plant factor. with R(W, ETP) being the reduction function of the potential The empirical derivation of the factors in the evaporation. The maximal evaporation that results from the computation equations for potential evaporation and the inclusion inclusion of the plant coefficients for the computation procedure of of different meteorological measurement values leads to a the potential evaporation should be substituted as ETP in Equation considerable disparity in the computation results of these 3.16 (see Subchapters 3.1.2 and 3.1.5.2). equations when applied to one and the same site. Comparative The dependence of the reduction function on the soil computations for climatic conditions in Germany are presented in moisture is described by various authors with step functions Schrödter (1985) and DVWK (1996). Choisnel de Villele and (Renger, Strebel and Giesel, 1974), hyperbolic (Wendling, Müller Lacroze (1990) carry out comparative calculations with numerous and Schwede, 1984; Klämt, 1988) or exponential functions (Disse, equations and equation variants for 20 stations in Europe and 1995). These functions are, as a rule, function systems whose arrive at an evaporation-climatic valuation of these equations. coefficients depend on the parameters of the water binding in the Because of the great number of procedures used and the soil. With these function types, the characteristic of the plants to differences in their computation results, data on potential continue evaporating potentially even when the soil moisture drops evaporation must be accompanied by a description of the to below the value at field capacity is recorded. Only when the soil computation procedure used. For practical applications in water water content drops below a certain limit which, depending on the balance models, care has to be taken that the input of the potential kind of soil, can sink to about 70 per cent of the available field evaporation is calculated according to the same procedure as for the capacity, there is a greater reduction in the actual evaporation than derivation of the water balance model. in the potential evaporation. In order to record the complex influences stemming from 3.1.5.3 COMPUTATION PROCEDURES FOR ACTUAL EVAPORATION soil and crop characteristics on the actual evaporation, water balance FROM AREAS OF LAND models are required that balance in time steps the water content in The amount of actual evaporation depends, on the one hand, on the the root zone from precipitation input and evaporation loss and take hydroclimatic site conditions that are described by the depths of into consideration the partial components of the evaporation precipitation and potential evaporation and, on the other hand, to a process — transpiration of the plants, evaporation of the intercepted great extent by the specific characteristics of the soil and the plant precipitation, evaporation from bare soil, etc. With increasing time crop. In order to be able to take account of these influences resolution of the target parameter actual evaporation, the required emanating from soil and vegetation, the actual evaporation has to number of necessary parameters and partial models to be included be calculated on the basis of areal units, so-called hydrotopes, which increase, because the nonlinearities of, for example, the can be considered more or less homogeneous as to soil and phenological phase of plants, the plant physiological processes and vegetation. the transport process of water in the soil should not be described by A much applied formulation which is used both in simpler simplified empirical formulae. methods for estimating actual evaporation as well as in complex evaporation models can be expressed in a general form by the Interception evaporation following equation: When the mean values of the actual evaporation are computed, the part of the interception evaporation is usually implicitly included in ETA = k ETP (3.15) the model. For time steps of less than one month, the use of a partial model to take into account the evaporation of the intercepted For example, the annual development of quotients from precipitation (or irrigation water) is recommended. actual evaporation and of potential evaporation for the most The water clinging to the plant surface following a important agricultural land uses — according to Haude — as a precipitation event forms the interception storage which evaporates function of soil water content were empirically determined and directly into the atmosphere without having benefited the soil water given in tabular form by Sponagel (1980). These tables allow a reserves. As long as the plant surface is dampened, no stomatal rapid, rough estimation of monthly values of the actual evaporation transpiration takes place. However, under the same weather of the field crops in question. Owing to the empirical derivation, a conditions, transpiration and interception evaporation have different generalization is not necessarily obvious. intensities (Szeicz, 1970). The evaporation from agricultural crops Instead of the quotient k, complex evaporation models use (maize, grain) in the maturity phase consists mainly of the functional relations R between actual and potential evaporation component interception evaporation while the transpiration tends to which are usually referred to as reduction functions (of the potential zero. evaporation). The relation between actual and potential evaporation Complex interception models record the processes of in an observed time interval depends essentially on the soil water filling and emptying the interception storage with reference to the content W in the root zone, which is actually available for the capacity of this storage and the actual weather conditions, i.e. the evaporation process and which is balanced for the time steps of the frequency and abundance of the precipitation events and the evaporation model. Especially with a high potential evaporation, the potential evaporation (Rutter, Mortan and Robins, 1975; Braden, reduction of the potential evaporation also shows a dependence on 1985). The capacity of the interception storage depends on the the value of the potential evaporation itself (Morton, 1983; structural arrangement of the plant crop — height of growth, crop CHAPTER 3 — DATA DERIVED FROM METEOROLOGICAL MEASUREMENTS 23 density, roughness of the plant surface, geometry and arrangement these, it is possible to determine the effectivity parameter n from the of the foliage, etc. Especially with forests of multi-storey structure, type of soil and land use and, subsequently, to determine the actual the interception processes can be estimated only with the aid of evaporation from the amount of precipitation, amount of potential complicated models. If the interception evaporation is considered evaporation and effectivity parameter. only as a partial process of the total crop evaporation in a soil water On the basis of extensive evaluations of lysimeter balance model, then simplified regression models are usually measurements, the authors derived effectivity parameters for a great applied. Such formulae compute the interception evaporation with number of possible land uses in a test area or for non- reference to the storage capacity (leaf area index) and the amounts meteorological site factors. For example, with the parameter n, the of precipitation and potential evaporation, which are already used sealing of the land surface in populated areas, the age of forest stock as input in the soil water balance model (Hoyningen-Heune, 1983b; or the yields of agriculturally-used areas can be taken into account Thompson, Barrie and Ayles, 1981). when considering their impact on the amount of evaporation. Furthermore, there are instructions for the application of the Bagrov Evaporation from sealed urban areas model with regard to the irrigation of arable land and for sites The evaporation from sealed areas is analogous with the process of affected by groundwater. The Bagrov model forms the methodical the evaporation of intercepted water on plants. The interception basis (Glugla et al., 1999) for the preparation of the maps “actual evaporation model used for plant covers can be utilized for the evaporation” and “runoff” in the Hydrological Atlas of Germany. computation of evaporation losses from sealed areas if their storage capacity for precipitation water is inserted. Penman-Monteith formula While numerous usable measurement results for the The further development of the Penman combination equation interception capacity of agricultural and forestry crops are found in (Equation 3.12) by Monteith (Thom, 1975; Monteith, 1976) the literature, little experience has been gained in the storage includes the specific influences of the plant crop on the humidity capacity of sealing material used in towns and industrial areas. The exchange with the help of the so-called crop resistances ra and rs. interception capacity of many sealing materials is probably very The prerequisite of a constantly wet surface as for the Penman small (e.g. asphalt); less than 0.5 mm can be taken as an equation is no longer necessary, so the actual evaporation ETA of a approximate value. For porous material (e.g. bricks), the capacity is plant crop is computed as follows: considerably larger. Apart from the characteristics of the material, ρ s(Rn + G)/L + cps (e Ð e)/ra the state of the sealing (existence of gaps and hollows, angle of ETA = (3.18) γ inclination in the case of roofs) is important for the storage capacity s + (1 + rs/ra) or evaporation losses. Wessolek et al. (1990) gives some Apart from the quantities of the Penman equation, empirically-determined data of storage capacity. Equation 3.18 contains the crop resistances ra and rs and the density ρ Taken together, it can be assumed that the amount of and specific heat cp of the air. evaporation from sealed areas is very small compared to that from The bulk surface resistance ra states more precisely the areas covered with vegetation. This difference in the depths of empirical wind function of the Penman equation. ra depends on evaporation is, however, of particular interest for certain tasks, e.g. the height, density and structure of the plant crop and is the evaluation of changes in runoff as a result of increasing sealing, determined from measurements of the vertical wind profile. The or the calculation of storage basins. bulk stomata resistance rs includes the total water transfer from the pores of the soil through the intercellular spaces and stomata of the Bagrov procedure plants into the atmosphere and depends for one on the soil water With the help of the Bagrov procedure, the mean annual values of content and for the other on the physiological control processes the actual evaporation from land areas, which are required for many that are specific to each kind of plant. rs becomes 0 when only the fields of application in hydrological and water resources practice, film of water on the leaves, the interception water, evaporates. In can be calculated taking the type of soil and land use into this case, a stomatal transpiration does not take place and the consideration: Penman-Monteith formula changes into the Penman equation for the evaporation from an open body of water (Equation 3.13). With d ETA n sufficient soil water supply in the root zone (prerequisite of the = 1 Ð (ETA/ETP) (3.17) d P potential evaporation), the bulk stomata resistance assumes a value specific for plants of rsmin > 0. The data required to determine the mean annual amount The empirical determination of rs as a function of soil and — of evaporation ETA from a hydrotope characterized by the type of plant conditions requires complicated tests with weighing Ð soil and crop are the mean annual depths of precipitation P lysimeters. Monteith (1965) refers to the clear diurnal pattern of rs (including correction of the systematical measurement error) and of during the vegetation phase. For this reason, the actual evaporation — the potential evaporation ETP which describe the hydroclimatic should be calculated in time steps of less than one day. A distinct conditions of the site. Furthermore, a Bagrov or effectivity annual pattern of mean rs values for England are revealed by Wales- parameter n valid for the unit of area is included in the computation. Smith and Arnott (1985). Available data for rs in the literature vary This records summarily the influences of the soil and plant crop on immensely and indicate the necessity of further research on this the evaporation. quantity. The differential equation (Equation 3.17), which can only Equations for the parameterization of ra and rs are found, be solved numerically, was further developed for practical for example, in Thompson, Barrie and Ayles, 1981; Brutsaert, application by Glugla (see Dyck and Peschke, 1983). Apart from 1982; Hoyningen-Heune, 1983a; Bultot and Dupriez, 1985; the numerical method of solution, nomograms were set up. With Wicke, 1988. 24 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

Grass reference evapotranspiration (WMO, 1981c). Such evaluation permits the computation of the A special application of the Penman-Monteith equation (Equation statistical mean of the heavy precipitation amount to be expected as 3.18) is the so-called grass reference evapotranspiration developed a function of the rainfall duration (minutes to days) and the by Allen et al., 1994. This is defined as evaporation from a recurrence interval (annuality). No certain conclusion can be made permanently homogeneous grass crop, 0.12 m high, with no water for recurrence intervals (in years) more than three times the number stress (soil water content greater or equal to 70 per cent of the of measured years. For this, a 30-year series of precipitation available field capacity). With this prerequisite, the grass reference amounts must be available to determine a recurrence interval of evapotranspiration is equal in importance to potential evaporation. almost a hundred years. The heavy precipitation point statistics Following extensive research, the authors determined the values for gleaned from a station are valid in the case of short-duration the crop resistances in the Penman-Monteith equation with the said precipitations, mostly only for areas of 1 km2, and in the case of conditions as follows: long-duration precipitations for areas of approximately 20 km2 around the station. -1 -1 -1 rsmin = 70 s m ; ra = 208 s m at a wind velocity of 1 m s . These point results can be transferred to other locations by various regionalization methods; however, knowledge about The maximal amounts of evaporation (i.e. occurring with existing relationships between extreme value distributions and sufficient soil water content) of other kinds of plants normally available climate analyses as well as correlations with orography deviate from the amount of the grass reference evapotranspiration. should be taken into account (Subchapter 3.3.3). This applies, for The application of the grass reference evapotranspiration as a example, to the relationship between heavy daily precipitations and hydrometeorological reference quantity for the estimation of the the isohyetal analysis of average annual precipitation amounts or to maximal plant evaporation requires a crop coefficient which, for the derivation of showery precipitations with the aid of the exposure agricultural crops, has a distinct annual pattern corresponding to the of a location (altitude gradient of orography as well as windward phenological crop development. and lee-side influences). The practical importance of the grass reference In hydrological practice, however, a relationship must be evapotranspiration lies in the fact that the potential evaporation, established between the precipitation amounts from the point which is defined only qualitatively and is therefore indistinct, is determinations of heavy rainfall statistics and the runoff from the replaced by a quantity which is exactly defined by means of set crop catchment basin. Consequently, in this case too, the areal resistances. The problem described above of differing results precipitation amount must be known. Thus, the question arises as produced by using different computation procedures for potential to the conditions under which areal precipitation amounts can be evaporation is thus eliminated. As opposed to the advantage of a obtained from point statistics. The main problem in estimating areal defined potential evaporation, the data requirements of the Penman- precipitation amounts through the extrapolation of point results lies Monteith formula for practical applications (evaporation mappings, in the spatially limited extent of the homogeneous or quasi- long time series of evaporation, etc.) are excessive. Wendling (1995) homogeneous precipitation fields for showery or continuous rainfall tried to calibrate simpler empirical procedures for potential as described above. evaporation (Subchapter 3.1.5.2) with the help of the grass reference In comparison to the point heavy rainfall statistics, the evapotranspiration. Such a calibrated formula was used to prepare spatially averaged intensity of the precipitation events decreases the evaporation maps in the Hydrological Atlas of Germany (BMU, with increasing areal extension. In order to take that areal effect 2000). into account, many depth-area-duration relationships of the precipitation were developed in the form of areal reduction curves 3.2 HEAVY RAINFALL POINT STATISTICS (Table 3.1 and Figure 3.1). In this context, the factor by which the Many hydrological problems can be solved only with the aid of amount of the statistically-determined point precipitation of a climatological fundamentals. This means the concerted statistical given duration and recurrence interval is to be multiplied to obtain evaluation of series of meteorological data as long as possible. The the areal precipitation depth for a given area for the same duration statistical evaluation of extreme values of heavy rainfall events, for and recurrence interval is designated as a reduction factor. As instance, is required for the entire sector of planning and design for some investigations show, for certain regions, characteristic the construction or redevelopment of hydraulic structures, such as reduction factors exist which vary mainly only in relation to the the dimensioning of urban sewage networks or sewage treatment size of the area and the duration of the precipitation. A synopsis of plants but also for the construction of retention basins as flood the results of the individual investigations compiled in the two alleviation structures in small and large catchment areas. Heavy figures for regions in which the climates are not at all uniform rainfall is a form of precipitation events which, compared to their shows altogether a relatively good agreement (Deutscher duration, show a high intensity and therefore occur seldom, e.g. on Wetterdienst, 1986). average twice a year. They may occur as precipitation events of However, in dealing with practical hydrological short duration (showers) with usually high, sometimes rapidly problems, it must be taken into account that the size of the area changing, intensity and a narrowly limited precipitation field and as and the precipitation duration are not independent of each other, long-lasting precipitation events (continuous rain) with low, usually but short precipitation durations are related to small areas and hardly changing, intensity and an extended precipitation field. The precipitations of long duration to large catchment areas. mechanisms causing the development of precipitation fields vary Accordingly, the reduction factor mostly has a value of around 0.9 accordingly (Chapter 5). to 1.0. Considering the systematic errors in precipitation A statistical evaluation of the extreme values of such measurements as well as the uncertainty of statistically determined rainfall events is made following the selection of annual or partial point precipitation amounts, a reduction can usually be dispensed series and is adjusted through mathematical distribution functions with. CHAPTER 3 — DATA DERIVED FROM METEOROLOGICAL MEASUREMENTS 25

Table 3.1 Reduction factors for duration and area (National Environment Research Council, 1975)

Duration Area (km2) 1 5 10 30 100 300 1 000 3 000 10 000 30 000

1 minute 0.76 0.61 0.52 0.40 0.27 —— — —— 2 minutes 0.84 0.72 0.65 0.53 0.39 —— — —— 5 minutes 0.90 0.82 0.76 0.65 0.51 0.38 ———— 10 minutes 0.93 0.87 0.83 0.73 0.59 0.47 0.32 ——— 15 minutes 0.94 0.89 0.85 0.77 0.64 0.53 0.39 0.29 —— 30 minutes 0.95 0.91 0.89 0.82 0.72 0.62 0.51 0.41 0.31 — 60 minutes 0.96 0.93 0.91 0.86 0.79 0.71 0.62 0.53 0.44 0.35 2 hours 0.97 0.95 0.93 0.90 0.84 0.79 0.73 0.65 0.55 0.47 3 hours 0.97 0.96 0.94 0.91 0.87 0.83 0.78 0.71 0.62 0.54 6 hours 0.98 0.97 0.96 0.93 0.90 0.87 0.83 0.79 0.73 0.67 24 hours 0.99 0.98 0.97 0.96 0.94 0.92 0.89 0.86 0.83 0.80 48 hours — 0.99 0.98 0.97 0.96 0.94 0.91 0.88 0.86 0.82 96 hours ——0.99 0.98 0.97 0.96 0.93 0.91 0.88 0.85 192 hours —— —0.99 0.98 0.97 0.95 0.92 0.90 0.87 25 days —— ——0.99 0.98 0.97 0.95 0.93 0.91

3.3 REGIONALIZATION IN 1921). Although its database was very poor, it nevertheless HYDROMETEOROLOGY contained a high proportion of empirical values and physical 3.3.1 Introduction considerations concerning atmospheric processes. The aim of In conventional measuring networks, climatic values are obtained objective regionalization methods with the aids available today only at points of irregularly distributed individual stations. For must therefore be to convert subjective methods of transferring each determination of areal values, these point results are to be point results to the area based on empirical methods as far as transferred to the area by means of areal analysis. Since the possible to objective computational methods with the development beginning of meteorological measurement, attempts have been of an expert system, taking into account physical processes and made to represent these climatic elements in the form of maps. The interdependencies. first Climate Atlas of Germany appeared in 1921 (Hellmann et al.,

1.00 T=1 h 12h G 12h h (-) 12h N 30 12h 60min 15 0.75 60min 30min 60min 30min 15min 15min 30min 60min

15min 0.50 30min

15min Hannover, 6 stations, 2.5 years, 35-100 km2 0.25 Hamburg, 18 stations, 8 years, 80-300 km2 Hmburg, 5 stations, 12 years, 80-115 km2 (Abraham, et al., 1976) Land, Sweden, 12 stations, 3 years, 2.5-25 km2 (Niemczynowicz, 1982) Emscher-Lippe area, 58 stations, 28 years, 31-156 km2

0 50 100 150 200 250 km 2 300 Figure 3.1 — Comparison of different reduction curves for duration and area (Verworn and Flender, 1986). 26 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

Figure 3.2a — Forms of representation of regionalized precipitation Figure 3.2b — Forms of representation of regionalized precipitation data: areal precipitations for hydrologically-limited areas. data: Isohyetal analysis for mean annual precipitation amounts.

[mm] < 60 60 to < 70 70 to < 80 80 to < 100 100 to <120 120 to < 140 140 to < 160 160 to < 190 190 to < 210 210 and more

Figure 3.2c — Forms of representation of regionalized precipitation data: climatically similar regions with allocation of representative Figure 3.2d — Forms of representation of regionalized precipitation point statistics. data: grid fields. CHAPTER 3 — DATA DERIVED FROM METEOROLOGICAL MEASUREMENTS 27

3.3.2 Cartographic representation of regionalized temporal and spatial representativeness of the point results as well as precipitation data the computational methods used for the description of the point Depending on the regionalization method and purpose of results: application, different forms of spatial representation of the (a) Requirements for the testing of temporal representativeness are: precipitation are usual (Figures 3.2a to 3.2d): (i) Investigations for statistically-sufficient long periods (a) The meteorological value which is directly related to the of evaluation; hydrological value of surface runoff is areal precipitation (ii) Standardized instrumentation during the period of (Chapter 3), which is oriented on the hydrologically-limited evaluation; areas of the topographic divides and gives a mean value for (iii) Standardized data processing; this reference area (Figure 3.2a); (iv) Guarantee of the stationary position and homogeneity (b) The most frequently used form of representation in meteo- of the time series (changes to the stations, relocation rology, e.g. for the mean annual precipitation amount is the of the stations, etc.); isohyetal map (Figure 3.2b). From the isohyetal lines, a (b) Requirements for the testing of spatial representativeness conclusion is drawn about the behaviour or the distribution are: of precipitation amount between the measuring stations (i) Investigations of the density and distribution of exist- (Schirmer and Vent-Schmidt, 1979). The climatologist ing stations; acquires knowledge about this not only through a thorough (ii) Standardized instrumentation throughout the entire analysis of the measurement data but also extends his measuring network; knowledge by continual observations of the precipitation (iii) Standardized measuring periods for all stations; events and possibly also by carrying out special measure- (iv) Standardized statistical evaluation methods for all ments in dense measuring networks. This knowledge is stations; transferred in the drawing of isolines; the variation range of (c) In order to be able to describe and weigh the frequency of the precipitation element is also taken into consideration so heavy rainfall events during a certain measurement period that each map contains a high proportion of subjectivity in (outlier problem), as well as estimates for very rare events, it the lines. The position of the isolines is therefore to be taken is necessary to apply extreme value evaluation methods. As a as an aid to orientation, delimiting from one another the result, other tests are needed: areas in which the amount of precipitation fluctuates more or (i) Usability of the computational statements applied to less around a mean value. describe a partial data set; In Figure 3.2c, climatically similar regions are (ii) Deviations between the different computational subjectively put together on the basis of the empirical data available, procedures; representative heavy rainfall statistics having been assigned to each (iii) Deviations between the measured and adjusted region (LfU BW/DWD, 1975). Transitions between the regions are distributions. indistinct. Thus, the determination of an area of confidence (tolerance interval) For a complete area coverage of the heavy rainfall for the validity of extreme value computations is possible. evaluation, described in detail in Subchapter 3.3.3, coloured grid fields on a scale of 1:2.5 million are selected, the side lengths of the 3.3.3 Cartographic representation of regionalized heavy grid being 8.45 km and the area 71.5 km2 (Figure 3.2d). rainfall amounts Precipitation amounts of a certain range are assigned to each of The aim of coordinated heavy rainfall regionalizations in Germany these grid fields. Because of the range, it is possible to take into was not only to process heavy rainfall point statistics, but also to account uncertainties due to different disturbing influences (Bartels, represent the spatial distribution of heavy rainfall amounts for Albrecht and Guttenberger, 1990; Bartels et al., 1997). Another selected durations (between five minutes and 72 hours) and relevant example of the grid field representation of nationwide overviews of recurrence intervals (between once a year and once every 100 years) values derived from hydrological measurements are the maps in the on maps with adequate spatial resolution for all durations and Hydrological Atlas of Germany. The grid resolution of the Atlas recurrence intervals (Bartels, Albrecht and Guttenberger, 1990; tables is 1 km × 1 km (BMU, 2000). Bartels et al., 1997). The purely computational methods used to transfer the point Therefore, an evaluation of randomly-occurring events values to the areas are based mainly on the fact that a distance-depen- within a 30-year measuring period and an extrapolation on very rare dent weighting of the irregularly-distributed surrounding stations is precipitation events was planned. The extreme values statistics carried out (see Chapter 5). The spatial distributions thus obtained, statement starts for every duration D (precipitation duration however, offer no guarantee that, with the given density of the measur- including interruptions) of a partial or annual series which is ing network, an area analysis true to nature can be obtained. Paul determined from a series of registered or measured precipitation (1988) tried various computational interpolation methods and amounts. Each series of heavy precipitation amounts hN(D;T) is compared them with manually-analysed isohyetal distributions. These adjusted to the theoretical distribution function by a regression investigations show that none of the computational methods applied calculation, where T is the recurrence interval (annuality) (DVWK, is an improvement on the cartographical representation enriched with 1985). The distribution function is represented on squared paper the experience values of an analyst well-versed in climatology. with hN(D;T) on the y-axis and ln T on the x-axis as a straight line. Complementary to the computational methods, methodical regional- The parameter u(D) is the ordinate section for ln T = 0 and the ization investigations must first be made to check the effect of possible parameter w(D) gives the gradient of the straight line. In order to disturbing influences and thus the meaningfulness of spatial distribu- obtain clear precipitation amounts over all durations, a double tions (Bartels, 1992). Of special importance are observations on the logarithmic compensation of the parameters u(D) and w(D) in 28 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES duration range I (5 to 60 minutes) and duration range II (60 minutes Tests of spatial representativeness to 12 hours) is carried out. The spatial deviations can be investigated by reconstructing the Building on these point evaluations, a regionalization of extreme value statistics obtained station by station exclusively from heavy rainfall events covering the area was then planned with the the surrounding station values. The difference between the field aid of a complex regionalization method which especially contains distribution obtained by excluding individual station results and the the orographically-modified variogram analysis. The results are field distribution with availability of station results can be recorded recorded in grid representations with a resolution of 8.45 × 8.45 as a percentage of the over- or underestimation. km per grid field. In this way, the KOSTRA values of the heavy precipitation amounts are obtained, with the help of which heavy Estimation errors by using extreme value computational methods rainfall point statistics can finally be derived, even for locations As a basis for the computation of the probability-dependent without precipitation measurement series (Table 3.2). confidence intervals of the extreme value computations, a risk The evaluation of disturbing influences should follow threshold of e.g. 10 per cent can be introduced i.e. 10 per cent of all from the KOSTRA investigation covering the area of Germany. The disturbing influences are no longer covered by the confidence following remarks concentrate on the tests for temporal and spatial interval (WMO, 1981c). representativeness of heavy rainfall statistics using Gumbel and The decisive criteria for a regionalization procedure result exponential distribution (DVWK, 1985). from the comparison of temporal and spatial variations for the Standardized instrumentation and data processing available stations over a long time period. Both influences can cause should be guaranteed in the measuring networks of the deviations with negative or positive signs. The addition of both Meteorological Services. The application of statistical test effects, therefore, leads to a reinforcement of an over- or procedures regarding the stationary position and homogeneity of underestimation or to a compensation of both disturbing influences. the mean values and deviations of the time periods investigated If the methodical estimation error using the statement of yield as a rule satisfactory results in comparison with other computational method is now added and if the greatest error disturbing influences. influence is investigated for each grid field, then an interesting picture is presented for Germany. The estimation error as a result of Tests of temporal representativenes using a distribution function is decisive for 90 per cent of the total The natural breadth of variation of rain is known to be very high, area of Germany. Later, the spatial (6 per cent) and then the so that an evaluation period as long as possible should be laid down. temporal variability (3 per cent) follow. The remaining error The order of magnitude of the uncertainties caused by temporal influences are negligible. It should be taken into account that the variation can, thus, be established so that moving time intervals of choice of the distribution function is to be classified as rather 30 years following Figure 3.3 are shown (1931Ð1960, 1932Ð1961, conservative, i.e. as a rule, the estimation of precipitation amounts etc.) for stations with long time series. gives too high values.

Table 3.2 Location-related KOSTRA heavy rainfall amounts hN (mm) and precipitation yields RN (l/(sáha)) depending on the duration D and the recurrence interval T for an example grid field (RN = (166,6/D) hN with D in minutes)

T 0.5 1 2 5 10 20 50 100

DhNRNhNRNhNRNhNRNhNRNhNRNhNRNhNRN

5 min 5.5 185 7.0 233.8 8.5 283.0 10.4 347.9 11.9 397.1 13.4 446.2 15.3 511.2 16.8 560.3 10 min 6.6 110 8.6 143.4 10.6 177.3 13.3 222.1 15.4 256.0 17.4 289.9 20.1 334.7 22.1 368.5 15 min 7.2 80.5 9.7 107.8 12.2 135.0 15.4 171.1 17.9 198.3 20.3 225.6 23.5 261.6 26.0 288.9 20 min 7.8 64.6 10.6 88.0 13.4 111.4 17.1 142.2 19.9 165.6 22.7 189.0 26.4 219.8 29.2 243.2 30 min 8.5 47.3 11.9 66.1 15.3 84.9 19.8 109.8 23.1 128.6 26.5 147.4 31.0 172.2 34.4 191.0 45 min 9.3 34.6 13.4 49.7 17.5 64.8 22.9 84.8 27.0 99.9 31.1 115.0 36.5 135.0 40.5 150.2 60 min 9.9 27.6 14.6 40.6 19.3 53.5 25.4 70.7 30.1 83.6 34.8 96.6 40.9 113.7 45.6 126.7

90 min 11.1 20.5 16.0 29.7 21.0 38.8 27.5 51.0 32.5 60.1 37.4 69.3 44.0 81.4 48.9 90.6 2 h 11.9 16.6 17.1 23.8 22.3 30.9 29.1 40.4 34.3 47.6 39.4 54.8 46.3 64.3 51.4 71.5 3 h 13.3 12.3 18.8 17.4 24.3 22.5 31.5 29.2 37.0 34.3 42.5 39.3 49.7 46.1 55.2 51.1 4 h 14.3 10.0 20.1 13.9 25.8 17.9 33.3 23.2 39.1 27.1 44.8 31.1 52.4 36.4 58.1 40.3 6 h 15.9 7.4 22.0 10.2 28.1 13.0 36.1 16.7 42.2 19.5 48.3 22.3 56.3 26.1 62.4 28.9 9 h 17.7 5.5 24.2 7.5 30.6 9.4 39.1 12.1 45.6 14.1 52.0 16.1 60.6 18.7 67.0 20.7 12 h 19.1 4.4 25.8 6.0 32.5 7.5 41.4 9.6 48.2 11.1 54.9 12.7 63.8 14.8 70.5 16.3

18 h 21.4 3.3 28.7 4.4 36.0 5.6 45.6 7.0 52.9 8.2 60.1 9.3 69.7 10.8 77.0 11.9 24 h 23.8 2.8 31.6 3.7 39.4 4.6 49.7 5.8 57.5 6.7 65.4 7.6 75.7 8.8 83.5 9.7 48 h 29.9 1.7 38.8 2.2 47.7 2.8 59.5 3.4 68.3 4.0 77.2 4.5 89.0 5.2 97.9 5.7 72 h 34.2 1.3 43.7 1.7 53.2 2.1 65.8 2.5 75.3 2.9 84.9 3.3 97.5 3.8 107.0 4.1 CHAPTER 3 — DATA DERIVED FROM METEOROLOGICAL MEASUREMENTS 29

The procedure described is an iterative process which can may not exceed a certain threshold. In addition, the determination lead to the following consequences: can be used to show that because of the predominantly advectively- (a) Limitation to a shorter evaluation period in favour of a influenced precipitation events, the spatial similarities between higher density of stations and standardization of the time winter short-term precipitations and long-term precipitations are period; much more strongly pronounced and cover greater distances than in (b) Limitation of the grid fields to a specified area size; the other seasons. Thus, various correlation partners from the much (c) Limitation of the class range of the precipitation amounts to denser measuring stations which measure daily precipitation are a specified number of classes; available for the distribution of total areas in representative regions (d) Supplementary investigations for the derivation of other for stations with heat table recording precipitation gauges (so-called auxiliary functions. DIGI stations), which can be used to optimize limits in an iterative Such auxiliary functions may result from extreme value process. statistics of predominantly advectively-determined long-term The basic criteria including an orographic database are: precipitations (D ≥ 24 h) through a relationship with the mean (a) Agreement between the point evaluations and the existing annual precipitation amount. With the aid of this relationship, KOSTRA grid field occupancy (with maximum 20 per cent existing maps of the mean annual precipitation amount, which were deviation between DIGI station values and grid values); analysed by a climatological expert on the basis of a great number (b) Mean monthly values of the precipitation amount by region- of empirical values concerning the effects of weather situations, can ally-typified monthly values of the precipitation; be referred back to. (c) Mean annual precipitation amount as spatial analysis; The extreme value statistics of the mostly convectively (d) Distance between each DIGI station and the assigned region. formed short-term precipitations (D ≤ 60 min) do not show such a relationship with mean annual precipitation amounts. However, in Conclusions this case, a good relationship with orography can be found in For all regionalization methods, the various error influences Germany. A measure for this is, for example, the exposure height (uncertainties) must first be weighed against each other in order to in the low mountain region, which can be derived from the arrive at a given maximum possible spatial resolution with a certain gradient and orientation of the slope to the prevailing weather meaningfulness (confidence range) on the basis of the density and situations. distribution of the stations as well as the evaluation period and the A special regionalization procedure was developed to computational method. Purely computational methods to transfer derive the spatial distribution of winter heavy rainfalls. This was the point results to the areas (Chapter 3) can normally only deliver a done by establishing the spatial representativeness of the individual very rough spatial distribution because, in this case, significant station values on the surrounding grid fields, whereby the deviations influences through the weather situation with differently operating

73034 Thurnau 73046 Nordhalben mm mm 80 80 T = 100a T = 100a

T = 50a T = 50a

60 T = 20a 60 T = 20a T = 10a T = 10a

T = 5a T = 5a 40 40 33-62 42-71 51-80 (a) 33-62 42-71 51-80 (a) Interval Interval

73052 Presseck 73053 Steinbach a.Wald mm mm 80 80 T = 100a T = 100a

T = 50a T = 50a

60 T = 20a 60 T = 20a T = 10a T = 10a

T = 5a T = 5a 40 40 33-62 42-71 51-80 (a) 33-62 42-71 51-80 (a) Interval Interval Figure 3.3 — Heavy rainfall amounts for the duration D ≥ 24 h and the recurrence intervals T = 5a to T = 100a for 30-year moving evaluations of the time period 1931Ð1980 from selected stations with daily values. 30 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES precipitation mechanisms and orography with windward and lee- aid of cartological representations of regionalized extreme values of side effects remain ignored. Without such additional information, the precipitation supply (Günther et al., 2000). the station results, both for short-term and long-term precipitations, are valid only in a relatively small area around the station. If, on the 3.4 PROBABLE MAXIMUM PRECIPITATION (PMP) other hand, the various influences with extensive error analysis and 3.4.1 General the combination of computational methods, existing empirically Heavy rainfall statistics allow, given the normally available time calculated documents and orographic concepts are included in the periods of about 30 years, conclusions of, at the most, recurrence investigation, then distributions with a relatively high spatial periods of up to 100 years. Flood structures, which are meant to be resolution covering the area can be derived. built with greater safety, cannot be based on these statistics. The probable maximum precipitation (PMP) is, therefore, defined 3.3.4 Cartographic representation of regionalized extreme conceptually as the theoretically greatest precipitation amount values of precipitation supply physically possible of a given duration for a specified catchment In the winter months, in addition to the precipitation, the area at a specified season (without regard to climate changes). This precipitation supply — the sum of rain and melt water release — is definition represents the upper limit of a possible heavy of importance. If the breakdown of the snow cover and the precipitation event (WMO, 1986a). precipitation stored in it come together with heavy rain, extreme In effect, there are three possibilities for quantifying PMP: values can appear which, in their amounts and intensities, influence (a) By means of purely statistical methods (Hershfield, 1965); the extreme value statistics in the snow-covered hydrologically (b) With the aid of deterministic physical models (Haiden et al., relevant areas. This is especially true of areas in medium and high 1991); places in the mountains (higher than 400 m). (c) Empirically, by physically-based evaluation of meteoro- Precipitation values are obtained from measurements; logical data (DVWK, 1997). values of precipitation supply are found with the aid of models. The Statistical methods are used for areas up to an area size of snow cover model SNOW-K enables the daily calculation of melt about 1 200 km2; physical methods can cover catchment areas up water release out of the snow cover, taking into account the liquid to 50 000 km2. Surface areas beyond these measurements are very precipitation, by continuous simulation of the development of the difficult to handle, both as models and empirically, because of the snow cover (Rachner, Matthäus and Schneider, 1997). The spatially different weather influences. optimization of the model parameters and the verification of the The resulting very high precipitation amounts, according derived results are obtained, relative to the location, by means of all to the definition of PMP, are therefore not directly used in available measured values of the water equivalent of the snow cover. dimensioning hydrological structures. The safety elements designed The daily values of the precipitation supply arrived at in this way into barrages approach these values in their calculations. In this way, are subject to an extreme value statistical analysis (REWANUS) the use and the potential danger of the installation in the case of with the support of the KOSTRA investigations. Regionalization is failure are weighed up against each other. Also taken into obtained from station-related point results. Mean winter consideration is an estimation of the construction methods as well as precipitation plays an important role in this regionalization the necessary building and running costs for a specified construction procedure. Regional climatological influencing factors in the river in a specified region. An optimization of these various needs, basins of Germany were taken into account with very well however, can only be arrived at through close co-operation among confirmed regression relationships. On the basis of these meteorologists, hydrologists and engineers. Parallel to this, relevant regressions, the regionalization of station-related extreme values, extreme values of precipitation in Germany are processed for calculated for all climate and precipitation stations, was obtained by subsequent use in hydrological practice. In addition — depending distance-dependent interpolation on the middle point of the grid on the duration — the area between the KOSTRA heavy areas (resolution of 8.45 × 8.45 km per grid field). As a result of the precipitation amount of the 100-year recurrence interval and the REWANUS investigations, grid-oriented extreme values of the PMP is to be sounded out. precipitation supply in the hydrological winter half year for The procedures for the evaluation of PMP cannot be durations of d = 0.5d to D = 10d are available with blanket coverage standardized. There is also no objective procedure to permit sure in the form of regionalized cartological representations. statements about the accuracy of the PMP estimates. Nevertheless, The KOSTRA heavy precipitation amounts for the winter observations about accuracy can be made. In relation to this, the period are based on measurements both of liquid as well as solid following factors should especially be taken into account: precipitations, whereby the precipitation fallen as snow is liquefied (a) Excess of estimated PMP over the maximum observed before the precipitation amount (in millimetres) is determined. precipitation amounts for the surrounding climatologically- Stored in the snow cover, the precipitation does not immediately homogeneous region; have to turn into runoff, at least not completely. On the contrary, the (b) Number and severity of the available recorded precipitation values of the precipitation supply provide the hydrologically- events in the region in question for the longest possible time effective contribution that is decisive for the formation of the runoff. series; Therefore, the REWANUS values, the information about the (c) Number, character and interrelationship of the maximizing occurrence probability of extreme values of the precipitation supply, steps; are of great significance, above all in the snow-covered (d) Reliability of the model used in relation to the precipitation hydrologically-relevant regions. description and other meteorological variables; With regard to the extreme values of the precipitation (e) Probability of the individual meteorological variables used in supply for durations from D = 0.5d, tabulated summaries similar to the model being excessive in relation to the occurrence of the heavy precipitation statistics (Table 3.2), can be derived with the very rare events. CHAPTER 3 — DATA DERIVED FROM METEOROLOGICAL MEASUREMENTS 31

3.4.2 Estimation of PMP using a physically-based and 3.6 show the comparison of regional precipitation amounts with evaluation of meteorological data worldwide extreme precipitation amounts. With maximization, an empirical path was taken by the Deutscher Wetterdienst to investigate the PMP amounts in 3.5 SNOWMELT Germany. To be able to estimate PMP, knowledge of the Snowmelt forecasting has top priority in short-term runoff meteorological variables and processes which impose a limit on forecasting (Kirnbauer, 1993). Decisions on definite measures to be precipitation amount is necessary. Obviously, the water vapour taken, such as operational planning and flood warnings, can be content of the air plays an important part. By multiplying the based on routinely established and regularly released forecasts. precipitation amount measured in an extreme heavy precipitation Forecasting models should, above all, produce good results in event by the maximizing factor gives the maximized situations which are important for the user (Gutknecht, 1991). precipitation amount. The maximizing factor is the relation — Experience in the use of snowmelt models for runoff forecasting has depending on the region and the season — of the theoretically been gained for instance in Switzerland (Schädler, 1992). Short- greatest water vapour content of the atmosphere to the water term forecasting takes into consideration changes in snow cover. vapour content which is available over an investigated area Reliable forecasting of snow water runoff under extreme conditions during an observed extreme heavy precipitation event of a certain (rainfall and strong wind) remains difficult. In Norway, by using duration. By adopting an optimal degree of effect of the remote-sensing data simultaneously, information concerning snow precipitation process, this maximizing factor already implicitly cover development is included in a flood early warning system contains the appropriate wind influence. (Haddeland, 1996). Continually-working models and event models Within the project entitled “Regionalization of maximized can be differentiated according to their application. The former have regional precipitation amounts in Germany” (MGN), the large important variables of state, e.g. the water equivalent (Rachner, number of computationally-evaluated probable maximum regional Matthäus and Schneider, 1997), and the latter need initial values in precipitation amounts for selected regions was regionalized, i.e. order to function (Kirnbauer, 1993). transferred to the other regions of Germany, and represented in In order to quantify the estimations of runoffs from snow- maps of the maximized regional precipitation amounts (MGN) for covered catchment basins, information on the areal values of the durations from D = 1 h to D = 72 h (DVWK, 1997). The main focus water runoff from melting snow cover is required. In the station was on relatively short precipitation durations of less than a day and networks run by the Weather Services, the snow depth and, to some relatively small regional areas. In the follow-up project entitled extent, the water equivalent of the snow cover are measured at the NIEFLUD, probable maximum regional precipitation amounts were precipitation stations and the meteorological data at the synoptic- processed for heavy precipitations of several days’ duration and climatological stations. The snowmelt runoffs have to be derived covering a wide area (> 1 000 km2) in Germany’s river basins from these data. Assessment and simulation of snowmelt runoffs (Malitz and Günther, 2001). The precipitation database did not have therefore mainly require the following steps: to be restricted to the approximately 200 stations recording (a) Processing of meteorological input data for snowmelt precipitation during the day and having high time resolution of models (e.g. precipitation, global radiation, air temperature, around 30-year precipitation time series. Data could be obtained air humidity, wind velocity) in spatial-temporal resolution; from the around 4 500 stations measuring precipitation once a day (b) Estimation of snowmelt rates and snow cover runoffs, taking and having a longer observation period (often 60 years). The high into consideration the rain falling into the snow cover and spatial density of these stations permitted the regional precipitation the content of free water; amounts for Germany’s river regions to be calculated directly from (c) Transformation of snow cover runoffs (snow and rain water) the measured point precipitation amounts. Therefore, neither the use into areal runoffs. of precipitation-duration-area relationships necessary for the Any step can be made with models of different quantification of the MGN values nor the execution of complexity. The main problem with regard to the snowmelt and its regionalization were applicable. modelling is that information obtained with point melt models must Individual investigations using maximization calculations subsequently be transferred to the area. The related deficit in for a few stations have also been carried out by other authors in the knowledge is made particularly clear in the study by Braun (1985), past. Although these investigations were also carried out very while in WMO (1986b) it has, to a large extent, been disregarded. meticulously, they always hold the danger that — because of the However, in 1988, the WMO Rapporteur on Data Inputs for small number of stations and the comparably short length of the Hydrological Models, on the recommendation of the WMO available precipitation series as well as the sometimes Rapporteur on Meteorological Systems for Hydrological Purposes, unrepresentative investigation period with regard to time — the was entrusted with the task of examining the methods for values calculated are by a long way not the probable maximum quantitative inputs from snowmelt. precipitation amounts in the investigated regions (Haugk, 1983). As The validity range of snowmelt models is, strictly opposed to this, all individual results in a temporal and spatial speaking, limited to the location at which the required context were evaluated at every step in both Deutscher Wetterdienst meteorological input data are recorded. The transfer of data and maximization projects in the search for the probable maximum results recorded at individual locations to the surrounding area or regional precipitation amounts in Germany. larger regions has recently been the subject of intensive research Figure 3.4 demonstrates the regionalized maximized areal (Blöschl, 1990; Kleeberg, 1992; Pfützner et al., 1992; Kirnbauer, precipitation amount for duration D = 72 h and an area size of about Blöschl and Gutknecht, 1994). For most purposes, the generali- 1 000 km2 in the summer season in Germany. The maximized zation is at present carried out using empirical procedures, whereby regional precipitation amount for a region of interest — indeed in the regional distribution of individual data is derived from an the middle of this region — can be read from this map. Figures 3.5 adequate amount of information from the stations. This course of 32 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

6 12 24 48 72 100 000

50 000 6 12 24* 48 72

12 48 72 6 24 * 12 48 10 000 6 24* 72 5000 6 72 12 24 48

6 X 24 48 72 )

2 12 24 48 72 1000 6 12X 24 Area (km Area 500 6 X 48 72 12 48 h. 24 h. 72 h. STORM A 6 h. 12 h. STORM A Latitude (°N) X STORM A 48 100 STORM A 6 X 12 24 72 STORM A 1 500 mm STORM A 1 300 mm 50 * 1 100 mm

900 mm 6X 12X 24 48 72

700 mm

550 mm 10 450 mm 0 100 200 300 400 500 600 700 800 900 350 mm 275 mm Probable maximum precipitation (mm) Figure 3.6 — Envelope of the Amount-Areas-Duration curves of the PMP (WMO, 1986a).

rain water predominantly produced at the surface of the snow cover, Longitude (°E) of the thermal retention capacity and of the free water content of the Figure 3.4 — Regionalized maximized areal precipitation for duration snow cover, and its hydraulic properties varying considerably with D = 72 h in an area size of about 1 000 km2 in the summer season in Germany. time as compared to soils in conjunction with the mechanic (structural) retention capacity. In the absence of experience, a action is a compromise. Despite its disadvantages due to storage term determined by mathematical optimization is therefore unsatisfactory physical reasons it will continue to be a procedure often used for simulation models. However, no general statements that is taken seriously in the future (Becker, 1992; Braun and can be derived since the coefficients used in the models cannot Rohrer, 1992). normally be clearly interpreted physically. The difficulty of simulating with a model the behaviour According to a state-of-the-art report on snow hydrology by of runoff from snow cover is that the runoff from an extremely Herrmann (1986), only specific experiments should be based on theo- nonhomogeneous and anisotropic hydrological storage, such as the retical considerations concerning water storage, the movement of snow cover, constitutes a mathematically not clearly assessable and seepage water and thus the behaviour of runoffs from snow covers in generally valid function of the input in the form of melt as well as the temporal resolution, e.g. Colbeck (1972), de Quervain (1973) and

2000

1000 2540.0 600 400 CHERRAPUNJI, INDIA 200 475 BELLENDEN KER, QUEENSLAND CHERRAPUNJI, INDIA 254.0 100 R = 166 D HSIN-LIAO, TAIWAN 60 BELOUVE, LA REUNION 40 SMETHPORT, PA. D'HANIS, TEXAS Rainfall (in) 20 Rainfall (cm) ROCKPORT, W. VA. HOLT, MO. 10 25.4 CURTEA DE ARGES, OUMANIA 6 PLUMS POINT, JAMAICA FUSSEN, BAVARIA 4

2

UNIONVILLE, MD. 1 2.54 1 2 4 6 10 20 40 60 2 3 6 9 12 18 24 5 10 20 30 2 3 6 9 12 24 { { { { Minutes Hours Days Months Duration Figure 3.5 — The greatest observed precipitations (point) in the world (WMO, 1986b). CHAPTER 3 — DATA DERIVED FROM METEOROLOGICAL MEASUREMENTS 33

Wankiewicz (1979). In particular, systematic analyses of runoff func- the only practicable approximation procedure for the estimation of tions based on Denoth et al. (1979) — including information from melt water inputs to hydrological catchment basins. In this respect, no snow profile studies and environmental isotopes as tracers (Herrmann, further decisive progress has been made in the field of realistic simu- Martinec and Stichler, 1979) — are promising. lation of snowmelt runoffs since the introduction of degree-day Meteorological-climatological requirements of point melt factors, snow cover detection according to aerial or satellite images models estimating actual or potential melting rates or potential and analyses of the recession hydrographs. runoff rates for snow covers decrease in the following order: Even valuable studies such as that by Braun (1985) do not (a) Energy balance method (Amorocho and Espildora, 1966; alter this fact. What is lacking, in particular with regard to Anderson, 1976; Aguado, 1983); operational water level and runoff forecasts for snow-covered (b) Combined energy-balance-index method (Anderson, 1973; catchment basins, are improved grid-oriented and operational model Knauf, 1980); systems with related meteorological input data as a prerequisite and (c) Index method; more adequate transformation functions for melt water input, (d) Temperature-wind index method (Martinec, 1960; Braun, despite the relatively positive results of simulations of the meltwater 1985); runoff (WMO, 1986b). (e) Temperature (melt) index (factor) method (Zingg, 1951; U.S. Because of the large number of snowmelt models now Army Corps of Engineers, 1956). available, a categorization is desirable. Classification is, however, The energy balance method has been more widely applied only possible with “fuzziness”. only since the late 1960s, i.e. following the development of Generally, however, it can be said that the input data electronic data processing, as it was then possible to represent requirements for a model increase proportionally to how a adequately the complex physical processes of snow cover physically-based model is planned. If the “process proximity” of the accumulation and, in particular, of ablation through mathematical simulation is used as a differentiation criterion, then the following relationships. Slight modifications of this approach take more or can be differentiated (Kirnbauer, 1993): less full account of the heat fluxes supplying the available melting (a) Physically-based models; energy, including the changes in the latent heat content of the snow (b) Conceptual models; cover. Therefore, for the computation of the atmospheric heat (c) Regression models. fluxes, it is important to have the following parameters: net The data requirements decrease in the given order, radiation, air temperature, air humidity or vapour pressure, wind comparable to the above-mentioned examples. velocity and, if necessary, the temperature of the rain water. In Positive results with the operational analysis and forecast of addition, for the determination of the latent heat content of the snow snow cover development and melt water release from the snow cover cover, the temperature and density of the snow are needed. were obtained, for example, in Germany with the model system As meteorological input data for simulation models of SNOW-D (Rachner, Matthäus and Schneider, 1997). SNOW-D snowmelt according to the energy budget approach, values for time enables the daily calculation and forecast of grid point values of the intervals of up to one hour, depending on the respective application, water equivalent of the snow cover and its melt water release. The are required. Above all, insufficient availability of data prevents the snow cover development is computed with the help of physically- wider application of this method for operational purposes such as, based model components which describe accumulation (build-up, for instance, runoff forecasts for snow-covered catchment basins. increase), metamorphosis (conversion, change) and ablation (decrease, Combined energy balance index methods prove to be melting of snow cover). Model input consists of data on: more suitable. Taking into account the usual lack of data, with these (a) Six-hour interval averages of air temperature and vapour methods energy terms determinable only with a considerable pressure; amount of measuring work are simplified or replaced by equivalent (b) Global radiation/duration of sunshine and precipitation totals values which can be determined more easily. As examples, the of the last 24 hours; approximation of net radiation via the global radiation reduced by (c) Additional data three times a week from a part-time network empirical albedo values for characteristic snow cover surfaces or (depth of snow cover, water equivalent of snow cover); simplified estimations of the sensible and latent heat fluxes with (d) The output data of the regional numerical weather prediction parameter optimization should be mentioned. Unlike the index model of the Deutscher Wetterdienst is used as input for the methods, mathematical relationships are to a large extent physically calculation of forecasts. substantiated. In detail, the following information on model output is On the other hand, the temperature index methods — that available: give information on heat budget processes — include the air (a) Current values of the snow cover (reference point 0600 UTC): temperature and produced mostly only unsatisfactory results, in (i) Snow depth (cm); particular with regard to melting due to radiation. In this context, (ii) Water equivalent (mm); the degree-day method is widely used; hereby, a relationship is (iii) Specific water equivalent (mm cm-1); established between positive daily mean temperatures of the air (b) Forecast values of snow cover development (forecast interval (degree days) and melt water losses supplying degree-day factors. maximum 48 hours, forecasting for six-hour intervals): Slight improvements can be achieved for prevailing advection (i) Water equivalent (mm); melting, e.g. through the extension by wind or even vapour pressure (ii) Precipitation supply, defined as the sum of meltwater terms. release and rain (mm). While the index methods prove, to a large extent, to be The results are provided grid-oriented and with a blanket unsuitable for hydrologically-usable simulations of point melting and coverage for Germany. A summary of the grid values can be made for time resolutions of more than one day, they represent at present for any area required. CHAPTER 4

ESTIMATION OF AREAL PRECIPITATION AND AREAL EVAPORATION

4.1 FUNDAMENTALS Weighted averaging The direct measurement of all hydrologically relevant climatic data The method established by Thiessen (1911) is incorrect insofar as (model input data) supplies point values. To what extent such point the weighting through polygons (polygon method, mid-vertical) values are also representative of the neighbouring region depends originates from Horton (1923). Equation 4.1 is extended by weight on the spatial homogeneity of the influencing factors. An areal value factors K(I): for even a small catchment basin can be determined in this way at 1 N best only in special cases. Alone, remote sensing methods combined Z = ∑ K(I) Z(I) (4.2) with ground measurements (point measurements as so-called N I = 1 ground truth) produce, due to their low spatial resolution, areal values, e.g. precipitation measurement by radar (quantitative with the standardization: assessment) and the estimation of the type of land use via satellite- N measured radiation of the Earth’s surface in the visible and infrared 1 Z = ∑ K(I) =1 (4.3) spectral range (e.g. LANSAT). In this chapter, the most important N I = 1 methods and models for the conversion of point data into areal data have been compiled. This graphic method is, however, difficult to programme when Although the methods described in Subchapter 4.2 are stations are changed. applied mainly to the target parameter areal precipitation, they can also, in principle, be applied to other climatic data such as potential Triangle method evapotranspiration, air temperature and water vapour saturation. This is a variation of the polygon method which is recommended However, this matter is relatively seldom dealt with in the literature. for a concentration of stations in the boundary area (Giesecke and For this reason, the general treatment of areal values in Subchapter Meyer, 1984). 4.2 speaks only of target parameter. In the following discussion, the symbol Z is used for this parameter. This Z(N) value estimated for N Isoline area method spatially-distributed individual measurements Z(1), ...., Z(N) Because of the inclusion of empirical values for the manual analysis deviates from the true value of the target parameter to a degree so of the isoline map, this method is considered relatively accurate. far unknown. However, it requires considerable effort. According to a definition, e.g. a DIN definition, the areal M value is invariably a value of the target parameter averaged over a Z= ∑ K(J) ZZ(J) (4.4) specified area. This area can be the catchment basin under J = 1 consideration (e.g. input for block models) or an areal element of that catchment basin, e.g. isoline areas: areas between isolines as in M: Number of isoline areas; the case of the isoline method (Equation 4.3) or quadratic areas as in W(J): Weight percentage of isoline area J; the case of the grid interpolation procedure with equidistant grid ZZ(J): Z-value of isoline area J. points (Equation 4.5). To avoid misunderstandings, it should be emphasized that Grid interpolation procedure the meteorologist understands the term areal evaporation as the In using the target parameter in a model with detailed areas, it is advis- transition having already been made from a point value to a able to prepare beforehand a gridding of the area. Z is then, for relatively closely limited areal value related to a vegetation crop or instance, the arithmetic mean of the grid values of the total area or of a homogeneous sector. The hydrologist, on the other hand, imposes a definite pixel set. Generally, the target parameter ZR(L) has to be esti- no limitation regarding homogeneity. mated for grid point L of a grid placed over the area. In the equation:

N Z(L)= ∑ A(L, I) Z(I) (4.5) 4.2 METHODS FOR THE COMPUTATION OF I = 1 AREAL VALUES A(L,I) means the weights of N = 8 measurement points (reference Non-weighted arithmetic averaging points) with regard to L, independent of Z(I). These weights are so chosen that the expected value of the 1 N interpolation error equals zero and its variance with regard to A(L,I) Z = ∑ Z()I (4.1) is minimal. This leads to a system of equations, the solution to N I = 1 which is called optimal interpolation or the Kriging method. It has recently been described in detail and applied by Jensen (1989). The This simplest of all conceivable methods for the estimation of areal BONIE method (see Subchapter 4.4), operationally applied by the values should be used only if it is known that there is no Deutscher Wetterdienst, is based, with a few supplementations, also considerable spatial scattering of the Z(I) values. on the method of optimal/statistical interpolation. CHAPTER 4 — ESTIMATION OF AREAL PRECIPITATION AND AREA EVAPORATION 35

A simple variation of approximation consists in using, for error increases with an increasing spatial gradient of the characteristic events, empirical values of the target parameter for target parameter. The use of relative values according to the reference point L and the measurement point I and to weight the Equation 4.6 increases the accuracy of computation measuring point I with the reciprocal square of distance D(I)-2. By (Deisenhofer, Kumm and Wollkopf, 1982); dividing the area around L into four quadrants and selecting the (c) Size of area: closest measurement point (N = 4) for each, Equation 4.5 changes The relative computation error decreases with increasing size into: of the area; (d) Size of the target parameter (e.g. mm, cm3, ¡C, m s-1): N ∑ D(I)Ð2 Z(I) The relative computation error decreases with increasing size = of the target parameter; ZR(L)= C(L) I 1 (4.6) N (e) Time interval: ∑ D(I)Ð2 The relative computation error decreases with increasing I = 1 time interval; (f) Accuracy of point measurement: In this case, C(L) and C(I) are the characteristic values of This is affected by a systematic and a statistic (random) the target parameter for grid point L or the N measurement points I, error. The error of an individual measurement becomes respectively. These may be typical (characteristic) precipitation dominant with increasing density of the station network distributions as in the case under consideration or long-term mean (Mendel, 1979). precipitation amounts. The advantages of this method are: 4.4 COMPUTATION OF AREAL PRECIPITATION (a) Relatively high accuracy of calculations; The application of the methods for the computation of areal values (b) Simple automatic plotting; mentioned in Subchapter 4.2 is relatively simple and reliable (c) Simple programming. because: The disadvantage is the considerable amount of (a) The methods have been developed expressly for the compu- mathematical work in the case of optimal interpolation. tation of areal precipitation or have already often been successfully applied; Mathematical method (b) The network of precipitation stations, unlike that for the In this case, a given function is adjusted to the measurement data of measurement of other climatic values (e.g. water vapour the target value. pressure, air temperature, wind velocity), is relatively dense; Applications are: (c) The spatial precipitation distribution, in particular in the case (a) Multiple regression; of flood-causing precipitation over an extended area, also (b) Bicubic spline function; unlike that relative to other climatic data, is subject to fewer (c) Polynoms; variations; (d) Fourier’s estimations; (d) Point precipitation can be measured directly; (e) Finite element methods. (e) The accuracy of areal precipitation computation can be esti- Since the amount of mathematical work and the deviation mated at least in a first approximation, e.g. by remote of the mathematical function between the reference points can be sensing. considerable, mathematical procedures are not often applied in As shown in Subchapter 4.2, the various methods of the practice. grid interpolation procedure for the computation of areal precipitation can be recommended. 4.3 ERROR SOURCES IN AREAL COMPUTATION Concerning the accuracy of areal precipitation Independent of the target parameter, for which an areal computation computation, reference is made to Subchapter 4.3. Figure 4.1 shows is to be carried out, several fundamental error sources can be that the mean computational error for daily values in an area size of indicated. Areal precipitation computation permits quantitative 600 km2 for a normal density of the synoptical station network (one statements on this subject. However, in this field, and in particular in station per 1 000 km2 in Germany) can exceed values of ± 30 per the field of areal evaporation computation, there is still an obvious cent. Only for a density of one station per 80 km2 (precipitation need for research. The following error sources are known: station network in Germany) is the mean error reduced to about (a) Computation procedure: ± 10 per cent. Non-weighted arithmetic averaging (Equation 4.1) and Convective rainfalls (showers), however, require a higher approximation via mathematical functions (Subchapter 4.2) station density because of their high spatial-temporal heterogeneity can lead to considerable errors. This applies where the distri- with the same mean error; in this case, an additional indirect bution of measuring stations is not adjusted to the spatial measurement through remote sensing (radar) is advisable. gradient of the target parameter, or in the case of uncon- Inaccuracy in point measurement, however, has not been trolled deviations of the mathematical function between the taken into account in this argumentation. As already pointed out in measurement points and in the boundary area; Subchapter 4.3, it becomes dominant when the station density (b) Density and spatial distribution of measuring points: exceeds a certain value. Consequently, a high station density is The computation error increases for an invariable spatial reasonable only if individual point measurements are carried out gradient of the target parameter with decreasing density of with a high degree of accuracy. The result of an investigation in the the measurement point network. Conversely, in the case of a catchment basin of the Neckar River (13 800 km2) into the fixed measurement point configuration, the computation precipitation event of 22 May 1978 was that for a time interval of 36 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES one day, a non-exceedance of about 50 km2 per station does not lead reference period. These are grid values on a geographical grid of 60 to any increase in computational accuracy; this is a total of 257 geographical seconds longitude and 30 geographical seconds stations with a grid point distance of 10 km, altogether 141 grid latitude for Germany. The procedure presumes that with the points (Mendel, 1979). regionalization of precipitation, reference values of a 30-year period The polygon method (Equation 4.2) has, in the past, depending on height, geographical longitude and latitude, exposure frequently been improved for areal precipitation computation by direction of the site and the amount of exposure, the significant taking into account the number and location of the stations as well climatological characteristics of the precipitation distribution are as the season of the year (Diskin, 1969; Israelsen and Riley, 1972). recorded. The remaining deviations in the actual precipitation The isoline method by Meinardus (1900) is a combination measurements at the stations are classified as, due to the weather, of the isoline area method (Equation 4.4) and the grid interpolation non-climatological. procedure (Equation 4.5). With this method, a value is determined The actual amounts of precipitation at the stations can through interpolation from an isoline map for each grid point of a therefore be interpolated according to distance in the form of values grid field. The arithmetic mean of the values then directly produces relative to the precipitation reference value and transferred onto the the areal value. grid (background area method). The relative values interpolated for The grid interpolation procedure in the form of Equation each grid area are converted into data (in millimetres) by 4.6 for the computation of areal precipitation was first developed multiplying with the absolute amounts of precipitation of the and applied in the United States (NOAA, 1972) but subsequent reference area. The corresponding mean monthly precipitation positive experience is also available (Mendel, 1977). A special case reference areas for the period 1961 to 1990 are used as background of the grid interpolation procedure is the so-called catchment areas for the computation of actual daily amounts of precipitation. concentration procedure (Deisenhofer, Kumm and Wollkopf, 1982). The amounts of areal precipitation are obtained as the arithmetic It is routinely applied by the Deutscher Wetterdienst for mean of the grid values that are within the boundaries of the area. hydrological purposes in calculating the amounts of monthly areal BONIE is a procedure which can, on the one hand, be precipitation for about 4 000 catchment basins (base areas). For this, used operationally for flood forecasting and, on the other, for in a derived version of Equation 4.6, the reciprocal station distances hydroclimatological investigations (Kremser and Reich, 1991; from the main centre of the catchment and the monthly means are Reich, 1998). An arbitrary number of spatially-irregular used. measurement values are copied onto a regular network of grid At the Deutscher Wetterdienst, apart from the catchment points, the mesh size of which is at present 6’ in an east-west concentration procedure, two further procedures have been direction and 4’ in a north-south direction. The mesh size can be developed. With these, the actual areal precipitation values of altered but this requires certain preparations. measurements made at intervals of up to one hour can also be made BONIE is based on a learning algorithm derived from the available in a user-friendly form for special hydrological theory of artificial intelligence. On the basis of sufficiently long applications. series of daily precipitation values dependent on relief and weather, The computation of amounts of areal precipitation BONIE separates various characteristic spatial distribution patterns according to the REGNIE method is based on the spatial in a particular analysis region. It derives the various mathematical- equalization of the actual daily, monthly and annual precipitation statistical attributes of this pattern, above all the spatial correlation, distributions by using the regionalized precipitation values of a but also the dependence on the relief and the geographical position, and ‘recognizes’, on the basis of the measurement values of one 80 particular day, the corresponding pattern. These patterns are referred 2 to as background area. [%] 2 1 000 km With the background information taken from this pattern 50 600 km 2 2 it is possible to determine the spatial distribution of precipitation 2 000 km 40 2 more accurately than with any other procedure, also when there are 4 000 km limited gaps in the data. Instead of the background areas derived 30 10 000 km 2 from climatic data, radar measurements can also be used. 20 000 km The analysis regions have an area of approximately 20 10 000 km2 and are usually identical with the river catchment areas or parts thereof. The grid point values are interpolated with the help of the Mean error (%) Mean error 10 statistical interpolation method similar to that of Kriging on the basis of all measurement values available in the area (and not just the nearest as in the quadrant method). The estimated value of the area size f (corresponding to z in Subchapter 4.2) is calculated for 5 the grid point g from the measured values at the stations i according to the following equation:

3 b o b 2 200 500 1000 2000 5000 fw(fˆˆ=+f ∑ Ð f) (4.7) gtgt i it it Area per station (km2) i Figure 4.1 — Mean error of daily areal precipitation computations as a function of the station density for six different area sizes (US where the superscript b refers to the value of the background area Weather Bureau 1947; NOAA 1972; Mendel, 1977). and the superscript o to the measured size of the area. The aim of CHAPTER 4 — ESTIMATION OF AREAL PRECIPITATION AND AREA EVAPORATION 37 the estimation of the interpolation weights wi is to minimize the For a period of 20 years it can be assumed that the sum of interpolation error in the average of all measurements: the positive and negative storage changes ∆W is about zero:

M — — Ð 221 ET = P Ð Q (4.12) ε = ∑ (f Ð fˆ ) = Min (4.8) A A g M gt gt t=1 In the selection of the year series, care should be taken The calculation of the interpolation weights is based on that weather and soil water conditions are more or less average at the following equation: the beginning and end of the period because, if the conditions differ too much from these, then ∆W Ð >0 cannot be assumed. The N calculation of areal evaporation according to Equation 4.12 ∑ (a + o ) w = a ; j= 1 ,..., N (4.9) ij ij i gj presupposes a compact catchment area, the above-ground and t=1 under-ground watersheds of which correspond, which shows no whereby aij and agj stand for the correlation coefficients of the diversion or feed of imported water. The given presuppositions and o b anomalies (fit Ð fit ) from the background area at the stations i and the long period of measurements limit the obvious practicability of j or at the grid point g and at the station j. oij refers to the correlation the procedure for many hydrological tasks. The procedure is mainly of the measured errors at the stations i and j. Without any further used for the adaptation and evaluation of the results of areal discussion, it should be explained that those errors caused by evaporation models. measuring at stations i and/or j, i.e. by their lack of representation Already the older procedures presented in the literature compared with the surrounding area1, are not correlated with each calculate the areal evaporation from meteorological standard data ≠ other, i.e. oij = O for i j (errors due to faulty equipment have to be with the help of regression equations. Kalweit (1953) correlated the rectified using another method). amount of areal evaporation with the saturation deficit and prepared If the formulation of Equation 4.9 is generalized on a with this method a time series of the monthly areal evaporation for system with G grid points, i.e. G right sides, then the matrix is the Spree catchment area. The regression functions were based on written: lysimeter measurements and evaluations of the water balance (A + O)WG = AG (4.10) Equations 4.11 and 4.12. Regression equations are also used for determining areal with (A + O) as a quadratic matrix with N lines. WG and AG, on the evaporation with satellite-based data. The International Satellite other hand, are matrices with N lines and G columns. The wi are Land-Surface Climatology Project (Becker, Bolle and Rowntree, included in Equation 4.7. 1988) computes, for example, the areal evaporation ETA over parts With this equation, interpolation errors in the statistics of of Mali, south-east France and Libya according to the following averages are minimal. An example for the operational application equation: ∆ of BONIE at the Deutscher Wetterdiens is shown in Figure 4.2. ETA = b1 + b2R + b3 T (4.13) The areal value is calculated by mathematically averaging ∆ the grid point values. If the distance between grid points is Radiation balance R and the difference T = T0 Ð T sufficiently small, this arithmetic mean will correspond to the spatial between surface and air temperatures were each measured for daily integral of the measurements — the actual areal value. intervals and for a spatial scale that depends on the resolution of the satellite (NOAA, LANDSAT; MOS). Equation 4.13 can be 4.5 AREAL EVAPORATION interpreted as an abstraction of a Penman variation presented by The areal evaporation of a catchment area can be calculated by Schrödter (1985) and Petznick (1988): b2R is the energy term and ∆ balancing the regional mean values of all water balance elements of b3 T is the ventilation term. Satellite-based measurements for the this area (see Subchapter 4.1): computation of areal evaporation are used within the framework of scientific research. Worldwide, intensive work on the development ∆ PA + ETA + Q + W = 0 (4.11) of models for the quantitative, satellite-based computation of areal where: evaporation is being carried out. PA Areal precipitation (with correction of the systematic error in measurement); ETA Areal evaporation; Q Discharge (above and under the ground); W Available water supply in the area, ∆W Change in available water supply in the period of observation. The evaluation of a possibly existent underground discharge component and especially the change in available water ∆ supply W for e.g. monthly periods requires complicated measuring mm in order to prepare soil moisture and ground water level 75 50 hydrographs. The water balance equation is therefore essentially 25 10 used for scientific investigations into regional water balance and 7.5 5 areal evaporation (Ernstberger, 1987; Jaworski, 1985; Schädler, 2.5 1980; Szabo, 1985). 1 ______Figure 4.2 — Example of a BONIE operational routine plot precipita- 1 In geostatistics (Kriging) the term “Nugget effect” is used for this. tion depths from 03.09.2000 6 a.m. to 04.09.2000 6 a.m. 38 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

In hydrological and water management practices, an calculation is recommended. The areal evaporation is shown inductive procedure is usual for the computation of areal as an arithmetic mean of the amounts of evaporation from the evaporation. The region in question is not homogeneous, thus it is grid areas. Furthermore, the regional distribution of the divided into hydrotopes with a quasi-homogeneous land use, on amounts of evaporation can be shown graphically. The size which the methods described in Subchapter 4.2 are used for the of the grid areas depends on the density of the available data calculation of evaporation from vegetation-bearing areas (taking and the accuracy of the meteorological and non- into consideration the kind of soil and vegetation), sealed areas and meteorological input data as well as the accuracy of the bodies of water. The evaporation mean for the area — the areal evaporation models used (model errors). This calculation can evaporation — results from the amounts of evaporation of the be carried out easily with the help of geographical individual hydrotopes and from their proportion of the area. On the information systems (GIS). In the GIS procedure, the grid basis of this calculation method, areal evaporation time series, values of the soil and land use data intersect with the — which are of interest for many tasks, are prepared, taking into possibly regionally — interpolated meteorological model consideration land use, land use changes, land use scenarios and the data. effects of these on areal evaporation and areal water balance. The models BAGROV (Glugla et al., 1999) and VEKOS Various methods can be used to calculate the mean value (Klämt, 1988) and the water balance model AKWA (Golf and of areal evaporation: Luckner, 1991) are examples of models that have been developed (a) For small areas or for a small number of hydrotopes, the in Germany for the computation of areal evaporation. These areal evaporation is usually easily calculated as the weighted models have been set up for routine and practical use. As input mean value of the evaporation of the individual hydrotopes variables, they use data from the meteorological standard network according to their areal proportions; and the data available for the whole of Germany for soil (b) For larger areas, within which regional changes of the parameters and for land use (see BMU, 2000). meteorological data used in the models occur, a grid CHAPTER 5

QUANTITATIVE PRECIPITATION FORECASTING

5.1 PRECIPITATION mesoscale systems (order of magnitude: 100 km), in lines along Precipitation reaching the ground is the final product of a number fronts and squall lines or in compact form as a cloud cluster. The of processes in the atmosphere. Dynamic, thermodynamical and most impressive phenomenon of organized convection is the cloud-physical processes developing on various scales are involved. tropical cyclone. Owing to the high water vapour content of the air, They extend from the synoptic scale with cyclones via the enormous precipitation depths are observed. mesoscale, with frontal and convective systems up to the Convective overturnings may also occur in areas with microscale, within which the cloud-physical processes occur ascending air on a large scale, e.g. in fronts. Convective clouds are (Figure 2.1). There are interactions between all these processes. then embedded in layer clouds and the stable precipitation is Figure 2.8 shows the large variety of microphysical processes. intensified convectively. Precipitation worth noting normally falls from clouds of Figure 5.1 shows the high variety of precipitation patterns a certain depth. Clouds occur in many forms. They form when the in the area of an extratropical cyclone. Different types of mesoscale air is supersaturated with respect to water or ice. Supersaturation precipitation bands described by Matjeka, Houze and Hobbs (1981) occurs as a result of the cooling of the air following lifting or and Tetzlaff (1987) can be seen. Type 1a designates precipitation mixing processes, i.e. through vertical or horizontal transport fields ahead of and type 1b precipitation fields along warm fronts. In processes as well as through contact with the radiating surface. A the layer cloud, which consists predominantly of stably stratified predominant influence on cloud formation and consequently on layers, small pockets of cold air are observed. These lead to a small- precipitation comes from moisture convergence in the lower layers scale cell formation with the formation of ice crystals which trigger of the atmosphere and from vertical motion. All three aggregate off the Bergeron-Findeisen process in the cloud below and then lead conditions of the water (phases) play a role in precipitation to a locally intensified precipitation band. Type 2 represents a formation. In mid-latitudes, it is in particular the ice phase which precipitation band ahead of the cold front in the warm sector and contributes to the effectiveness of precipitation formation corresponds to a squall line. The precipitation bands along cold (Bergeron-Findeisen process). fronts are organized in accordance with circulation perpendicular to In precipitation forecasting, a distinction is made between the front (type 3) and are associated with clouds of varying vertical stable and convective precipitation depending on the mechanisms depth. In this case, depending on the intensity of precipitation, two of formation. classes are distinguished (type 3a and type 3b). Relatively rarely Stable precipitation forms whenever there is a large-scale cold air advances in upper levels ahead of the surface cold front. ascent of stably stratified and sufficiently humid air caused by Then precipitation of high intensity is observed locally. Precipitation dynamic processes, e.g. in warm fronts. The vertical velocity is only is arranged here in the form of bands or cells (type 4a and type 4b). within the range of a few centimetres. If condensation occurs, then Moreover, in the cold air following the cold front, precipitation extended layer clouds of the type Cirrostratus, Altostratus and bands may occur (type 5). Nimbostratus are formed. Considering the spatial distribution of stable precipitation, it can be said that it covers areas with characteristic lengths of 1 000 km at low horizontal variability. For this reason, it is also designated as large-scale precipitation. The local duration of the rainfall is several hours. On the windward side of mountains, the ascending motion is intensified due to the orography, leading to higher precipitation depth. Convective precipitation presupposes vertical mass overturnings with condensation in unstably stratified atmospheric layers (moist convection). Unstable temperature stratification occurs as a result of heating of lower and/or cooling of upper atmospheric layers due to radiation or advection. Convection clouds (Cumuli, Cumulonimbi) with varying vertical depth are the visible signs for mass overturnings related to moist convection. This occurs in cells with characteristic horizontal lengths of one to 10 km. The vertical velocity occurring in convective clouds is higher by two orders of magnitude as compared to that observed at large-scale lifting. The convective precipitation field is characterized by high temporal and spatial variability. The triggering of moist convection is also influenced by properties of the Earth’s surface (e.g. orography, soil water content). In mid-latitudes, moist convection with showers is a typical phenomenon within the area of cold air, occurring in the rear Figure 5.1 — Typical mesoscale precipitation patterns of extratropical of a cyclone. Convective clouds may organize to form major cyclones (Matejka, Houze and Hobbs, 1981). 40 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

5.2 QUANTITATIVE PRECIPITATION and predicted data. Derivation of the regression equations is made FORECASTING PROCEDURES separately for the individual annual seasons. According to Grebner Quantitative precipitation forecasting includes the spatial and (1982) the method can be evaluated as follows: temporal distribution of precipitation (occurrence, onset, duration, (a) Advantages: intensity) in addition to the forecasting of the type of precipitation (i) Input data (predictors) are variables used in synoptics; (e.g. rain, snow, hail). This requires the knowledge and description and of atmospheric processes on various scales relevant to precipitation (ii) Minor requirement for computer capacity and comput- formation and taking into account their interactions. ing time; First attempts to produce quantitative precipitation (b) Disadvantages: forecasts were based on synoptic or statistical methods. The (i) Limitation mainly to weather situations with large- problem of quantitative precipitation forecasting could be tackled scale precipitation; in the past decades on a mathematical-physical basis. This is due to (ii) Exclusively region-specific procedure; an enlarged knowledge in atmospheric dynamics, thermodynamics (iii) Great variance of results; and microphysics as well as with the aid of the facilities provided (iv) Fixed forecast period of estimations; by numerical weather prediction (NWP) (Chapter 5.2.3). (v) Precipitation estimation possible only in the form of The procedures for quantitative precipitation forecasting classes. presented hereafter reflect the historical development which involve Example: an increase in complexity. Subdivision and designation of the A regression procedure of the Swiss Meteorological Institute various methods are oriented on Bellocq (1980) and Grebner at Zurich produced quantitative precipitation forecasts exclu- (1982). For an evaluation of the procedures, reference could be sively for winter cyclonic weather situations on the northern made to Grebner (1982), who has given a survey on procedures side of the Swiss Alps up to about 1983. Basic meteorologi- applied in Europe. cal assumptions for the pressure and wind field are that the In this chapter, a distinction is made between the forecast area is situated in a frontal zone and the advected air following: masses show a high absolute moisture (Courvoisier, 1970). (a) Synoptic-statistical procedures; The statistical basic material is composed of the following (b) Stochastic procedures; data: the measurements from 25 precipitation stations, the variables (c) Dynamic models; of state of the atmospheric fields (taken from the calendar of (d) Dynamic-statistical model interpretation; and weather situations, tabular radiosonde values of Payerne, weather (e) Remote-sensing procedures. charts) and the four-day precipitation depths of 53 episodes for the The compilation was made mainly from a national point years 1951 to 1967 of at least 10 mm areal precipitation on the of view as at 1998 and presents the methods by means of northern side of the Swiss Alps. prototypical examples. Although the first two methods (a) and (b) Then follows an analysis of the relationship between are now of lower importance in synoptic meteorology, they have precipitation depth and the values of atmospheric variables and the gained renewed interest in climatology and global change analysis selection of the predictors. The four-day mean wind speed at the where variants of these methods are used to predict precipitation 500 hPa level and the four-day means of the dew points in 850, 700 from large-scale climate simulations (“downscaling”) (see also and 500 hPa show the highest correlation coefficients with Lettenmaier (1995), Friend (1998) and Rajagopalan and Lall precipitation depth (predictand). Therefore, regression equations (1999)). A complete international survey must remain reserved for a were set up for them (Figure 5.2). later report. Because of the importance of the mathematical-physical That method produces four-day precipitation depths (areal method, in this chapter emphasis is placed on the description of the means). Application to cases mostly in the winter months of dynamic models. 1967/1968/1969 showed good results. On days during which precipitation was to be expected, the predictor wind speed at 5.2.1 Synoptic-statistical procedures 500 hPa was taken from the American NWP charts. The dewpoints The synoptic-statistical procedures are based on the assumption that were determined from observations of the Payerne radiosonde precipitation on the ground is statistically interrelated with a station and from two stations situated upstream. combination of values characterizing atmospheric conditions as well Then the predictors were inserted in the derived regression as their variations up to the initial time of the forecast. These equation. According to Table 5.1, the mean difference between the interrelations can be described by regression relationships or a measured and the predicted precipitation depths is 8.7 mm. collection of similar cases. The forecast procedures based on these Just as in the case of dynamic models, low precipitation methods produce estimations for precipitation classes and will be depths are often overestimated and high amounts underestimated. explained hereafter under the terms “regression procedure” and The Swiss procedure cannot be directly transferred to other areas. “similarity procedure”. A further disadvantage of the procedure is in particular its limitation to precipitation caused by dynamic and orographic effects and Regression procedure consequently to specified precipitation situations. This procedure establishes regression relationships between predictors (variables of the state of atmospheric fields such as Similarity procedure pressure gradient, pressure tendency, moisture conditions at certain In this case, one proceeds on the hypothesis of similarity in a way pressure levels) and the value to be predicted (predictand, e.g. that cases independent from one another but similar because of precipitation depth). The forecast period is determined by the time synoptic criteria of similarity develop similarly and show period, for which significant relationships exist between measured comparable precipitation depths. The procedure is based on CHAPTER 5 — QUANTITATIVE PRECIPITATION FORECASTING 41 historical data sets (reference cases). For a concrete case, the most respective catchment basin are used as input data. As output synoptically similar weather situation is selected. The relevant data, the procedure produces an interpretation of critical precipitation distribution is used as a forecast for the concrete case. threshold values of areal means of hourly precipitation Unlike in the regression procedure, this method always resorts to intensities. the original, historical data material. There is no regression equation The procedure was developed for different catchment in between. basins, the hydrologically determined effects of the precipitation The advantages according to Grebner (1982) include: being taken as a basis in each case. For this purpose, first the (a) Information on the areal distribution of precipitation as specific limit values of the precipitation depths per event and of the compared to the regression procedure; and hourly precipitation intensities were determined in order to catch (b) Easily estimated precipitation amounts to be expected the only precipitation events which led to floods. The respective provided that there are several similar reference cases. values, e.g. for the Langdale Valley, Cumberland (north-west The disadvantages according to Grebner (1982) include: England) are 36 mm and 6 mm h-1, respectively (Holgate, 1973). (a) Suited only for large-scale precipitation caused by dynamic Between 1954 and 1963 these criteria were met in 35 cases. Based effects; on this experience, the following criteria for continuous heavy (b) Difficult determination of criteria of similarity; rainfalls were developed for that catchment basin (Figure 5.3): (c) Similar initial conditions do not automatically entail corre- (a) A low moves in an easterly or northerly direction along an sponding further developments; Iceland-Cornwall line before it is fully developed; (d) High storage requirement for reference cases; (b) The warm front of the low is being occluded when reaching (e) Valid for 24 to 48 hours, partial time periods of less than 24 the catchment basin; and hours being unpractical; and (c) The relative humidity in the warm air is high and this (f) Rare cases with high precipitation amounts badly repre- uniformly from the ground up to the 650 hPa level. sented in the absence of any special selection procedures. The procedure is not able to catch individual convection It is common to all synoptic-statistical procedures that the cells and the resultant showers. In general, more precipitation was quality of forecasts for convective precipitation is low. predicted than actually occurred so that more flood warnings than Example: necessary were issued. Table 5.2 shows the number of precipitation A similarity procedure for the estimation of heavy rainfalls for events which were above the critical threshold value and were catchment basins in north-west England and north Wales is correctly estimated as compared to the number of events which described by Holgate (1973). Temperature and moisture actually occurred. values from actual weather charts at surface, 700 and 500 hPa levels, as well as measurement data from radiosondes in the Table 5.1 Comparison of measured and computed mean precipitation mm depths from 25 stations on the northern side of the Swiss Alps Measured 4-day precipitation depth (Courvoisier, 1970) 60 Output date Computed 4-day Measured 4-day Difference of forecast precipitation precipitation computed — depth (mm) depth (mm) measured (mm) 40 02.10.67 32.2 24.2 + 8.0 16.10.67 34.2 20.0 + 14.2 31.10.67 33.0 29.8 + 3.2 20 15.11.67 28.6 43.0 Ð14.4 4-day mean of 26.11.67 22.3 15.0 + 7.3 wind velocity at 500 hPa 06.12.67 27.2 18.5 + 8.7 24.12.67 27.2 7.6 + 19.6 20 40 60 80 knots 29.12.67 21.7 28.0 Ð6.3

mm Measured 4-day 02.01.68 25.9 24.0 + 1.9 precipitation depth 60 09.01.68 25.8 29.0 Ð3.2 26.01.68 34.9 32.6 +2.3 16.02.68 13.8 6.4 + 7.4 10.03.68 24.1 27.0 Ð2.9 40 15.03.68 25.3 15.7 + 9.6 18.03.68 38.9 19.6 + 9.3 21.03.68 23.0 12.7 + 10.3 21.09.68 41.9 74.4 Ð32.5 20 24.12.68 36.7 48.5 Ð11.8 4-day mean of dewpoints at 850/700/500 hPa 07.02.69 13.7 11.0 + 2.7 12.02.69 18.3 14. + 3.5 05.11.69 23.9 19.1 + 4.8 -19 -17 -15 -13 -11 -9 °C 08.11.69 33.6 18.7 + 14.9 Figure 5.2 — Interdependence of four-day precipitation depth as areal 12.11.69 33.4 38.6 Ð5.2 mean on the northern side of the Alps and four-day mean of the wind 28.11.69 17.7 6.8 Ð 10.9 velocity at 500 hPa (above) and of the dew points at 850, 700 and 500 03.12.69 24.6 27.4 Ð 2.8 hPa (below) (Courvoisier, 1970). 42 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

Table 5.2 Results obtained with the similarity procedure according to 70 o Holgate for the Langdale region, Cumberland o 70 (Holgate, 1973 modified)

60 Period Critical threshold Number Number of Total number o value of observed correct of forecasts events forecasts 60o 50 o 1965Ð66 24 mm total 11 8 18 4 mm h-1 1966Ð68 36 mm total 13 5 13 40 o 6 mm h-1

o 50 1968Ð72 16 mm total 50 38 77 4 mm h-1 30 o procedures proceed on the assumption that within a series of measurements, there exist mathematically describable statistical 40o 30o 20o 10ow 0o correlations which can be used for probabilistic estimation of the Figure 5.3 — Track and development of cyclones from 29Ð31 October further development of measurement series. The statement whether 1965 (Holgate, 1973). precipitation will fall or not is not possible. A prerequisite for the estimation of precipitation with the aid of the stochastic procedure is 5.2.2 Stochastic procedures that the precipitation event has already started. The physical Stochastic procedures were developed in conjunction with runoff conditions of the precipitation process are disregarded in this case. models for catchment basins with a short system response time in Input data are exclusively measured precipitation data (depth, order to provide runoff forecasts as early as possible to have as duration, intensity). much time as possible for decisions on regulations between the time A reason for using this procedure is the poor forecast of the forecast and the arrival of the flood wave. Stochastic quality of dynamic models for small catchment basins (up to

Precipitation classification according to Temporal and spatial discretization metorological aspects - Event analysis Type K cold air situation - Areal precipitation Type W warm air situation (THIESSEN method) Type GS gradient-weak situation (point measurements because of absence of radar- based precipitation measurement)

Event definition - Only flood-causing events in the summer - Flood definition - Definition of flood-causing precipitation (for modelling only Type K to which most cases are assigned)

Analysis of total precipitation events

Parameters describing future precipitation: total depth, duration and intensity

Regression relationships between parameters or transformations thereof

Highest correlation between the logarithms of precipitation depth and duration, i.e. two-dimensional probability problem (intensity excluded)

Forecast is limited to precipitation depth (one-dimensional probability problem); determination of precipitation duration via regression relationship with precipitation amount

Determination of one-dimensional conditional distribution functions

Adjustment tests for determination of quality of adjustment of different distribution functions

Estimation of parameters of distribution functions according to the method of moments and the method of maximum likelihood

Pearson III distribution for lower precipitation depths Gamma distribution for higher precipitation depths

Figure 5.4 — Preparatory work for the establishment of a forecast algorithm of a stochastic model. CHAPTER 5 — QUANTITATIVE PRECIPITATION FORECASTING 43

500 km2). Stochastic procedures are suited only for such small regressions between the logarithms of precipitation depth and basins and are valid for a short period of one to three hours. On the duration is not so favourable (Klatt, 1983). basis of distribution functions to be determined, the depth, duration All in all, the deviations between estimated and measured and intensity of precipitation are estimated for any given precipitation in the direction of a too-high precipitation estimation exceedance or non-exceedance probabilty of a certain value. show a skew distribution and considerable variance (Figure 5.6). Example: Too-low precipitation estimations or underestimations of major A simple model by Klatt (1983) and Schultz (1987) is based events are relatively seldom. However, deviations are considerable. on the idea that after the onset of heavy rainfall, certain The selection of the adequate non-exceedance probability depends probabilities (i.e. depending on the measurements made up on the goal set. While for the operation of a multipurpose storage, a to the initial time of the forecast) can be found with regard to non-exceedance probability (Pu) of approximately 60 per cent the duration and intensity of precipitation, which can be appears reasonable, the total of all deviations considered over a expected for the rest of the event. The establishment of a longer time period equalling approximately zero; for flood corresponding forecast algorithm is preceded by the work forecasts, Pu > 60 per cent must be selected (Klatt, 1983). described in Figure 5.4. With the aid of the distribution func- For precipitation estimation, Markov-chain models can tions determined thereby, a precipitation estimation can be also be used. On the basis of empirically-determined conditional made according to the flow diagram in Figure 5.5 for each probability functions, they produce any occurrence probabilities for exceedance and non-exceedance probability. future precipitation depths and/or intensities. The verification results refer to non-exceedance For successful application of such stochastic procedures, probability, for which the estimations are valid. In this context, it is it is indispensable to carry out a correlation analysis of already examined whether the number of the precipitation depths which are available precipitation events, including their potential linear over or underestimated corresponds approximately to the given transformations. For a temporal estimation, only data or their linear nonexceedance probability. The result is that the distribution transformations should be used. They show a high degree of functions being used (gamma and Pearson-III distribution) are autocorrelations for as many successive estimation intervals as suited for future precipitation volumes. The result for the possible. corresponding durations of precipitation determined via linear These procedures, which are described, for instance, by Lattermann (1983) and Schilling (1983a, 1983b), do not take into account the precipitation depths of events nor simple non- Determination of non-exceedance probability Pu exceedance and exceedance probabilities, but probabilities of transition from one state into another in the course of an event. Spatial precipitation variability can hardly be described by means of this model technique, even if there is a significant Predicted precipitation depth = f(Pu) spatial correlation of stations. If spatial correlation is insignificant, the results are not usable.

5.2.3 Dynamic models Total precipitation depth = measured + 5.2.3.1 GENERAL CHARACTERISTICS AND SURVEY predicted precipitation depth Dynamic models are physical-mathematical models, so-called NWP models, which describe the spatial-temporal behaviour of

Relative frequency of Logarithmic transformation of total precipitation depth differences between 0.08 predicted and measured precipitation amount PU = 60.0% (positive sign: predicted value too high) In (total duration of precipitation) = 0.04 f (ln (total precipitation depth)) Relative frequency

0.00 -80 -60 -40 -20 0 20 40 Retransformation of the logarithm of Difference (mm) the duration of precipitation Relative frequency of differences between 0.08 predicted and measured precipitation amount PU = 60.0% (positive sign: predicted Predicted duration of precipitation = total duration value too high) of precipitation — duration of precipitation measured 0.04 up to the initial date of the forecast Relative frequency

0.00 -60 -40 -20 0 20 40 60 Determination of future precipitation intensity from known pre- Difference (mm) cipitation depth and related precipitation duration (see above) Figure 5.6 — Relative frequencies of deviations between predicted and measured precipitation at non-exceedance probability Pu = 60 per cent Figure 5.5 — Scheme of precipitation forecast ( Schultz, 1987). (Klatt, 1983). 44 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES atmospheric processes in a deterministic way. In all NWP models models (LAMs) (the model domain covers only a partial area of the which are at present being used operationally by the Meteorological Earth’s surface). The majority of the global models belong to the Services, the formation of precipitation is simulated. This means group of spectral models; limited area models, however, are mostly that apart from other weather parameters, such as surface pressure, grid point models. temperature, wind speed, etc., the weather parameter precipitation is Spectral global models used for weather forecasting have computed at any time. a truncated wave number of T126 to T319, i.e. the smallest wave NWP models are based on the physical laws, such as length which can be represented is between 300 and 125 km. The conservation of momentum, mass and energy. These are reflected models produce forecasts for a time period of at least four days; the in the equations of motion, continuity and thermodynamics. Other European Centre for Medium-range Weather Forecasts (ECMWF) important model equations are the equation of state and — for at Reading (United Kingdom) disseminates forecasts of its global most models — the hydrostatic equation. Models considering models for a time period of seven days. The models can simulate moisture processes use, in addition, equations for moisture satisfactorily the large-scale motion of the atmosphere for the first variables. days of the forecast. The quality of the forecasts decreases with From a mathematical point of view, a weather prediction increasing forecast period. model consists of a discrete system of non-linear partial differential In 1999, the Deutscher Wetterdienst introduced the global equations. After having set the initial and boundary conditions, the grid point model GME (31 layers, horizontal mesh width ∆s ≈ time development of the meteorological variables is computed 55 km) for operational use replacing the global spectral model approximately via the coupled equations (of a prognostic and (19 layers, T106). The GME uses an icosahedral-hexagonal grid diagnostic type) with the aid of special numerical procedures. (i.e. a nearly uniform triangular grid). Depending on the numerical treatment of the differential The production of global ensemble forecasts of the state equations, a distinction is made mainly between two groups of of the atmosphere has become commonplace at many of the world’s models, i.e. grid point models and spectral models. If the derivations operational prediction centres during the past decade (Molteni et al., with respect to time and space are approximated by finite 1996; Tracton and Kalney, 1993; Harrison et al., 1995). These differences, then the model concerned is a grid point model. The ensemble forecasts are predicated on the notion that the state of the fields of the variables are given as discrete values at the grid points atmosphere as derived from all available observations is not known of a three-dimensional grid. In the case of spectral models, however, precisely, but can be represented in terms of a probability the fields of the variables with regard to the horizontal are distribution. Operational ensemble forecast systems attempt to represented as a linear combination of spherical harmonics with sample this initial state probability distribution and then produce time-dependent coefficients. Horizontal derivatives in this case can samples of the resulting forecast probability distribution by be computed analytically (Haltiner and Williams, 1980). integrating each individual member of the sample independently in Important characteristics of the numerical structure of a a forecast model, usually a model developed for use in producing weather prediction model are: more traditional single discrete forecasts. This is called a single- (a) Number of vertical layers; model ensemble prediction approach. In some ensemble prediction (b) For grid point models: size of horizontal mesh width; systems the uncertainty of model formulation is also taken into (c) For spectral models: the highest possible number of waves account by selecting different model options of parametrization of which can be represented on a great circle (truncated wave horizontal diffusion, deep convection, radiation, gravity drag, number); and orography, etc. (Houtekamer et al., 1996). (d) Time stepping scheme. In addition to the single-model approach there also exists The rapid development of NWP models during recent a multi-model approach. This approach uses different models with decades has been closely linked to growing computer power. different physics, numerics and truncations. Each model starts from Increasing computational speed and storage capacity are its own analysis, thus differences in the initial conditions are also associated with increasing horizontal and vertical resolution and present. (Balzer and Emmrich, 1997; Ziehmann, 2000). an extended refinement of the parameterization of physical Many operational prediction centres now routinely deliver processes. a variety of ensemble-based products to forecasters. One example In operational weather prediction models, the following of such a product is a map of contours of the probability that physical processes are taken into account in a parameterized form precipitation amount exceeding 1 mm will fall during a given time in addition to the simulation of horizontal and vertical advection and interval. to the effect of pressure and Coriolis force: A higher spatial resolution (vertical and horizontal) is (a) Turbulent transport of momentum, sensible and latent achieved with the aid of limited area models. The objective of heat; higher resolution over a limited area is to obtain higher accuracy of (b) Formation of stable and convective precipitation within the the near-surface weather parameters (temperature and moisture at hydrological cycle; 2 m, wind at 10 m, cloud cover and precipitation) due to a better (c) Radiation; and representation, e.g. of the Earth surface characteristics. (d) Soil processes. Most of the limited area models are conceived for the Out of the large number of parameters describing the meso-β scale (processes with characteristic horizontal length characteristics of the Earth’s surface, at least orography is taken into between 20 and 200 km) (Figure 2.1). State of the art models are account in the models. characterized by horizontal mesh widths of 5Ð55 km, 15Ð45 layers, As far as the horizontal extension of the model domain is five to 15 of which represent the lowest two km of the atmosphere concerned, a distinction is made between global models (the model (boundary layer). Depending on the computer capacity available, domain covers the whole surface of the Earth) and limited area model domains with up to approximately 100 000 grid points at a CHAPTER 5 — QUANTITATIVE PRECIPITATION FORECASTING 45

Data bank Archive

Process

Archiving

Initial state Numerical Forecast fields Post-processing Post-processing Observations Assimilation (Analysis) forecast (DMO) e.g. MOS results

+ 6h Forecast

Verification

Skill scores

Figure 5.7 — Scheme of numerical weather prediction system at the Deutscher Wetterdienst. level are selected. The forecasts cover a period (forecast period) of boundary values from a global or meso-α model, whereas high- one to three days. Limited area models with the above-mentioned resolution regional models for the meso-γ scale take them from a or with a higher resolution are also called regional models because meso-β model. of the smaller model domain covered. The production of ensemble forecasts with the aid of Examples of regional models for the meso-β scale which limited area models is still a very difficult task. It is presently were in operational use for weather forecasting in 1999 are as confined to a few experimental systems. follows: Information on the state and development of operational Germany: LM (central and western Europe, 35 layers, weather prediction models of the individual Weather Services can ∆s ≈ 7 km); be found in the annually published Numerical Weather Prediction United Kingdom: Mesoscale Unified Model (north-western Progress Report (WMO, 1999). Europe/north-eastern Atlantic, 38 layers, ∆s ≈ 11 km); 5.2.3.2 SYSTEM OF NUMERICAL WEATHER PREDICTION Norway: High-resolution limited area model (HIRLAM) NWP models are embedded in the NWP system. Within this system (Version 1: Europe/North Atlantic, Polar they represent the central component, but predictions are not Basin, northern areas of Russia and America, possible without the entire NWP system. Other important 31 layers, ∆s ≈ 55 km; Version 2: north- components within the NWP system are meteorological western Europe, Norwegian Sea, 31 layers, observations obtained with the aid of the observing system, the ∆s ≈ 27 km); initial state (analysis, initialization) as well as verification. In United States: Eta-32 model (northern America, north-eastern Figure 5.7, the NWP system of the Deutscher Wetterdienst has been Pacific, north-western Atlantic, 45 layers, schematically represented. ∆s ≈ 32 km); The meteorological observing systems described in Canada: GEM model (northern America, 28 levels, Chapter 3 provide surface and upper-air observations as well as ∆s ≈ 24 km). remote-sensing data. Via the GTS, observations (messages) must be The Deutscher Wetterdienst plans to increase the rapidly and reliably transmitted to the individual Weather Service horizontal resolution of the operational model LM to ∆s ≈ 2.8 km in centres (Chapter 6). After having been decoded, sorted and checked 2002. In this second-stage application, the LM model will also for transmission errors and internal consistency, the messages are simulate processes on the meso-γ scale (processes with stored in a meteorological data bank. The observational data are characteristic horizontal length between 2 and 20 km). needed for the description of the atmospheric state at a certain time, Limited area models require time-dependent boundary i.e. for the analysis and for the verification of predictions. values from a model with a coarser resolution containing the area The starting point of any forecast is knowledge of the of the limited area models. Models for the meso-β scale take the initial conditions, i.e. of the three-dimensional state of the 46 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES atmosphere at the initial time of the forecast. For this purpose, the The schedule of global analyses and forecasts in the meteorological variables must be determined with the aid of an present NWP system of the Deutscher Wetterdienst distinguishes extensive procedure at the points of a three-dimensional grid between early runs and main runs. Early runs start about 2.25 hours covering the atmosphere (numerical analysis). Within the analysis after analysis time, whereas main runs start about 4 hours after procedure, meteorological data from irregularly distributed stations analysis time. The advantage of the early run is a quite early are interpolated onto the given grid points. completion of the forecast. The forecast of the main run is expected At the Deutscher Wetterdienst, global analyses are made to be more reliable, though, because more observational data are on the grid of the GME model (∆s ≈ 55 km) valid at 0000, 0600, available when the main run is started. 1200 and 1800 UTC. The following parameters are analysed: The GME computes two main runs a day, which produce surface pressure, geopotential height and wind components at 31 174-hour forecasts and are based on the analyses of 0000 and pressure levels, relative humidity at the lower pressure levels, sea 1200 UTC, respectively. Additionally, three early runs are computed surface temperature (0000 UTC) and snow depth (0000, 0600, 1200 for 0000, 1200 and 1800 UTC to provide lateral boundary values and 1800 UTC). for the LM. Accordingly, the LM is run three times a day to produce The commonly applied method of multivariate optimum 48-hour forecasts starting at 0000, 1200 and 1800 UTC. interpolation is used to analyse surface pressure, geopotential height The operational run time for a 174-hour GME forecast on and wind components. The six-hour GME forecast valid at the the CRAY T3E 1200 distributed memory massively parallel analysis time serves as the initial approximation (initial guess) for processor (MPP) of the Deutscher Wetterdienst is about 2 hours the analysis. 40 minutes. A 48-hour LM forecast run in conjunction with its With the sequence of GME analyses at six-hour intervals corresponding GME run takes about 1 hour. and six-hourly GME forecasts, the data assimilation is performed An operational NWP system which is to produce forecasts by adjusting the model atmosphere to the observations every six under operational constraints requires an efficient computer system: hours. The temporal and spatial aspect of this data assimilation is vector computer, possibility of parallel computations on several expressed in terms of four-dimensional data assimilation (Wergen, processors, computer performance of many GFlops (Giga floating 1984). point operations per second) and main storage for several Gigabytes. For the LM, analyses of the horizontal wind vector, Such conditions can be met only at a central agency of the respective potential temperature, relative humidity and near-surface pressure NMS so that only here can quantitative precipitation forecasts be are continuously made on the grid of the LM (central and western produced with the aid of a routinely operated NWP system. Europe, ∆s ≈ 7 km) by nudging the model atmosphere towards observations. Besides synoptic observations (which are confined to 5.2.3.3 PRECIPITATION FORECASTING WITH THE AID OF THE a fixed time schedule) also continuous observations, such as aircraft ROUTINE MODELS GME AND LM observations, are used. The analyses are stored at hourly intervals. The present NWP system of the Deutscher Wetterdienst consists of Prior to any forecast run of the model, the initialization of the global model GME (global, horizontal mesh width ∆s ≈ 55 km, the fields of variables must be carried out to remove initial noise forecast period 174 hours) for synoptic and meso-α scale from the forecast. This means fine tuning so that during forecasting, simulations and the high-resolution meso-β scale model LM the amplitudes of gravity-inertia waves are not unrealistically large, (western and central Europe, ∆s ≈ 7 km, forecast period 48 hours) to thus impairing the meteorologically important processes (Wergen, allow for short-range weather forecasting in central and western 1984). The amplitudes of gravity-inertia waves must be kept small Europe. Both models are grid point models: GME uses an in order to allow a quality control of observations within the four- icosahedral-hexagonal grid and LM uses a rotated latitude/longitude dimensional data assimilation with the aid of the forecasts over six grid. In the GME model, the atmosphere is represented by 31 layers, hours. With the models of GME and LM of the Deutscher and in the LM, by 35 layers. The GME works with a comparably Wetterdienst, a digital filtering scheme is used in order to reduce the high vertical resolution in the stratosphere, whereas the LM resolves amplitudes of the gravity-inertia waves during the simulation the lowest part of the troposphere best with 12 layers in the lowest (Lynch, Giard and Ivanovic, 1997). 2 000 m. In both models, moisture is included in all layers. The The model results must be subjected to careful values at the lateral boundaries of the LM forecast domain are interpretation by a meteorologist, in particular for local applications. interpolated from GME fields and are updated every hour. Only spatial systems which have a size of at least several grid points The physical complexity of both models is similar. A can be represented realistically, because the differing schemes survey of the numerical and physical structure of the GME and LM usually cause errors in the smallest wavelengths of the forecast field. models is given in Tables 5.3 and 5.4. Furthermore, attention should be given to the fact that the values of the variables at the grid points are representative of the respective Hydrological cycle volume element. They must not be considered as local forecast The hydrological cycles of the GME model and the LM model values for the location of the grid point. This also applies to include the following processes: grid-scale, convective and turbulent precipitation. Statistical procedures such as Perfect Prog and Model transport of moisture and cloud water content; condensation of Output Statistics are an important aid for local interpretation water vapour and evaporation of cloud water; formation of rain and (Subchapter 5.2.4). snow through various processes; freezing and evaporation of falling The systematic daily verification of numerical forecasts rain in the atmosphere; sublimation and melting of falling snow in serves for the control of success and further model improvement. the atmosphere; deposit of rain and snow on the ground. Soil At the Deutscher Wetterdienst, predicted fields of numerous processes are parameterized with the aid of two soil moisture layers variables are compared with the corresponding fields of analysis and a Penman-Monteith type transpiration parameterization and/or with observations. (Chapter 4.1.3.1). CHAPTER 5 — QUANTITATIVE PRECIPITATION FORECASTING 47

Table 5.3 Table 5.4 Global model (GME) Local model (LM)

Global icosahedral-hexagonal grid point numerical weather Non-hydrostatic meso-β scale regional NWP model for central prediction model and western Europe. Prognostic variables Prognostic variables ¥ Surface pressure, temperature, specific humidity, specific ¥ Pressure, temperature, horizontal wind components, vertical cloud water content, zonal and meridional wind components. velocity, specific humidity, cloud water content. Diagnostic variables Diagnostic variables ¥ Geopotential, vertical velocity. ¥ Air density, precipitation fluxes of rain and snow. Numerics Numerics ¥ Horizontal representation: icosahedral-hexagonal grid ¥ Rotated spherical grid, mesh size 0.0625¡ (~7 km), Arakawa (Sadourny, Arakawa and Mintz, 1968; Williamson, 1968), C-grid. mesh size ~60 km. ¥ Second-order horizontal and vertical differencing. ¥ Vertical representation: hybrid coordinate, 31 layers, finite ¥ Hybrid terrain following height coordinates, 35 layers, difference. 10 layers within 1 500 m above model orography. ¥ Split semi-implicit time integration, 200 s time step. ¥ HE-VI (horizontally-explicit, vertically-implicit) time-split- ¥ Eulerian advection for dry model. ting integration, 40 s time step. ¥ Semi-Lagrangian advection for specific humidity and cloud ¥ Lateral boundary formulation due to Davies (1976), boundary liquid water. values provided by the global model (GME) at hourly ¥ Fourth-order linear horizontal diffusion, slope-correction for intervals. the diffusion of temperature. ¥ Fourth-order linear horizontal diffusion. Physical parameterizations ¥ Rayleigh-damping in upper layers. ¥ Grid-scale precipitation including parameterized cloud Physical parameterizations microphysics. ¥ Grid-scale precipitation including parameterized cloud micro- ¥ Mass flux convection scheme after Tiedtke (1989). physics. ¥ Vertical diffusion (Louis, 1979) for the surface layer, an ¥ Mass flux convection scheme after Tiedtke (1989) modified extended level-2 scheme after Mellor and Yamada (1974) by using a CAPE-closure. higher up. ¥ Level 2.5 vertical diffusion scheme including a laminar δ ¥ -two-stream radiation scheme after Ritter and Geleyn (1992) boundary layer. for short- and long-wave fluxes, full cloud-radiation ¥ δ-two-stream radiation scheme after Ritter and Geleyn (1992) feedback. for short- and long-wave fluxes, full cloud-radiation feedback. ¥ Sub-grid scale orographic drag after Lott and Miller (1997). ¥ Soil model after Jacobsen and Heise (1982) with two soil ¥ Two-layer soil model after Jacobsen and Heise (1982) moisture layers and Penman-Monteith transpiration; snow and including snow and interception storage. Climate values interception storage included. Climate values changing changing monthly (but fixed during forecast) in third layer. monthly (but fixed during forecast) in third layer. Fixed analysed sea surface temperature. Topographic data sets Topographic data sets ¥ Mean orography derived from the GTOPO30 data set ¥ Mean orography derived from the GTOPO30 data set × (30'' × 30'') of the United States Geological Survey (USGS). (30«« 30««) of USGS. × ¥ Prevailing soil type from the DSM data set (5' × 5') of the ¥ Prevailing soil type from the DSM data set (5' 5') of FAO. Food and Agriculture Organization of the United Nations ¥ Land-fraction, vegetation cover, root depth and leaf area (FAO). index from the CORINE data set of ETC/LC. ¥ Vegetation cover from the Earth Observation System (EOS) ¥ Roughness length derived from the GTOPO30 and CORINE (NASA) data set (1 km × 1 km). data sets. Operational applications Operational applications ¥ Forecast: initial dates 0000, 1200 and 1800 UTC; integrations ¥ Forecast: initial dates 0000, 1200 and 1800 UTC; integrations up to 48 hours (early run) and up to 174 hours (main run). up to 48 hours. ¥ Assimilation cycle: 0000, 0600, 1200 and 1800 UTC; ¥ Continuous data assimilation by nudging, analysis available integrations up to six hours. every hour.

In the GME and LM models, as normally done in NWP subscale process must therefore be parameterized, i.e. it must be models, the distinction between the two types of precipitation described with the aid of the variables at the grid points. (Subchapter 5.1) is made quite formally. Stable (or large-scale) precip- itation is coupled with clouds filling up at least one grid point volume. Treatment of grid-scale precipitation Mostly, large-scale precipitation areas extend over several grid points. The grid-scale scheme allows four hydrological components: water This type of precipitation is determined in the model directly from the vapour, cloud water, rain, and snow. Between them, a couple of variables at the grid points and is therefore designated as grid-scale interactions is possible. Precipitation formation is initiated by precipitation. The second type of precipitation (convective precipita- autoconversion and nucleation, both temperature-dependent. Rain tion) is coupled with convective overturnings whose characteristic is generated by autoconversion from cloud water droplets which horizontal length is smaller than the mesh width of the model. This have formed after supersaturation of water vapour in a grid volume 48 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES element, initiated mostly by grid-scale dynamic uplifting. A evaporation at the surface is dominant. Convective cloud coverage is stratified cloud exists in this case. Rain drops may grow through not associated consistently with convective rain but is defined accretion from cloud droplets. Snow is formed from cloud droplets independently from the results of parameterized convective through a simplified nucleation process, and it may grow by the processes and is set 0.2. riming of supercooled cloud droplets. This latter process together with the depositing of water vapour on snow are very efficient for Provision of precipitation forecasts precipitation formation in mid-latitudes (Bergeron-Findeisen- Model output process). Further cloud-microphysical processes which are The GME/LM model output of precipitation consists of four parameterized are melting and sublimation of snow, freezing and components: grid-scale rain, grid-scale snow, convective rain and evaporation of rain and evaporation of cloud droplets. Rain and convective snow. During the model run, each precipitation snow fall in the same column where they have been generated, component is summed up at every grid point from the beginning of because horizontal advection of precipitation is neglected. the forecast and is stored in a meteorological data bank at hourly intervals as accumulated values at the model grid points (Direct Treatment of convective precipitation Model Output (DMO)). Convective precipitation results from moist convective processes Analyses of the GME are available daily for 0000, 0600, which occur on scales not resolved by the model. The mass flux 1200 and 1800 UTC; for the LM, the analyses are stored at hourly parameterization scheme after Tiedtke (Tiedtke, 1989) is used to intervals. Forecasts are available as follows (Deutscher Wetterdienst, describe the effects of subgrid-scale convective transports of 1999): momentum, heat and moisture on the grid variables. A stationary GME: one-dimensional bulk cloud model is used to simulate the processes full domain (163 × 842 grid points; 31 levels) taking place in an ensemble of convective clouds, such as up- and 1 h Ð 78 h: every 1 hour downdrafts and lateral mixing (entrainment/detrainment). There are 81 h Ð 174 h: every 3 hours three different types of convection: penetrative, mid-level and full domain (0.75¡ × 0.75¡Ð grid; 480 × 241 grid points; 14 shallow convection. However, at a certain grid point, only one type pressure levels) at a time is allowed. Penetrative (deep) convection is initiated if 3 h Ð 174 hour: every 3 hours horizontal advection of moisture is large, shallow convection if LM: full domain (325 × 325 grid points; 35 levels) 1 h Ð 48 h: every 1 h Wind speed and direction in differ- The above-mentioned fields are found in the WMO GRIB ent heights above code and stored in the meteorological data bank. Just after the the ground model run, data are transferred to the meteorological data archive and kept there for at least one year. The data bank serves as an Ageostrophic wind component interface to all follow-up programmes such as visualization, physical and statistical interpretation of forecasted precipitation (DMO) and hydrologic applications using QPF as input. The spatial distribution of DMO accumulated over specific intervals is visualized. In addition, the temporal evolution of Temperature in precipitation amount at selected grid points over central Europe different heights along with other weather elements is plotted at hourly intervals for T2m: 2 m above ground the model runs of LM (Figure 5.8). These graphical products are T850: in 850 hPa used at the Regional Forecast Centres of the Deutscher Wetterdienst T : in 500 hPa 500 for estimating areal precipitation and for issuing warnings of heavy precipitation.

Dissemination Degree of cloud cover Precipitation data forecasted by the models are disseminated via the Fog/mist GTS in GRIB and GRID code to other NMSs. In addition, data and plotted maps are distributed via Internet, the DWD’s satellite Convection distribution system FAX-Europe and telefax, and the precipitation forecasts are available on the DWD’s server. Several hydrologic Weather institutions in Germany receive LM-forecasted grid-point values up to 48 hours ahead at hourly intervals. Individual information and Ground pressure advice on heavy precipitation and interpretation of QPF are given + : accumulated by forecasters to hydrologists on demand. The Regional Forecast precipitation (12 h) Centre in Leipzig issues estimations of areal precipitation of the Hourly precipitation Snow depth Mulde and Spree river basins in east Germany valid for periods 0600Ð1200, 1200Ð1800 and 1800-0600 UTC once a day. Threshold values which are exceeded with a probability of 90, 50 and 10 per Figure 5.8 — Extract from the meteogram for the BKF grid point cent are subjectively estimated by the forecasters using DMO and ‘Trier’. 96-hour BKF forecast starting on 10 June 2000, 0000 UTC. model output statistics (see Subchapter 5.2.4) products. CHAPTER 5 — QUANTITATIVE PRECIPITATION FORECASTING 49

Quality of precipitation forecasting TSS = 100 (AD Ð BC) / ((A + B) (C + D)) (5.1) Statements as to the quality of precipitation forecasts by means of NWP models should be made only on the basis of the results of a The range of TSS is from Ð100 to 100 with the value being 100 comprehensive verification. Such verification is, however, difficult being for a perfect forecast. as compared to the verification of other weather parameters. Unfortunately, long-term verification results of the The weather parameter precipitation can be described by operational GME and LM forecasts are not available yet. Therefore, several characteristics: onset, duration, type of precipitation, verification results of EM and DM forecasts are additionally intensity curve and precipitation depth. A full verification must presented. The models EM and DM had been in operation for about cover all of these characteristics. seven years until they were replaced by GME and LM in December A basic problem is that the precipitation value predicted 1999. For hydrological applications, mainly the precipitation at a grid point is to be interpreted as a mean value over the area of forecasts with the models DM and LM are relevant. The LM now the grid square, while observations provide point values. For this supports hydrological applications in a similar way as the DM used reason, an areal mean must be derived from the precipitation to. measurements made at irregularly distributed stations, which may The main difference between LM and DM is an increase be situated at different altitudes. Precipitation analysis appears to be in horizontal resolution (DM: ∆s ≈ 14 km; LM: ∆s ≈ 7 km). The particularly difficult in the case of convective precipitation, showing LM also aims at a better description of the initial state by employing considerable areal variability. When making comparisons, attention the nudging technique. Furthermore, the physics of LM (e.g. soil should be given to the fact that for the models, the elevation of the model, turbulence scheme, convective parametrization) will be area assigned to the grid point normally corresponds to the elevation upgraded. The LM is expected to produce forecasts of a somewhat averaged over the grid square. better quality than the DM model. The current precipitation verification at the Deutscher To see how well the models capture the regional Wetterdienst is carried out in two modes (Damrath et al., 2000). In distribution of precipitation height, observed and forecasted the first mode, all available information from synoptic stations in precipitation distributions for February 1999 over Germany and Germany and Switzerland is used to assess the quality of the 12- over Switzerland are shown in Figure 5.9. It is obvious that with the hour precipitation amount twice a day. This set of observations high resolution DM and, especially LM, specific patterns of the covers about 240 stations. The second mode uses about 4 200 precipitation field can be simulated fairly well. This particularly stations in Germany only. These stations report only the 24-hour applies to orographically-forced large precipitation amounts over precipitation amount, but they do not give reports in real time. the mountainous regions of southern Germany and Switzerland. The determination of statistical measures for the quality However, both an increase in precipitation on the windward side of of QPF is based on the comparison of station measurements of the mountains and a decrease on the lee side are exaggerated by the precipitation amount with the forecasted precipitation amount at the models. This effect is observed not only in winter, when problems in model grid point nearest to the station. With these data, contingency the measurement of falling snow might lead to wrong conclusions tables (see Table 5.5) for different thresholds of precipitation height concerning the verification of precipitation, but also throughout the are evaluated. In the first (synoptic) mode, the thresholds 0.0, 2.0, summer (Damrath et al., 2000). and 10.0 mm/12 h are used. In the second (non real time) mode, the The following statistics (Damrath et al., 2000) thresholds are 0.0, 1.0, 2.0, 4.0, 8.0, and 16.0 mm/24 h. From the demonstrate the long-term trend of precipitation forecast quality, the results A, B, C and D in the contingency table (see Table 5.5) differences in forecast quality of EM and DM and the forecast various scores can be derived for ‘per cent correct’, ‘probability of quality with regard to day/night. Figure 5.10 shows the 12-month detection’, ‘false alarm rate’ and most importantly ‘true skill moving average of TSS for the thresholds 0.0 mm/12 h and statistics’ (TSS): 10.0 mm/12 h, respectively, for the forecasts starting at 0000 UTC and 1200 UTC. As can be seen, both the EM and the DM have nearly the same quality for the threshold 0.0 mm (Figure 5.10a and Figure 5.10b). Therefore, the decision ‘precipitation yes or no’ will be issued by both models with the same accuracy. This holds until September 1997, when the former 20-level version of DM was replaced with a 30-level version. Moreover, the area of DM was enlarged significantly (163 × 163 grid points instead of 109 × 109 grid points, the resolution remaining unchanged). The model forecasts for precipitation during the day are in general better than

Table 5.5 Contigency table for different thresholds of precipitation height

Forecast over a Observation over a given threshold given threshold NO YES

Figure 5.9 — Section from BKN grid (horizontal mesh width 127 km). NO A B × Marked are 6 7 = 42 grid points, of which predicted precipitation YES C D values are transmitted to the Federal Institute of Hydrology (Koblenz). 50 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

(a) (b)

(c) (d) True skill statisticsTrue skill statistics True True skill statisticsTrue skill statistics True

Figure 5.10 — Twelve-month moving average of TSS.

The verification statistics generally show better model performance during the winter months as compared to the summer months (Damrath et al., 2000). This points to a general problem of correctly parameterizing convective precipitation. This may result in comparatively poor QPF input to hydrologic models in cases of summertime flash floods. The use of the new models GME and LM is expected to improve this situation.

True skill statistics True Finally, it is demonstrated how the numerical QPF forecast quality varies the with forecast period and threshold value. Figure 5.11 presents the TSS of DM and EM forecasts for different thresholds (‘class limits’) and forecast periods (1Ð3 days ahead). Additionally, the TSS of a persistence forecast (PE) is shown. A persistence fore- Class limits [mm/day] cast assumes that the observed precipitation amount will persist in the Figure 5.11 — TSS of DM and EM forecasts. future. Such a forecast often serves as a baseline for minimum fore- cast quality. Figure 5.11 clearly shows a decrease of forecast quality those for nocturnal precipitation. The advantages of the high- with increasing forecast period. This is due to the chaotic nature of the resolution DM compared to EM are obvious for precipitation atmosphere. Furthermore, the forecast quality decreases with increas- amounts of more than 10 mm/12 h (Figures 5.10c and 5.10d). Only ing precipitation threshold. Thus, it is more difficult to forecast a during the first year of the moving averages presented in Figure 5.10 severe precipitation event than to issue the decision ‘precipitation yes can a positive trend in forecast quality be seen. This is somewhat or no’. surprising since for other weather elements, such as wind and A correct precipitation forecast is an important factor in temperature, an increasing forecast quality has been noted in recent the forecast of floods. Several examples of numerical precipitation years. Again, it is confirmed that ‘precipitation is notorious for forecasts relevant for flooding events are presented in Damrath et al. being difficult to quantitatively predict accurately’ as stated by (2000). In most cases, the former operational models of the Gaudet and Cotton (1998). Deutscher Wetterdienst (EM and DM) achieved good results, and In case of a threshold of at least 3.0 mm/12h, the EM the potential of the models of the current operational system (GME forecasts of QPF are superior to the forecaster’s judgement (Balzer, and LM) to realistically simulate these situations is even higher, this 1998). Only for thresholds of 0.5 mm/12h and 0.0 mm/12 h is the being very important for hydrologic applications. forecaster able to provide better forecasts than the EM model. In the The temporal evolution of precipitation in situations last five years, forecasters have not been able to improve further the which had led to flooding events in the federal state of Baden- EM/DM forecasts of rare precipitation events (Balzer, 2000). Württemberg (Germany) was investigated by Cress (1997). There CHAPTER 5 — QUANTITATIVE PRECIPITATION FORECASTING 51 was sufficient similarity between forecast and observations, if the predictions, time and location of the development of deep precipitation heights were averaged over an area. The comparison convection cannot be expected to be realistically represented by LM of precipitation at individual stations, however, revealed major due to the chaotic nature of convection (Frühwald, 1998), thus differences in terms of phase and intensity. Thus, the precipitation requiring a careful interpretation of these upcoming QPF products. forecast at a model grid point must not be interpreted as a prognosis of precipitation at a single location and a single point in time. It is 5.2.4 Dynamic-statistical model interpretation suggested to use rather a window in space and time to estimate the NWP models produce precipitation values which are assigned to a exceedence probability of a certain precipitation threshold at a grid point and are to be interpreted as a representative value for the specific location. It is expected that this view will become even respective grid square. In practical weather forecasting, however, more important when using the higher resolution meso-γ models in forecast values for individual locations are also required. It is future. possible that the elevation of such locations deviates considerably The results of precipitation verification described so far from the elevation assigned to the grid point, which implies that the give an impression of the average model behaviour with regard to weather parameters computed at the grid point cannot be simply season, day/night, model, threshold value and forecast period. This transferred to such a location. With the aid of statistical follow-up information can be used only to a limited extent for the estimation procedures, it is possible to interpret the model results with of the quality in a specific individual case. Minimization of the reference to the respective location. Two methods are distinguished systematic error in precipitation forecasting by means of statistical in this context: model interpretation is useful (Subchapter 5.2.4). For interpretation, (a) Perfect Prog (PP (Klein, Lewis and Enger, 1959); and it is helpful to use the results of simulations starting from preceding (b) Model Output Statistics (MOS) (Glahn and Lowry, 1972). initial dates, to consider the precipitation values at neighbouring Both methods are based on the fact that from the data set grid points, to have a look at precipitation forecasts of other models of the predictors and of the required parameter (predictand, e.g. or to use ensemble forecasts. observed precipitation at the respective location), a relationship between predictors and predictand can be derived in the form of a 5.2.3.4 OUTLOOK multiple linear regression equation by means of a statistical The computer capacity at the Deutscher Wetterdienst will be further analysis. Using that equation, the local predictor can be estimated upgraded during the coming years. A sustained speed of 0.5 in the actual case. That method is also applicable to areal TeraFlops on the MPP system is expected to be reached in 2002. precipitation, provided the area is smaller than the area of the grid This will allow a further improvement of the horizontal and vertical square (∆s)2. resolution of the operational GME and LM models. The final stage In the case of the PP method, the predictors are measured of the GME/LM system will be achieved in 2002. The horizontal or analysed meteorological parameters; in the case of the MOS resolution of the models will then have reached 28 km and 40 layers method, the predictors are selected from the set of parameters for GME and 2.8 km and 50 layers for LM. Output fields of LM predicted from the model (Pander, 1982). In both cases, particularly will be provided every 15 minutes resulting in an enormous demand those parameters which have a causal link to precipitation are suited for data storage (Frühwald, 1998). as predictors. In the case of the PP method, a more comprehensive By this substantial increase of resolution, the LM will database (long-term observations) is normally available than in the resolve processes of the meso-γ scale (2Ð20 km) which aims to case with the MOS method; for MOS, only model data from a time improve local weather forecasting. Large convective systems can be period during which no changes were made to the model can be resolved explicitly. This offers the possibility to forecast the used. A change made to the model requires recalculation of the transient precipitation patterns associated with such systems. In multiple linear regression equation; however, minor changes to the particular, in areas of strong orographic forcing, realistic convection model either do not have or have only a minor effect on the patterns may be produced even when the initial field is interpolated interpretation results. For a prognostic application of the PP method, from a model with a coarser mesh. Also, the influence of the ground the predictors have to be replaced by predicted model values. on the time evolution of weather systems will be simulated more However, no specific prediction model is binding. accurately by LM. At the Deutscher Wetterdienst, a statistical interpretation Cloud ice will be a prognostic variable allowing the scheme on the basis of MOS is applied to the GME results. It has simulation of Cirrus clouds. In the medium term, rain and snow will been developed in close cooperation with Knüpffer (1998) for be treated through prognostic equations. The three-dimensional general forecasts and for special application in aviation meteo- transport of hydrometeors may modify the precipitation pattern on rology. The MOS system was originally designed for the application the lee side of mountain ranges. to EM results. The data set for the evaluation of the regression The continuous data assimilation (nudging method) for equations consisted of daily EM forecasts up to 78 hours lead time LM will also exploit remote-sensing data such as radar and and of hourly synoptic observations of 120 German and 120 lightning data. LM is envisaged to run in ‘nowcasting on demand European stations. The data set covered the time period form mode’ in case of convective weather. Radar information has to be January 1992 until December 1998. The MOS system was then converted to a modified initial state at hourly intervals. Thus, a transferred to GME by rederiving the regression equations using the simulation with a forecast period of two to eight hours can start seven-year data set mentioned above and additional GME forecasts every one hour for estimating the propagation and evolution of recomputed for a two-year period. The GME-forecasts were given a severe weather associated with deep convection on the meso-γ scale. weighting factor of 2 in this process. The results so far have justified Results of LM test simulations on the meso-γ scale with this procedure. 2.8 km mesh size indicate that dynamics of deep convective cells The MOS system uses a set of about 150 potential will be reproduced rather realistically. However, for 24-hour predictors of different types. Basically, the DMO variables 52 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

(geopotential height, temperature, ...) are used. A second set of Statistical model interpretation is applied for local predictors consists of derived model variables (e.g. temperature precipitation forecasting at several Weather Services. Derivation of advection and vorticity advection). Persistency predictors such as statistical relationships requires a single, comprehensive statistical the latest observation or the latest statistical forecast are used as analysis of large data sets, but little computing time is required well. A variety of other predictors complement the set. The when the method of statistical interpretation is applied daily. predictands have been defined partly in categorical and partly in Interpretation can be carried out with reference to a location or to probabilistic form for general and special requirements. For QPF, an area (smaller than the area of the grid square and forecast periods both categorical and probabilistic results are evaluated, the latter of less than 48 hours are recommended). providing probabilities for the precipitation height exceeding 0.0, MOS corrects systematic model errors, with corrections 0.2, 1.0 and 5.0 mm for six and 12 hours (Damrath et al., 2000). being more successful for stable precipitation than for convective The regression algorithm has been optimized with respect precipitation. In the case of the PP method, the systematic model to root-mean-square error minimization of the predictands on error growing with increasing forecast period simultaneously entails independent data, thus avoiding problems due to statistical a decrease in the quality of the interpreted values. According to overfitting (Knüpffer, 1996). Some of the predictands have to be Brunet, Verret and Yacowar (1988), for the first section of a 72-hour transformed in order to adjust the sensitivity of the regression forecast period, the PP method is more suitable and for the second algorithm to specific ranges, as required. For QPF, a square-root section, the MOS method is more suitable. For rare extreme transformation both for DMO and for the predictand prevents the precipitation events, only a slight improvement is achieved with few cases with very high precipitation amounts from dominating the both methods. regression equation. This allows a better forecasting of cases with Apart from MOS and PP, a Kalman filter technique is lower precipitation amounts. often used to refine the DMO. The statistical correction by a Even with the refinement of QPF by MOS, the Kalman filter aims to reduce systematic model errors which are predictability of precipitation is less than the predictability of other often caused by a mismatch between model and station topography. weather elements. This was revealed by a predictability study by Knüpffer (1966) on the basis of data from the European Centre for 5.2.5 Remote-sensing procedures Medium-range Weather Forecasts. The MOS forecasts of different Unlike conventional measuring networks, radar and satellite data elements were compared to a reference forecast, which was an provide areal information. Measuring methods and accuracy of optimum combination of persistency and climate. The predictability radar-based precipitation measurements and satellite-based was defined as the lead time of the MOS forecasts, for which the precipitation estimations are described in Subchapter 2.2. reduction of variance of the MOS forecasts, as compared to the Irrespective of the primary purposes, such remote-sensing data can reference forecasts, became less than 10 per cent. The predictability also be used for quantitative precipitation forecasting in different depends on the weather element. For this study, it was much more ways. than eight days for temperature, about 5.5 days for wind and five For so-called nowcasting (forecasts for periods up to days for QPF. The situation becomes much worse if the true skill 2 hours), the cloud and precipitation fields observed at 15Ð60 min statistics (TSS) method is used. The skill of a six-day temperature intervals are moved together with the observed or predicted wind or forecast is similar to the skill of a three-day wind forecast or of a in accordance with the travel direction of the precipitation area one-day QPF. observed so far. Assuming that the prediction intensity observed is The MOS-system runs several times per day on the basis maintained, this extrapolation method is a practical method for of the operational model runs. The results are presented to the quantitative precipitation prediction for time periods of up to forecasters in numerical and graphical form. The MOS forecasts are 2 hours. In addition, estimates of changes in intensity can perhaps also integrated into the operational verification scheme. Here only, be made from the time development of the weather system as a an ensemble of 14 stations in Germany is considered. The MOS- whole. system for EM forecasts (EMOS) leads to a degree of refinement In NWP, remote-sensing data serve the improved similar to that of the MOS-system for GME forecasts (GMOS) description of the initial state of the atmospheric parameters. This (Balzer, 2000). The results for EMOS for the period from October is particularly true for humidity over areas with only a few 1998 to March 1999 are shown in Table 5.6. For temperatures and observations at the surface or in the free atmosphere. Vertical clouds, the MOS product for EM greatly enhances the quality of the profiles of humidity derived from satellite-based measurements EM forecasts. The improvement for QPF is, however, not as high and/or information on clouds and precipitation from remote-sensing and depends on the rainfall rate. Rainfall amounts of more than data are used to get information on the distribution of humidity. 10 mm/12 hours even show a degradation of the forecast quality by Also, precipitation observations can be used to estimate MOS. This seems to be a result of the square root transformation roughly the latent heat release which is associated with the mentioned above (Damrath et al., 2000). formation of precipitation. This latent heat then serves as a forcing Table 5.6 agent during the first hours of the simulation. Thus, vertical motion, EMOS results (October 1998ÐMarch 1999) which influences precipitation to a decisive extent, can be simulated RMSE TSS (RR/12 h) more realistically during the first hours of the model run, as far as structure and amount is concerned. Jones and Macpherson (1997) T TX N dd ff >.0 >.5 >3 >10 proposed a latent heat nudging technique which shortens the period EM 2.22 2.23 33.3 35.5 1.79 61 61 57 47 of adjustment of the mesoscale model (up to 12 hours) and thereby DM 2.32 2.27 33.4 31.8 1.72 62 60 59 58 improves precipitation forecasting. There are worldwide efforts (in western and eastern EMOS 1.55 1.49 23.6 28.3 1.50 60 65 62 35 Europe, Canada, Japan, the United States, and other countries) to CHAPTER 5 — QUANTITATIVE PRECIPITATION FORECASTING 53 develop combined measuring systems from conventional and whether such a QPF is available, where there is a lack of confidence remote-sensing data, and to make them usable for quantitative in QPF or whether the runoff model is not suited for an application weather forecasting through data exchange. The western European of QPF. compound project COST 73 (Cooperation in science and QPF of the operational NWP models at the Deutscher technology) of the European Commission, supported by 16 Wetterdienst are widely used to issue flood warnings and to provide countries, should be mentioned. The most developed system was input data for hydrologic models used by the water authorities of the FRONTIERS plan of the United Kingdom which was some of the federal states of Germany. At several flood forecast presented already in Subchapter 2.3. The assimilation of radar data centres in Germany, QPF at model grid elements for specified into operational NWP models is in its infancy, but there is much periods is visualized for monitoring the meteorological situation potential for growth in this field. The COST-717 Working Group with respect to a potential flood situation. During the flood event, on Using Radar Information for Assimilation into Atmospheric the further development of the flood situation is assessed with QPF. Models was established in autumn 1999. It offers a framework for In addition, forecasted areal precipitation for catchment areas European radar and NWP scientists to achieve this potential. computed from DM (now LM) grid-point values are employed (Fell In order to make a global exchange of remote-sensing and Prellberg, 1999). In mountainous areas, such as in the Bavarian data via the GTS possible (Subchapter 6.1.3), the BUFR code was portion of the Alps, fast catchment response to rainfall QPF is the developed. only basis for pre-warning heavy precipitation. Task forces, such as the police and local fire brigades, can be put on alert for observing 5.3 NEED FOR, AND REQUIREMENTS OF, QPF ON the potentially approaching flood situation. BEHALF OF HYDROLOGY AND WATER At the Flood Forecast Centre in Karlsruhe (Baden- RESOURCE MANAGEMENT Württemberg, Germany), QPF of DM (now LM) is the input Population growth in conjunction with increasing colonization of parameter to a regression algorithm estimating the water levels at the alluvial plains signals an increasing need for water stage and about 30 gauging stations in Baden-Württemberg for 48 hours ahead discharge forecasting. Moreover, there is a varying intensification (Homagk, 1996). If the forecasted water levels exceed a specified of floods caused by man, related with an acceleration and increase threshold value, this information can be used for switching the data in flood waves. An unfavourable distribution of precipitation transmission of the automatic ombrometer network to an hourly caused by a change in the climate, which is being discussed, could interval or for putting the flood forecasting system on alert also have a negative effect on the discharge characteristics. (Homagk, 1999). Figure 5.12 shows observed and forecasted Precipitation forecasts are particularly needed by hydrographs at the Heidelberg gauge (River Neckar, Germany). operational water stage and discharge forecasting, i.e. they are Using DM-forecasted precipitation amount for the catchment area transmitted to the user. However, two restrictions should be of River Neckar, the forecasted maximum and the timing of the mentioned in this context: maximum correspond better to the observed values. (a) A discharge forecast for a time period during which the The most important use of QPF is made for precipitation- discharge hydrograph is predominantly characterized by the runoff models where rainfall is the most sensitive parameter after water stored in the catchment basin can be computed snow. In Germany, precipitation amount forecasted by the former according to experience made without using measured and BKF model and statistically improved with the MOS approach was predicted precipitation (e.g. 18-hour forecast for the Rhine 3 -1 gauge Worms; Bundesanstalt für Gewässerkunde, 1987); and Runoff (m s ) (b) Precipitation forecasts available at present and in the past for water stage and discharge forecasts have not always met the requested requirements for accuracy. This applies, for example, to the mesh width for grid point forecasts and the influence of the orography. Recalculations with subsequently observed instead of predicted precipitation amounts indicate the influence of the QPF error (Bundesanstalt für Gewässerkunde, 1987). A high variety of applications of operational water stage and discharge forecasts is known. This is reflected in the evaluation of a global questionnaire. Out of 107 questionnaires sent to different agencies, 78 were returned. According to those questionnaires, flood protection is a main reason for application, but also navigation, energy and water supply (irrigation, drinking water, low flow augmentation, recreational value) and water quality are mentioned. The size of the catchment basin is between 100 km2 and 100 000 km2, the forecast period between two hours and six days. In 90 cases, precipitation data are also used, but unfortunately it remains unclear to what extent predicted precipitation is included. In this case, just as in other literature, the question as to the reasons for a possible non-application of the QPF is not raised. 20 December 1993 21 December 1993 22 December 1993 It would therefore be interesting to know an answer to the questions Figure 5.12 — Hydrograph at the Heidlberg gauge (River Neckar). 54 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES successfully utilized for discharge forecasts at gauging stations in tutions responsible for flood forecasting. The number of stations the Leine river basin within a pilot study more than 10 years ago measuring the water equivalent of snow cover will have been (Holle, 1987). At present, precipitation-runoff models are widely enhanced to about 600 in Germany by 2004. applied in Germany where they are mostly combined with flood On the whole, operational quantitative precipitation routing models (BMU, 1997). forecasting is of high economic importance and not only through A flood forecasting system based on these types of early warnings to flood-prone areas. Among those who profit are models has been in operation since 1995 for the catchments of River not only public utility companies responsible for irrigation, drinking Neckar and for the eastern tributaries of the upper Rhine River. This water and electricity, but also agriculture, forestry and the building system is the first one which employs both precipitation amount industries as well as traffic and tourism. measured by ombrometers at 120 stations and forecasted by DM In recent times, short-term (up to 72 hours) precipitation (now LM). By using forecasted precipitation amount, the accuracy forecasts have gained new importance in conjunction with water of water level forecasts has been improved, the forecast range pollution through agriculture, industry, urban drainage and extended, e.g. at the gauging station Heidelberg from 12 to households. Thus, for instance, the bridging of low-flow periods 24 hours, and forecasts at some additional gauging stations provided predicted on the basis of high pollutant concentrations in the (Homagk and Schulz, 1999). respective water and reduction of that concentration through low The River Moselle forecasting system was tested daily by flow augmentation or the deliberate discharge of polluted waters using DM forecasts. This comprehensive three-model system during predicted flood periods are, though generally applied, developed and operated at the Bundesanstalt für Gewässerkunde in dubious methods. In this context, the increasing pollution of shallow Koblenz comprises a hydrodynamical, a statistical, and a waters and bank storage should also be mentioned. The quantitative precipitation-runoff model (Krahe, 1998). The latter covers sub- problems linked with the problems of quality therefore increasingly basins of the River Saar, which is one of the larger tributaries of the force drinking water works to make use of precipitation forecasts. River Moselle. It was coupled to DM data, such as precipitation, The requirements to be met by QPF for the purposes of 2 m temperature and humidity, 10 m wind speed, and cloudiness, at operational water stage and discharge forecasting can therefore be 6-hour intervals. Evaluations of test runs show, on the one hand, reduced to the following: good results concerning the forecasted onset/duration of the (a) Forecast as early as possible of the beginning and duration of precipitation event. On the other hand, a tendency for a precipitation event — While QPF is made continuously for overestimating the precipitation amount was found (Krahe, 1998). various reasons, water stage and discharge forecasts are of The precipitation-runoff model is now being used as a tool to verify interest only as regards extreme conditions, in particular the QPF forecasted by LM. floods. In this case, one speaks of event forecasts. Flood- Reservoir management is another application of QPF. causing precipitation must therefore be predicted early so Controlled release is carried out, e.g. for the Bigge reservoir in the that the data set can be updated and a forecast can be made Lenne river catchment in Nordrhein-Westfalen (Germany) to ensure of the rising water stage hydrograph. This leads to the minimum discharge along the lower reaches of the River Ruhr and second requirement; to protect the town of Altena from flooding (Ruhrverband, 1994). (b) Validity of the forecast as long as possible — The term “as QPF of the Deutscher Wetterdienst is used for the precipitation- long as possible” in this context describes the forecast period runoff model for the River Lenne (Göppert et al., 1998) which is using a QPF which shows sufficient accuracy for the user the main tributary of the Ruhr River. In order to estimate the and exceeds the forecast period of forecasts produced bandwidth of discharge, three runs with the precipitation-runoff according to other methods (e.g. empiric). Normally, the model are performed based on original QPF values and precipitation forecast is short term, i.e. two days ahead. Two-day forecasts amounts derived from the weather bulletins issued by the Regional even in major catchment basins draw fully on the retention in Forecast Centre in Essen and the Agrometeorological Advisory the catchment estimated on the basis of precipitation Office in Bonn (Morgenschweis, 1999). This ensemble technique, measurement and precipitation prediction. Because of a which has been applied in medium-range weather forecasting for shorter concentration time (maximum flow period up to fore- several years, seems to be a promising approach to get information cast, gauge) in the smaller catchment basins for the same on the reliability of the forecast results in hydrology. period of forecast, emphasis is on QPF so that it has a In May 1999, the successive reservoir management of the stronger impact on the discharge curve. The period of the Sylvenstein reservoir (Bayern, Germany) based on DM forecasts water stage and discharge forecast in smaller catchment protected the town of Bad Tölz from inundation by the Isar River basins with less than 1 000 km2 therefore is relatively short, during the Bavarian flood event (Kästner, 1999). The catchment of normally less than one day. But in this case, the question of the Sylvenstein reservoir covering an area of 1 100 km2 was hit by the spatial distribution of the predicted precipitation becomes an excessive rainfall of about 200 mm in 48 hours. relevant, which leads to the third requirement; Last but not least, precipitation (6-hour totals) of the (c) Spatial and temporal discretization as high as possible of the Deutscher Wetterdienst is an essential input parameter to the SNOW predicted amounts — This means that QPF screening suffi- model (Rachner, Matthäus and Schneider, 1997). This model provides ciently covers the area of the forecast as well as the spatial both diagnostic and forecasted water equivalent of snow cover and precipitation distribution. meltwater release. The model domain covers all of Germany, but a dense network of stations measuring the water equivalent of snow 5.4 EVALUATION OF PROCEDURES AND THEIR cover three times a week exists only in four of the sixteen federal FURTHER DEVELOPMENT states. Output values of snow are provided on a 7-km grid to be used Comparison of procedures for quantitative precipitation forecasting as input to precipitation-runoff models operated by hydrologic insti- and their evaluation depends on the respective applications. CHAPTER 5 — QUANTITATIVE PRECIPITATION FORECASTING 55 1 km × Remote-sensing procedures (radar) (radar) procedures 2 h 1 h after the beginning of the event 1 h after the beginning Workstation Some minutes ≤ 1 h 2 about factor Accuracy Area of radar network 1 km 48 h interpretation ≤ Dynamic-statistical grid square Workstation minutes Several 12 or 24 h Local interpretation possible; systematic forecast error is reduced; cannot be captured events rare extreme Area covered by the model Area covered places; also areas Individual ≤ Mostly Some minutes after completion of model simulation which is a prior condition for the application of dynamic-statistical interpretation Dynamic models 30 minutes for 24 h forecast Ð 1/4 of hemisphere for limited area models ≤ 5 h after main observation times: 0000, 1200 UTC 5 h after main observation ) ) Global 7 d (global models) 78 h (limited area models) Ð b a Satisfactory for grid-scale; not yet satisfactory for for grid-scale; not yet satisfactory Satisfactory precipitation (further details are designed convective to be assessed by general inquiry) processors)/several computer (several Vector 15 Mwords Grid square ≤ ≤ 1 h ( ( After completion of model simulation, about 3 Table 5.7 Table 2 100 km PC Some minutes precipitation and (advective) a single event Ð Suitable only for large-scale precipitation and (advective) a single event Hydrological relevant characteristics of the different procedures procedures of the different characteristics relevant Hydrological Hydrological relevant characteristics of different QPF procedures characteristics of different relevant Hydrological 48 h Hours 10 minutes Ð Ð Workstation 5 Regression proceduresRegression Similarity procedures models Stochastic Area covered by the procedures' individual locations by the procedures' individual Area covered Up to about 96 h 24 Unsatisfactory. Only local interpretation for frequently occurring Unsatisfactory. precipitation weather situations and large-scale PC Some minutes Spatial aspect area (total area) Forecast Smallest subarea aspect Time period Forecast Shortest forecast interval Quality of precipitation forecast Computing resources required Computer/memory capacity Computing time Availability of forecast valuesAvailability After completion of calculations Suitable only for large-scale 56 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

Answers to questions concerning the area affected by precipitation, Intercomparison tests are desirable for NWP models the time of onset, the duration and depth of precipitation are all (Anthes, 1988). Within the European Working Group on Limited important. Attention should also be given to operational aspects Area Modelling (EWGLAM), forecast results from limited area such as the required computing resources and the time at which the models of several European Weather Services are already predicted precipitation data are available. exchanged each month (one forecast per month); the results include, In Table 5.7, the individual methods for quantitative inter alia, precipitation forecasts over two 12-hour intervals. precipitation forecasting are classified with the aid of hydrologically The results of the weather prediction models may be relevant characteristics, prototypes of which are described in refined by applying the dynamic-statistical model interpretation Subchapter 5.2. The characteristic spatial aspect should be (Perfect Prog, MOS) described in Subchapter 5.2.4. Local understood to mean the total area and the smallest subarea for which interpretation of the precipitation value at the grid point for a forecasts can be produced. The same applies, mutatis mutandis, to location or an area smaller than a grid square permits correction of the characteristic time aspect. Statements on the quality of the the systematic model error and indirect consideration of local forecast, on required computing resources and on the availability of factors having a permanent effect on the precipitation value to be results follow. It is left to the user from the hydrological field to predicted. The user thus obtains improved precipitation forecasts on select, with the aid of the survey, the method serving his purpose. average, but not for each individual case. A description of the international state-of-the-art of On the whole, the methods of statistical interpretation operationally-used methods for quantitative precipitation have proved efficient. They are widely applied to forecasts by forecasting is not possible at present. For this reason, a relevant models with mesh widths ∆s ≥ 50 km. There are also attempts to WMO inquiry is proposed. Main emphasis should be placed on develop operational MOS systems for models with even smaller dynamical models (weather prediction models). Specified mesh widths (e.g. Collins, 2000). questions on the hydrological cycle and on verification should Research and development are required for forecast make a qualitative estimation of the quantitative precipitation intervals over 2Ð12 hours, i.e. for the part of very-short range (up forecasting possible. to 12 hours) exceeding the range of nowcasting (up to 2 hours). In Regression and similarity procedures are no longer of this case, a practicable linking of nowcasting methods with the importance for Weather Services where operational NWP systems deterministic method of NWP models is required. In this are operated. The results obtained through these two procedures connection, it should be taken into account that the adjustment can also be obtained with the aid of NWP models. These models period (spin-up time) of the weather prediction models coincides describe the essential processes of thermodynamics and with that time period. microphysics of clouds relevant for precipitation in a Stochastic models are suited for small areas (20Ð500 km2) parameterized form and produce precipitation as an explicit and short forecast periods (1Ð3 hours). Based on measurement data forecast value. The models are being further developed with of a current precipitation event, such models produce the probability respect to higher resolution, improved modelling of physical on its further development with respect to time. A further processes and more accurate determination of the initial state. For development through a synoptic orientated typification of the initial moisture fields (specific humidity, cloud water content), precipitation event taking relevant meteorological parameters into observations based on remote-sensing systems will be increasingly account could lead to further improvements (Fleer, Franke and used in the future. Lecher, 1986). CHAPTER 6

WORLD WEATHER WATCH

6.1 DESCRIPTION subprogrammes aimed at studying essential fundamentals for the The World Weather Watch (WWW) is a global system for the improvement of weather forecasting. To support those Members collection, analysis, processing and dissemination of weather and who are unable to make their own contribution to the global system environmental information, and thus represents the very core of the as planned by WWW, the Voluntary Cooperation Programme (VCP) meteorological services. It was set up in 1963 (WMO, 1988d). The was established. operation of WWW is based on the fundamental concept that each of the 160 Members of WMO undertakes, according to its means, 6.1.1 Global Observing System (GOS) to meet certain responsibilities in the agreed global scheme, so that While progress in creating a larger density of the conventional all countries may benefit from the consolidated efforts. network of ground stations and in improving GTS connections is The WWW has three main components: relatively slow, a breakthrough in the bridging of spatial and (a) The Global Observing System (GOS) that uses all technical temporal gaps in the observation of rapid weather formations due facilities on land, at sea, in the air and in higher atmospheric to the development and use of new meteorological satellites layers for the observation and measurement of meteorologi- (Meteosat second generation — MSG) and automated observing cal elements; systems (Measuring Network 2000) can be seen. (b) The Global Telecommunication System (GTS), a worldwide Thus GOS today is considerably influenced by satellites data communication system with special circuits for the orbiting between the North and South Poles (800 to 1 000 km from rapid exchange of observation data, analysed and processed the Earth) collecting and transmitting global data on cloud covering, information including weather forecasts which are produced vertical temperature and humidity profiles, sea and land surface by the third main component; and temperatures, as well as snow and ice cover. The geostationary (c) The Global Data-processing System (GDPS), a network of satellites are positioned at a distance of approximately 36 000 km world and regional centres for computerized data-processing from the Earth. The present situation of the geostationary satellites including, for example, computer modelling for numerical changes constantly. Thus, on the position GOES/E there is at weather prediction, statistical analyses and graphic processing. present an older METEOSAT — INSAT has been replaced by a A survey of the activities of WWW from the point of view modified GOES and GOMS has not yet become a reality. Between of the National Meteorological Services (NMSs) is given in 75¡N and 75¡S, these satellites provide continuous images of cloud Figure 6.1. cover and serve as a telecommunication system for the collection The WWW, with its three main components, is active on and transmission of data (Figure 6.3). three levels- national, regional and global- according to the various scales of meteorological processes. The integrated system concept Global satellite update of the WWW with GOS, GTS and GDPS, the World Meteorological (a) Europe: Centres (WMCs), the Regional Specialized Meteorological Centres Meteosat-5 arrived at 63¡E, 36 000 km above the Indian (RSMCs) and the National Meteorological Centres (NMCs), as well Ocean on 19 May. Following successful imaging tests, it as the various fields of application, are shown in Figure 6.2. entered into service for the INDOEX experiment on 1 July. Also closely linked with WWW activities are education Meteosat-6 was relocated to a standby position at 10¡W after and training as well as technical and scientific research. One of the Meteosat-7 took over as the operational satellite at 0¡ on first and most comprehensive research programmes was the Global 3 June; Atmospheric Research Programme (GARP) with many (b) United States: GOES-9 (West) began suffering serious problems in its atti-

Upper-air tude control system in July as it neared the end of its planned Geostationary, Satellite ground Special aircraft Voluntary ship orbiting satellite station station lifetime. The in-orbit back-up satellite, GOES-10, replaced it Satellite meteoro- Buoys, moored, Automatic Manned Weather radar surface station surface station logical centre drifting User User

GOS GDPS

Numerical Area meteoro- Archives weather forecast logical watch Observing RTH RSMC Telecommunication station Centre User oriented Short-range Climatological special forecasts forecasts warnings data advisories GTS Observing station NMC WMC RSMC

Observing Severe weather Health, toursim station Agriculture Transport Power plants protection recreation sea land Environment Media Water resources air Construction protection User User

Figure 6.1 — The activities of WWW from the NMSs’ point of view. Figure 6.2 — Integrated concept of the WWW. 58 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

MTSAT-2 is scheduled for 2004; (f) India: INSAT-2B, the second generation satellite, positioned at 93.5¡E and INSAT-1D of the first generation at 83¡E, are the current operational satellites with INSAT-2A in standby mode at 74¡E. ) As already explained in Subchapter 2.2, at present, satellite-based measurements alone are not sufficiently accurate. Radar and conventional measurements have to be used for interpretation and computation. The GOS has thus developed into a composite system whereby not just one observational component or measuring technique is applied, but the whole set of data available is Figure 6.3 — The presently existing polar-orbiting and geostationary satellites (Image 9, 1998). required (Table 6.1).

on 28 July. Launched in 1994 and past its projected lifetime 6.1.2 Global Data-processing System (GDPS) of three years, GOES-8 (East) continues to function with no GOS already existed in a similar form prior to the establishment of significant changes in the past 28 months. GOES-11 was WWW, whereas the GDPS was created by WWW (Table 6.2). The launched in May 1999. NOAA-K, the first in a new series of introduction of supercomputers and highly sophisticated numerical five polar-orbiting satellites was launched on 23 May. Re- analysis and prediction models has shown that a distribution of tasks named NOAA-25 in orbit, it is now in a commissioning is essential. The system of three WMCs was designed to ensure the phase and will replace NOAA-12 at a future date. NOAA-14, processing of information on a global scale (Subchapter 2.1) for launched in December 1994, continues to function well; both real-time and non-real-time applications. These output (c) Russia: products are used by the RSMCs which, as shown in Table 6.2, The next Russian polar-orbiting meteorological satellites carry out either regional or problem-related functions. These include Meteor-3M-1 and 3M-2 were launched in 1999 and beyond prediction products on the mesoscale or large scale. Relevant 2000, respectively. The future GOMS Electro-geostationary information is given in WMO, 1993. meteorological satellite was launched in 2000; The GDPS is being continually extended so that the (d) China: number and quality of the output products increase from year to The next FY-1C polar orbiting meteorological satellite was year. The WMCs now produce almost 350 analyses and forecasts launched in 1999 and the next FY-2B geostationary meteoro- and the RSMCs make available over 2 000 products every day. The logical satellite in 2000; improvement in quality can be measured in terms of forecast (e) Japan: GMS-5, the current operational geostationary meteorological Table 6.2 satellite, will continue until the next generation, MTSAT. This Functions of various centres in the Global will be a dual-role (telecommunications and meteorology) Data-processing System (WMO, 1987) geostationary satellite. MTSAT-1 was launched in 1999 and Centre status Functions World Meteorological To prepare global/ hemisphere: Table 6.1 Centre ¥ analyses Global Observing System (WMO, 1987) ¥ short- and medium-range forecast Surface-based system Space-based system products supplied by: supplied by: ¥ climate diagnostic products Surface synoptic stations Polar-orbiting meteorological Regional Specialized To prepare: satellites Meteorological Centre ¥ regional fine-scale analyses Upper-air synoptic stations Geostationary meteorological with geographic ¥ regional fine-scale forecast products satellites specialization (48 hours) Observing stations in aircraft Satellites for environmental ¥ adjustment of global products to regional research conditions Aeronautical meteorological stations Research satellites resulting in: Regional Specialized To prepare specialized products, e.g. for: Research stations on vessels images and quantitative data Meteorological Centre ¥ long- and medium-range forecasting Special stations on vessels (including surface temperatures, with activity ¥ monsoon forecasting Climatological stations cloud upper limits, wind data specialization ¥ drought monitoring Agricultural meteorological stations from cloud displacement, total ¥ tropical cyclone advisory service Weather radar stations ozone concentrations, water ¥ specialized services to shipping and Radiation monitoring stations vapour contents, radiation aviation Stations in reconnaissance aircraft balances) with near-global ¥ regional climate diagnosis Stations with meteorological rockets coverage Ozone-measuring stations National Meteorological To prepare: Centre ¥ nowcasts and very short-range forecasts Background-pollution measuring ¥ short-, medium- and long-range forecasts stations for the public Planetary boundary-layer observing ¥ severe weather warnings stations ¥ national climate diagnostics Tide-monitoring stations ¥ other products as decided nationally CHAPTER 6 — WORLD WEATHER WATCH 59 accuracy. Today, the quality of weather forecasts in the northern the GTS: hemisphere for the next seven days is as good as it was in 1960 for (a) The transition from the X.25 transmission method that up the next three or four days. Without the GDPS, WMO Members until now used almost exclusively the procedure of message would not be able to maintain the variety, complexity and high switching systems in store and forward to methods based on standards of the meteorological services. IP (TCP sockets, FTP). File distribution technologies have thus become increasingly important; 6.1.3 Global Telecommunications System (GTS) (b) The changeover from the use of rented telephone and digital The GTS components linking the GOS via the Regional transmission lines to Managed Data Networks (MDN) Telecommunication Hubs (RTHs) with the WMCs, the RSMCs and whereby the transmissions maintenance is left to a large the NMCs must function properly if the observations and derived extent to the MDN server. In Regional Association VI products are to be collected, processed and disseminated to the (Europe), the MDN is about to commence operations under NMSs in forms suitable for application. At present, the GTS the name of Regional Meteorological Data Communication conveys over 15 million bytes of data and 2 000 weather charts Network. Other regions have started with preparations for daily. Transmission is via satellites, cable or VHF/UHF radio. the changeover. Hence, the transmission speeds have today The GTS itself is organized on three levels: become more a question of requirements and less of techni- (a) The Main Telecommunication Network (MTN) that consists cal possibilities and costs. of 24 point-to-point circuits linking the three WMCs (with RTH) and the 15 RTHs on the MTN (Figure 6.4); 6.1.4 Future developments (b) In addition, a further 14 RTHs (not on the MTN) supplied As WWW has to adapt itself to the progressive technical via regional connections; and possibilities in all sectors, this system will always be open to (c) The 149 NMCs within the scope of the RTHs that are linked improvements. Future developments will include: with the 32 RTHs via international connections. (a) A continual increase in transmission speeds, as required; Apart from the “official” GTS circuits, numerous (b) A gradual optimization of international data exchange by connections with the “on-bilateral agreement” status have been set improving the organization of the data networks, by chang- up where necessary, to transmit additional mainly national ing to modern technology and by guiding developing meteorological data (Resolution 40 (Cg-XII)). countries towards the level of the technology-developing The general guidelines of these systems are compiled in countries; the Manual on the Global Telecommunication System (c) A standardization and an improvement of the transmission (WMO-No. 386). Today, the normal data transmission speed for code, e.g. by increasing the transition to CREX, BUFR and MTN circuits is 64 Kbps, but, where necessary, the regional and GRIB codes instead of the traditional FM formats. national connections are also increasingly being operated at Apart from technical and organizational adjustments, the 64 Kbps. WWW will in this decade probably be influenced by changing Two important conversion processes are taking place in requirements and priorities. In this context, the following should be

Figure 6.4 — Circuits connecting the world and regional centres (RTH). 60 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES mentioned: the Hydrological Services do not need to set up their own (a) Safety of life and mitigation of natural disasters; transmission networks and can thus save costs. In some regions (e.g. (b) Agriculture and food production for an increasing population central Europe) where this was not possible up until now because under variable and, in many cases, unfavourable climatic the WWW network was used at maximum capacity for the conditions; transmission of meteorological data, the use of the network is now (c) Conservation of resources, particularly of water and energy, possible as a result of the introduction of new transmission and the management of natural resources; techniques with high transmission speeds. (d) Study and exploitation of marine resources (oil, gas, miner- Where free capacities in the WWW network still exist, als, fisheries); greater use of the WWW should be made by the Hydrological (e) Protection of the environment (atmosphere, inland waters, Services for the transmission of their data. In regions which do not oceans, soil and biota); yet have an extensive network for the transmission of hydrological (f) Transportation (land, sea and air); data at their disposal and where WWW capacities are not available, (g) Building and industry; an extension of WWW would be reasonable. (h) Health, recreation and tourism. For the exchange of hydrological data, WMO has developed a special code, HYDRA (WMO, 1995) which is adapted 6.2 DEMANDS OF HYDROLOGY AND WATER to the WMO code for the exchange of hydrological data. RESOURCES MANAGEMENT ON THE WWW Application of this code facilitates the exchange of hydrological As Chapter 1 shows, for most hydrological and water resources data and the taking over of meteorological data. In addition, there applications, non-real-time data are required. Such data can in is a code for hydrological forecasts (HYFOR) (WMO, 1995). This general be made available without difficulty to the hydrological code is designed to facilitate the exchange of hydrological forecasts, services by the Meteorological Services. The situation is different e.g. flood forecasts. in the case of real-time forecasts. Here, the data are needed immediately after measurement. 6.3 PILOT PROJECTS WWW can be used by hydrology and water resources In the mid-1970s a study was carried out to investigate to what management in two ways: extent WWW could be used for hydrological purposes. The Rhine (a) Use of the meteorological data exchange via the WWW; basin in Europe and the John River basin in North America were (b) Use of the WWW for the current transmission of hydro- selected for the purposes of this study. logical data. For the Rhine basin, the result was that meteorological Some Hydrological Services use meteorological data data were already being intensively used by the Hydrological exchanged via the WWW for their forecasts. Generally, the network Services, however, in a conventional way, e.g. on data carriers of the meteorological (synoptical) stations, whose data are (magnetic tapes) or by telex. An exchange of data via WWW was exchanged via the WWW, is not very dense, so there is considerable not possible because the WWW system was already overloaded. By inaccuracy in the areal values determined from the climatic data increasing the transmission speed and by setting priorities and non- transmitted. For this reason, many Hydrological Services have real-time data determination, the use of the WWW system would established additional automatic stations from which data can be today be possible. recalled. The situation in the John River basin is different. Unlike In many countries, the WWW is used for the current in the Rhine basin, the public telephone system is wide-meshed so transmission of hydrological data. For this purpose, measured data for this basin it appears quite reasonable to use WWW’s are fed into the WWW network and transmitted to the data centre telecommunication system. continuously but also in real time for forecasts. In such countries, CHAPTER 7

HYDROLOGY AND WATER RESOURCES PROGRAMME

7.1 HYDROLOGY AND WATER RESOURCES 7.2 THE HYDROLOGICAL OPERATIONAL PROGRAMME MULTIPURPOSE SYSTEM (HOMS) Within the framework of the United Nations, a number of HOMS is a sub-programme of the Programme on Basic Systems in international organizations deal with water, inter alia, WMO, the Hydrology. The basic document for HOMS is the HOMS Reference Regional Commissions, the United Nations Educational, Scientific Manual (HRM), which is available on the World Wide Web and Cultural Organization (UNESCO), the United Nations (http://www.wmo.ch/web/homs/homshome.html). Through the Environmental Programme (UNEP), the World Health Organization Manual, the Hydrological Services of the WMO Members offer and (WHO), the Food and Agriculture Organization of the United exchange interesting methods and computer programs which can be Nations (FAO), the United Nations Development Programme used for the implementation of daily operational tasks. Each method (UNDP) and the World Bank. The WMO Hydrology and Water or computer program, a so-called component, is described on two Resources Programme (HWRP) covers the operational aspects of standard pages following a certain pattern. Technically-related hydrological services. components are combined to form so-called sequences. The purpose of the programme is to promote activities The technology is usually in the form of descriptions of in operational hydrology and to further close cooperation hydrological instruments, technical manuals or computer programs, between Meteorological and Hydrological Services (paragraph material which has been made available for inclusion in HOMS by (e) of Article 2 of the Organization’s Convention). The overall the Hydrological Services of WMO Members from the techniques objective of the HWRP is to apply hydrology to meet the needs which they themselves use in their normal operations. This is an for sustainable development and use of water and related important aspect of the HOMS philosophy as it ensures that the resources, to mitigate water-related disasters and to encourage technology transferred is not only ready for use but also works effective environmental management at the national and reliably. international levels. HOMS is organized as a cooperative effort of WMO The current HWRP consists of five components: Members. Participating countries designate a HOMS National (a) Programme on Basic Systems in Hydrology (the Reference Centre (HNRC), usually in the National Hydrological Hydrological Information Referral Service, the Hydrological Service. This Centre provides national components for use in Operational Multipurpose System (HOMS) and the World HOMS, handles national requests for HOMS components to be Hydrological Cycle Observing System (WHYCOS)); supplied by other HNRCs, advises users on HOMS, and (b) Programme on Forecasting and Applications in Hydrology generally coordinates and publicizes HOMS activities in the (climate and water, flood and drought hazards); country. (c) Programme on Sustainable Development of Water Resources In 1999, a major effort to update the HRM was (participation of Hydrological Services in the national undertaken. The version now available online is the result of this planning); effort. The number of components has decreased significantly (d) Programme on Capacity Building in Hydrology and Water compared with the 1998 version. This decrease in quantity is, Resources (developing and strengthening human resources in however, compensated by a corresponding increase in the hydrology and management of water resources); usefulness of the individual components, as they have all been (e) Programme on Water-related Issues (cooperation with other updated quite recently. It must be noted that for the HRM to keep organizations). this high level of usefulness, it must be updated continually. The activities under the HWRP concentrate on the Therefore, HOMS users are encouraged to visit the HOMS pages measurement of basic hydrological elements from networks of frequently to keep abreast of developments. hydrological and meteorological stations; the collection, processing, storage, retrieval and publication of hydrological data, 7.3 WORLD HYDROLOGICAL CYCLE OBSERVING including data on the quantity and quality of both surface water SYSTEM and groundwater; the provision of such data and related 7.3.1 Introduction information for use in planning and managing water resources The World Hydrological Cycle Observing System (WHYCOS) is projects; and the installation and operation of hydrological WMO’s response to recommendations of the 1992 United Nations forecasting systems. Conference on Environment and Development (UNCED), which Hydrological problems are also dealt with in other WMO called on Governments to establish and maintain effective programmes such as the Tropical Cyclone Programme (TCP) and information and monitoring networks and promote further the the World Climate Programme (WCP). Support of the WCP is exchange of information regarding both surface water and mainly provided by the component programme on climate and groundwater, with respect to quantity, quality and uses, as well as water. ecosystems. It was initiated in 1993 by WMO with the goal of Two major sub-programmes of the HWRP are explained improving the knowledge and understanding of the hydrological below in more detail. cycle at the national, regional and global scale, and thus to support rational water resources management. 62 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

WHYCOS International Advisory Group (WIAG)

President of CHy (chairperson)

WMO Secretariat

Representatives of Regional Hydrological HYCOS Projects Advisers of the regions where HYCOS MED-HYCOS projects are being developed SADC-HYDCOS..... or implemented

CHy Advisory Working Regional Hydrological Grooup Advisers Member resposible for The World Bank WHYCOS European Union.....

Observers Regional GRDC grouping GEMS-Water e.g. ECA GCOS FRIEND IGAD, etc. IAHS ICID, etc.

Figure 7.1 — Coordination mechanism for WHYCOS.

WHYCOS is modelled on WMO’s WWW, which collects been achieved by the establishment of an in-house large amounts of data, mostly in real time, available globally, and WHYCOS Coordination Group (WCG) consisting of the has long existed in the field of meteorology. It offers a similar Deputy Secretary-General, the Assistant Secretary-General structure for hydrological purposes and is being designed and and the Directors of the Departments concerned; implemented in independent regional components called the (b) External coordination mechanism: To ensure world-wide Hydrological Cycle Observing System (HYCOS), each of them operational linkage among the various HYCOS components comprising of National Hydrological Services (NHSs) operating and to coordinate on all technical aspects of the programme. within well defined hydrological regions. WHYCOS does not This has been achieved by the establishment of the replace existing hydrological observing programmes, but WHYCOS International Advisory Group (WIAG) composed supplements them. And the National Services maintain their of the president of the Commission for Hydrology (CHy) and sovereignty over all data and information they share within representatives of the HYCOS regional centres, the CHy WHYCOS and participate on a voluntary basis. An important output Advisory Working Group, the external supporting agencies of WHYCOS is regional data sets that are of consistent quality and as well as the WMO Secretariat and the Regional can be used in preparing products for water resources assessment Hydrological Adviser concerned. and management. With WHYCOS as a vehicle, data from remote The WHYCOS priority lies in the regional needs of the sensing measurements in rivers will be transmitted via participating countries. The objectives of WHYCOS are: meteorological satellites to national and regional centres. Here the (a) To establish a global network of national hydrological data will be processed to assist in decision-making for planning observatories for the collection of high-quality data development and operation of projects such as retention measures, concerning the hydrological cycle and to transmit them in flood control, a decrease in the effects of drought, the combat of real time to national and regional data centres via the water pollution and regional water resources management. GTS; Furthermore, the information obtained could be used on research on (b) To strengthen the technical and institutional capacities of the the effects of El Niño and impacts of possible changes in climate National Hydrological Services by providing the relevant organization and objectives. hydrological information on water resources, trends, hazards, etc.; and 7.3.2 Organization and objectives (c) To promote the dissemination and use of such information, To ensure the successful implementation of the programme, its for example, via the World Wide Web. activities are coordinated by a special mechanism as follows (Figure 7.1): 7.3.3 WHYCOS stations (a) WMO internal coordination mechanism: to link the inputs of The stations are selected by the participating countries. Focus is the various Departments of the WMO Secretariat. This has being placed on existing stations that can be upgraded to satisfy the CHAPTER 7 — HYDROLOGY AND WATER RESOURCES PROGRAMME 63

Figure 7.2 — HYCOS regional components around the world. requirements of both national priority objectives and those of 7.3.4 Regional components WHYCOS. The criteria for station selection are: WHYCOS has been conceived as a vehicle for technology (a) Availability of long time series; transfer, training and capacity building. Its regional components (b) A stable water level-discharge relationship; are being developed and implemented in various parts of the (c) Regional significance of the data. world with financial support provided by the World Bank, the The minimum set of basic data recorded consists of: European Commission and the Government of France. At (a) Water level/flow; present, three projects are under implementation covering the (b) Precipitation; Mediterranean (MED-HYCOS), Southern Africa (SADC- (c) Temperature; HYCOS) and west and central Africa (AOC-HYCOS). In (d) Humidity. addition, there are 12 other projects at different stages of Other variables to estimate the potential development in such areas as East Africa, Congo, Danube, evapotranspiration as well as selected physical and chemical Amazon and La Plata river basins, the Aras Sea, Baltic Sea, parameters are required. Provided that countries agree to the Black Sea and Caribbean Sea basins, the Pacific Islands and the exchange of data, the use of satellites under the umbrella of WMO Himalayan and Artic region (Figure 7.2). will be free of charge. REFERENCES

Aguado, E. (1983): An energy-budget model of snowmelt in Becker, A. (1992): Methodische Aspekte der Regionalisierung. southern and east-central Wisconsin. Ph.D. thesis, University DFG-Mitt. XI der Senatskommission für Wasserforschung, of Wisconsin, Madison. pp. 16-32. Allen, R.G., M. Smith, A. Perrier u. L.S. Pereira (1994): An Becker, F., H.-J. Bolle and P.R. Rowntree (1988): The update for definition of reference evapotranspiration. ICID International Satellite Land-Surface Climatology Project BULLETIN 43, No. 2, pp. 1-34. (ISLSCP). ISLSCP-Report 10, FU Berlin, 100 pp. Altenhofen, J. (1985): A modified atmometer for on-farm Bellocq (1980): Operational models of quantitative precipitation ET-determination. ASAE Publication 14Ð85, pp. 117Ð184. forecasts for hydrological purposes and possibilities of an Amorocho, J. and B. Espildora (1966): Mathematical simulation intercomparison. WMO, Annex, R/NWP/5. of the snow melting processes. Wat. Sci. and Engng. Paper Benesch, B. (1989): Gro§ma§stabige digitale Klimakarten. No. 3001, University of California, Davis CA. Statusbericht des BMFT Projektes Landoberflächen- Anderson, E.A. (1973): National weather service river forecast klimatologie; Deutscher Wetterdienst, Offenbach. system — Snow accumulation and ablation model. U.S. Blaney, H.F. u. W.D. Criddle (1950): Determining water require- Dept. of Commerce, NOAA Tech. Mem. NWS-HYDRO-17, ments in irrigated areas from climatological and irrigation Silver Spring MD. date. US Dep. Abr. Div. Irrigation and Water Conservation, Anderson, E.A. (1976): A point energy and mass balance model of SCS TP Nr. 96, 48 pp.. a snow cover. U.S. Dept. of Commerce, NOAA Tech. Rep. Blöschl, G. (1990): Snowmelt simulation in rugged terrain Ð The NWS 19, Silver Spring MD. gap between point and catchment scale approaches. Wiener Anthes R.A. (1988): Computational and physical factors affecting Mitteilungen. Wasser-Abwasser-Gewässer Nr. 91. forecast skill in regional numerical models. In: Workshop on BMU (1997): Stand und Einsatz mathematisch-numerischer limited-area modelling intercomparision. Boulder, Colorado, Modelle in der Wasserwirtschaft (Status and application of pp. 1Ð16. mathematical-numerical models for water management). Austin, B.N., I.D. Cluckie, C.G. Collier and P.J. Hardaker (1995): Bundesministerium für Umwelt, Naturschutz und Radar-based estimation of probable maximum precipitation Reaktorsicherheit, Bonn, 63 pp. and flood. Final Report prepared for the Water Directorate, BMU (2000) Bundesministerium für Umwelt, Naturschutz und DOE, by the Meteorological Office and the Dept. of Civil Reaktorsicherheit (2000): Hydrologischer Atlas Engineering, University of Salford, UK Met. Office, Deutschlands, Berlin. Bracknell. Braden, H. (1985): Ein Energiehaushalts- und Verdunstungs- Balzer (1998): Verifikationsbericht zur Güte lokaler modell für Wasser- und Stoffhaushaltsuntersuchungen Wettervorhersagen (Verification report on the quality of local landwirtschaftlich genutzter Einzugsgebiete. Mitt. Dt. weather forecasts), No. 17. Offenbach, 40 pp. Bodenkdl. Gesell. 42, pp. 294-299. Balzer (2000): Verifikationsbericht zur Güte lokaler Braun , L.N. (1985): Simulation of snowmelt-runoff in lowland Wettervorhersagen (Verification report on the quality of local and lower alpine regions of Switzerland. Zürcher Geogr. weather forecasts), No. 21. Offenbach, 59 pp. Schr. 21, Zürich, 166 pp.. Balzer, K. and P. Emmrich (1997): Verification and inter- Braun, L.N. and Rohrer, M.B. (1992): Schneedeckenverteilung comparison 1996. In: Verification of ECMWF products in im Mikro- bis Meso-skaligen Bereich. In: Member States and Co-operating states Ð Report 1997. Regionalisierung in der Hydrologie, DFG-Mitt. XI der Available from ECMWF, Shinfield Park, Reading RG2 9AX, Senatskommission für Wasserforschung, VCH-Verlag, UK, pp. 40-49. Weinheim, pp. 185-199. Barret, E.C., C. Kidd and J.O. Bailey (1988): The Special Browning, K.A. (1979): The FRONTIERS plan: a strategy for using Microwave Sensor Images (SMMI): a new instrument with radar and satellite imagery for very-short-range precipitation rainfall monitoring potential. Int. J. Remote Sensing. forecasting. Met. Mag. 108, pp. 161Ð184. Bartels, H., G. Malitz, S. Asmus, F.M. Albrecht, B. Dietzer, Brunet, N., R. Verret and N. Yacowar (1988): An objective T. Günther and H. Ertel (1997): Starkniederschlagshöhen comparison of model output statistics and “perfect pro” für Deutschland. KOSTRA-Atlas, DWD, Offenbach am systems in producing numerical weather element forecasts. Main. Weather and Forecasting 3, pp. 273Ð283. Bartels, H. (1992): Regionalisierung am Beispiel der flächendeck- Brutsaert, W. (1982): Evaporation into the atmosphere. Reidel enden Starkniederschlagsauswertung für die Bundesrepublik Publishing Company, Dordrecht, 299 p. Deutschland. In: Regionalisierung in der Hydrologie, DFG, Budyko, M.I. (1974): Climate and life. Von D.H. Miller Mitteilung XI der Senatskommission für Wasserforschung, herausgegebene englische Übersetzung, Academic press New Weinheim. YorkÐSan FransiscoÐLondon, 508 pp. Bartels, H., F.M. Albrecht and J. Guttenberger (1990): Bultot, F. and G.L. Dupriez (1985): Casebook on operational Starkniederschlagshöhen fur die Bundesrepublik Deutschland assessment of areal evaporation from a river basin. OHR 22, (Teil 1). DWD, Offenbach am Main. (WMO-No. 635), Geneva, pp. 80Ð105. REFERENCES 65

Bundesanstalt für Gewässerkunde (1987): Jahresbericht 1986. forecasts at the German Weather Service. Submitted to II/15ÐII/16 resp. II/12ÐII/13, Koblenz. Journal of Hydrology. Bundesministerium für Forschung und Technologie (1989): de Quervain, M. (1973): Snow structure, heat and mass flux Klimaforschungsprogramm. Konferenzbericht, heraus- through snow. IAHS Publ. 107, Vol. 1, pp. 203-226. gegeben von der GSF, Neuherberg, 543 pp. Deisenhofer, H.E., H. Kumm and H. Wollkopf (1982): CEC (1988): Status Reports on National Activities: COST-73 Berechnung von monatlichen Gebietsniederschlägen mit dem Weather Radar Networking, Doc. EURO-COST 73129188, Sammelgebietsverfahren am Beispiel des Iller- Commission of the EC COST Cooperation. Einzugsgebietes. Deutscher Wetterdienst, Offenbach, 40 pp. Chandrasekar, V., V.N. Bringi, N. Balakrishnan, and D. Zmic Denoth, A., W. Seidenbusch, M. Blumthaler, P. Kirchlechner and (1990): Error structure of multiparameter radar and surface W. Ambach (1979): Study of water drainage from columns of measurements of rainfall. Part III: Specific differential phase. snow. Proc. Hanover Workshop, CRREL Rep. 79-1, J. Atmos. and Oceanic. Tech 7, pp. 621-629. Hanover NH. Choisnel., E.; E. de Villele u. F. Lacroze (1990): Une approche Deutsche Forschungsgemeinschaft (1979): Hydrologischer Atlas uniformisée du calcul de l’évapotranspiration potentielle pour der Bundesrepublik Deutschland. Boldt Verlag Boppard, l’ensemble des pays de la Communauté Européene. Rappport 365 pp. final, Direction de la Métérologie Nationale Paris, 69 pp. and Deutsche Forschungsgemeinschaft (1987): Methoden der Annexes. Fernerkundung in Hydrologie und Wasserwirtschaft. Colbeck, S.C. (1972): A theory of water percolation through snow. Arbeitspapier der DFG ad-hoc Arbeitsgruppe Fernerkundung J. Glaciol. 11, pp. 369-385. in Hydrologie und Wasserwirtschaft, im Auftrage der DFG- Colbeck, S.C. and M. Ray (Eds.) (1978): Modeling of snow cover Senatskommission für Wasserforschung, Selbstverlag der runoff. Proc. of Hanover Workshop, CRREL Rep. 79-1, DFG, 56 pp. Hanover NH. Deutscher Wetterdienst (1979): Beschreibung aller Basis- und Collier, C.G. (1991): NIMROD Ð A system for nowcasting and Sammelgebiete mit Gebietsniederschlagsberechnungen. initialization for modelling using regional observational Offenbach. data. Preprints, 25th Int. Conf. on Radar Met., 24-28 June, Deutscher Wetterdienst (1986): Hydrometeorologie. Promet 2/3, Paris, France, American Meteorological Society, Boston, Offenbach. pp. 21-24. Deutscher Wetterdienst (1990 und 1991): Weltzentrum für Collier, C.G. (1996): Applications of weather radar systems. John Niederschlagsklimatologie (WZN). FE-Antrag zum BMFT- Wiley and Sons, Chichester, New York, Brisbane, Rexdale, Vorhaben WZN und Vortrag zum Workshop Iphofen bzw. Singapore. Workshop Neustadt. Collier, C.G. and P.J. Hardaker (1996): Estimating probable Deutscher Wetterdienst (1999): Quarterly Report of the maximum precipitation using a storm model approach. To Operational NWP Models of the Deutscher Wetterdienst, appear in Journal of Hydrology. No. 21, 1 September until 30 November 1999. Offenbach, Collinge, V. K. and C. Kirby (1987): Weather radar and flood fore- 73 pp. casting. John Wiley and Sons, Chichester, New York, Diekmann, F.-J., S. Happe, M. Rieland, W. Benesch, G. Czeplak Brisbane, Toronto, Singapore. and F. Kasten (1988): An operational estimate of global solar Collins, C.T. (2000): Developing Local Model Output Statistics irradiance at ground level from METEOSAT data: results (MOS) Equations to predict Temperature and Probability of from 1985 to 1987. Meteorol. Rundschau 41, pp. 65Ð79. Precipitation. Preprints of the 15th Conference on Probability DIN 4049, (1994): Teil 1, Teil 3: Hydrologie, Beuth Verlag and Statistics in the Atmospheric Sciences, 8-11 May 2000, GmbH, Berlin und Köln. Asheville, North Carolina. pp. 30-33. Diskin, M.H. (1969): Thiessen coefficients by a Monte Carlo Conway, B.J. (1987): FRONTIERS an operational system for procedure. Journal of Hydrology 8, pp. 323Ð335. nowcasting precipitation. Proc. Symp. Mesoscale Analysis Disse, M. (1995): Modellierung der Verdunstung und der and Forecasting, Vancouver, Canada, 17Ð19 August, ESA Grundwasserneubildung in ebenen Einzugsgebieten. IHW SP-282, pp. 233Ð238. Univers. Karlsruhe (TH), H. 53, 180 pp. Courvoisier, H.W. (1970): Die quantitative Niederschlagsprognose Doorenbros, J., W.O. Pruitt (1977): Guidelines for predicting crop winterlicher zyklonaler Witterungslagen auf der water requirements. FAO Irrigation and Drainage Papers No. 24. Alpennordseite der Schweiz. Veröff. Schweizer. Met. DVWK (1980): Empfehlungen zum Bau und Betrieb von Lysimetern. Zentralanst. 15, 14 pp. DVWK-Regeln zur Wasserwirtschaft, Heft 114, Bonn. Courvoisier, H.W. (1980): Quantitative Niederschlagsvorhersagen. DVWK (1985): Niederschlag-Starkregenauswertung nach Schriftenr. d. Deutschen Verbandes f. Wasserwirtsch. u. Wiederkehrzeit und Dauer. DVWK-Regeln zur Kulturbau (DVWK) 51, pp. 67Ð81. Wasserwirtschaft 124, Parey-Verlag, Hamb. Cress, A. (1997): Bereitstellung von Niederschlagsvorhersagen DVWK (1990): Stand der Verdunstungsermittlung in der mit dem Deutschland-Modell des DWD für historische Bundesrepubilk Deutschland. DVWK-Schriften, Heft 86, Hochwasserereignisse (NIEVO). Abschlu§bericht für den pp. 1-77, Bonn. Zeitraum vom 01.07.1994 bis 31.12.1996. Deutscher DVWK (1996): Ermittlung der Verdunstung von Land- und Wetterdienst, Geschäftsbereich Forschung und Entwicklung, Wasserflächen, DVWK-Merkblätter 238, 135 pp. 63 pp. DVWK (1997): Maximierte Gebietsniederschlagshöhen für Damrath, U., G. Doms, D. Frühwald, E. Heise, B. Richter and Deutschland. Mitteilungen des Deutschen Verbandes für J. Steppeler (2000): Operational quantitative precipitation Wasserwirtschaft und Kulturbau, Heft 29. 66 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

Dyck, S. and G. Peschke (1983): Grundlagen der Hydrologie. Niederschlagsganglinien. Beitrage zur Meteorologie der Berlin, 388 pp. Schweiz - Hydrologie, pp. 33. Dyer, A.J. and F.J. Maher (1965): The Evapotron: An instrument Giesecke, J. and H. Meyer (1984): Das räumlich-zeitlich variable for the measurement of eddy fluxes in the lower atmosphere. Niederschlagsangebot, ermittelt aus Bodenmessnetzen; Division of Meteor. Physics Techn. Paper No. 15, Melbourne, Niederschlagsvorhersage für Hydrologie und 31 pp. Wasserwirtschaft. Mitteilung der Senatskommission für Eimern, J. van and H. Hackel (1979): Wetter-und Klimakunde. Wasserforschung der DFG, VCH-Verlagsgesellschaft Stuttgart, 269 pp.. Weinheim. Emmrich, P. (1987): Aktuelle Verifikationsergebnisse der BKF- Glahn, H.P. and D.A. Lowry (1972): The use of model output Niederschlagsvorhersagen. In: Möglichkeiten der statistics (MOS) in objective weather forecasting. J. Appl. Niederschlagsvorhersage für Hydrologie und Meteor. 11, pp. 1203Ð1211. Wasserwirtschaft. Mitteilung VIII der Senatskommission für Glugla, G. et al. (1999): Entwicklung von Verfahren zur Wasserforschung der DFG, VCH-Verlagsgesellschaft, Berechnung langjähriger Mittelwerte der flächendifferen- Weinheim. zierten Abflu§bildung. DFG-Projekt GI 242/1-2, Ernstberger, H. (1987): Einflu§ der Landnutzung auf Verdunstung Bundesanstalt für Gewässerkunde, 16 pp. and annexes. und Wasserbilanz. Beitrage zur Hydrologie, 189 pp. Golding, B.W. (1995): Rainfall-analysis and short range predic- ESA (1981): Introduction to the METEOSAT System. SP-1041, tion. Paper 1 in Hydrological Uses of Weather Radar, British Paris. Hydrological Society, Occasional Paper No.5, ed. EUMETSAT (1988): 7th METEOSAT scientific users’ meeting. K.A. Telford, pp. 1-9. EUM P 04, Madrid, 27Ð30 September 1988. Golf, W. und K. Luckner (1991): AKWA Ð ein Modell zur Fell, E. and D. Prellberg (1999): Hochwassermeldedienst in Berechnung aktueller Wasserhaushaltsbilanzen kleiner Rheinland-Pfalz zur Hochwasservorsorge (Flood Warning Einzugsgebiete im Mittelgebirge, Acta hydrophysica Berlin Service in Rhineland-Palatinate in Support of Flood 35, Heft 1, pp. 5-20. Protection Measures). Wasserwirtschaft 89, pp. 64-69. Göppert H., J. Ihringer, E.J. Plate, and G. Morgenschweis (1998): Fleer, H., W. Franke, and K. Lecher (1986): Möglichkeiten der Flood forecast model for improved reservoir management in Niederschlagsvorhersage für hydrologische Zwocke. Dt. the Lenne River catchment, Germany. Hydrological Sciences- IHP/OHP-Nationalkomitee, IHP/OHPBerichte 7, Koblenz, Journal 43, pp. 215-242. 73 pp.. Grebner, D. (1982): Objektive, quantitative Niederschlags- Foken, T. (1990): Turbulenter Energieaustausch zwischen vorhersage im Rheingebiet. Ber. Nr. I-1 der Internat. Komm. Atmosphäre und Unterlage. Ber. Dt. Wetterdienst 180, f. d. Hydrol. d. Rheingebiets (1CHR), s’Gravenhage. Offenbach/M. 287 pp. Günther, T., H. Matthäus, M. Rachner and F.M. Albrecht (2000): Fortak, H. (1971): Meteorologie. Carl Habel Verlagsbuchhandlung Regionalisierte Extremwerte des Niederschlagsdargebots aus Berlin und Darmstadt. Regen und Schneeschmelze für Deutschland. REWANUS- Franke, W. (1987): Meteorologisch-statistische Verfahren zur Atlas, DWD, Berlin (in print). verbesserten Verwendung von BKF-Niedersch- Gutknecht, D. (1991): On the development of ãapplicable“ models lagsvorhersagen in Hochwasservorhersagemodellen für die for flood forecasting. IAHS-Publ. No. 201, pp. 337-345. Leine. In: Möglichkeiten der Niederschlagsvorhersage für Haddeland, J. (1996): Flood Forecasting in Norway. XVIII. Hydrologie und Wasserwirtschaft. Mitteilung VIII der Konferenz der Donauländer über hydrol. Vorhersagen und Senatskommission für Wasserforschung der DFG, VCH- hydrologisch-wasserwirtschaftl. Grundlagen. Schriftenreihe Verlagsgesellschaft, Weinheim, pp. 131Ð141. zur Wasserwirtschaft TU Graz, Band 19/1, pp. B 199-204. Friend, A.D. (1998): Parameterisation of a global daily weather Hagen, G. (1841): Handbuch der Wasserbaukunst. 1. Teil: Die generator for terrestrial ecosystem modelling. Ecological Quellen. Modelling 109, pp.121-140. Haiden, T., P. Kahlig, M. Kerschbaum and F. Nobilis (1991): Frühwald, D. (1998): Quantitative precipitation forecasts Ð current Einflu§ der Orographie auf Extremniederschläge: Ein deter- status and future developments at DWD. Proceedings of the ministischer Modellansatz. Mitteilungsblatt des Workshop on Flood Forecasting, Prague, Czech Hydrographischen Dienstes in Österreich, Nr. 64, Wien, 1-27. Hydrometeorological Institute, 4-5 May 1998, Halldin, S. (1988): Swedish Evaporation Research — A Review. pp. 77-85. Nordic Hydrology 19, pp. 303-340. Frühwald, D. (1999): Quantitative Niederschlagsvorhersage mit Haltiner G.J. and R.T. Williams (1980): Numerical prediction and numerischen Modellen Ð Gegenwärtiger Stand und zukün- dynamic Meteorology. John Wiley and Sons, New York. ftige Entwicklung im Deutschen Wetterdienst. Hamilton, J., P. Lemon and B. O’Donnell (1987): Objective IHP/OHP-Berichte Heft 13, Koblenz. pp. 193-204. Analysis of Climatological Fields of Temperature, Gaudet, B. and W.R. Cotton (1998) : Statistical Characteristics of Sunshine and Rainfall. Met. Service of Ireland. Techn. Note a Real-Time Precipitation Forecasting Model. Weather and No. 49. Forecasting 13, pp. 966-982. Gee, G. (1986): Four case Harrison, M.S.J., D.S. Richardson, K. Robertson, and A. studies illustrating some deficiencies of the bright band Woodcock, 1995: Medium-range ensembles using both the correction procedure. Nowcasting Research Group Report ECMSF T63 and unified models Ð An initial report. UKMO No.6, Meteorological Office, Bracknell, Berkshire. Tech. Rep. 153, 25pp [Available form United Kingdom Geiger, H, A. Stehli and K. Castellazzi (1986): Regionalisierung Meteorological Office, London Rd., Bracknell, Berkshire der Starkniederschläge und Ermittlung typischer RG12 2S2 United Kingdom]. REFERENCES 67

Haude, W. (1952): Zur Möglichkeit nachträglicher Bestimmung Israelsen, E.K. and J.P. Riley (1972): Distribution of precipitation der Wasserbeanspruchung durch die Luft und ihrer in mountainous areas. Geilo symposium, Norway, 31 JulyÐ Nachprüfung an Hand von Topfversuchen und 5 August 1972 (WMO-No. 326) pp. 552-564. Abflu§messungen. Berichte des DWD 32 (US-Zone), Itier, B., P. Cellier and C. Riou (1985): Actual evapotranspiration pp. 27-34. measurement by a simplified aerodynamic system. IAHS Haude, W. (1958): Uber die Verwendung verschiedener Publ. 14, pp. 201Ð207. Klimafaktoren zur Berechnung der potentiellen Evaporation Jackson, R.D., R.J. Reginato and S.B. IDSO (1977): Wheat und Evapotranspiration. Meteorologische Rundschau 11/3, canopy temperature: a practical tool for evaluating water pp. 96-99. requirements. Water Resources Research, 13/3, pp. 651-656. Haugk, E. (1983): Beiträge zur Wahl des Bemessungshochwassers Jacobsen, I. and E. Heise (1982): A new economic method for the und zum vermutlich größten Niederschlag, DVWK-Schriften computation of the surface temperature in numerical models. 62, Hamburg und Berlin. Beitr. Phys. Atm. 55, pp. 128-141. Hellmann, G., V. Elsner, G. Henze and K. Knoch (1921): Klima- Jaworski, J. (1985): Casebook on operational assessment of areal Atlas von Deutschland. Berlin, Reimer. evaporation. OHR 22 (WMO-No.635), pp. 106-120. Herrmann , A., J. Martinec and W. Stichler (1979): Study of Jensen, H. (1986): Regionalisierung der Verteilungsfunktion des snowmelt runoff components using isotope measurements. jahrlichen Maximums des Tagesniederschlags im Kanton Proc. Hanover Workshop 1978, CRREL. Hanover NH, pp. Zürich. Zürcher Geogr. Schriften 27, ETH Zürich. 288Ð296. Jensen, H. (1989): Räumliche Interpolation der Stundenwerte von Herrmann, A. (1986): Schneehydrologie-Aufgaben, Stand und Niederschlag, Temperatur und Schneehöhe. Zürcher Notwendigkeiten unter besonderer Berücksichtigung von Geographische Schriften 35, 70 pp, ETH Zürich. Mitteleuropa. Promet 2/3, pp. 23-30, Offenbach. Joe, P. (1996): Precipitation at the ground: radar techniques. In: Hershfield, D.M. (1965): Method for estimating probable Radiation and Water in the Climate System, Springer Berlin, maximum rainfall. Journal American Waterworks Ass., Vol. Heidelberg, New York, Barcelona, Budapest, Hongkong, 57, pp. 965-972. London, Milan, Paris, Santa Clara, Singapore and Tokyo, Hofmann, G. (1955): Die Thermodynamik der Taubildung. pp. 277-321. Berichte des DWD 18. Jones, C.D. and B. Macpherson (1997): A latent heat nudging Holgate, H.T.D (1973): Rainfall forecasting for river authorities. scheme for the assimilation of precipitation data into an oper- Met. Mag. 102, pp. 33-48. ational mesoscale model. Meteorol. Appl. 4, pp. 269-277. Holle, F.-K. (1987): Einbindung der Niederschlagsvorhersage in Kaestner, A. (1979): Numerische Analyse. Fortbildung- die Hochwasserabflussvorhersage (Integration of sveranstaltung über numerische Wettervorhersage der Precipitation Forcasting into Flood Forecasting). Deutschen Meteorologischen Gesellschaft, pp. 37Ð74. Institut für Wasserwirtschaft, Hydrologie und land- Kalweit, H. (1953): Der Wasserhaushalt, Verlag Technik, Berlin. wirtschaftlichen Wasserbau der Universität Hannover, Final Kästner, W. (1999): Personal communication, Bavarian State Report, 98 pp. Ofice for Water Resource Management, München. Homagk, P. (1996): Hochwasserwarnsystem am Beispiel Baden- Kirnbauer, R. (1993): Modellkonzeptionen zur operationellen Württemberg (Flood Warning System in Kurzfristvorhersage der Schneeschmelze. Schneehydrologie Baden-Württemberg). Geowissenschaften 14, pp. 539-546. Ð Modellierung der Schneeschmelze in Einzugsge-bieten. Homagk, P. (1999): Personal communication. Environmental Schriftenreihe der Forschungsinitiative des Verbundkonzerns. Agency of Baden-Württemberg, Karlsruhe. Band 13, pp. 87-110, Wien. Homagk, P. and W. Schulz (1999): Personal communication, Kirnbauer, R., G. Blöschl and D. Gutknecht (1994): Entering the Environmental Agency of Baden-Württemberg, Karlsruhe. Era of Distributes Snow Models. Nordic Hydrology 25, Horton, R.E. (1923): Accuracy of areal rainfall estimates. Monthly pp. 1-24. Weather Revue 51, pp. 348-353. Kirchhofer, W., A. Ohmura and H. Wanner (1984): Applied Houtekamer, P., L. Lefaivre, J. Derone, H. Ritchie, and H. Climatology. Zürcher Geogr. Schriften 14. Mitchell (1996): A system simulation approach to ensemble Kitchen, M., R. Brown and A.G. Davies (1994): Real-time correc- prediction. Mon. Wea. Rev. 124, pp. 1225-1242. tion of weather radar data for the effects of bright band, range Hoyningen-Heune, J. von (1983a): Forschergruppe Wasser- und and orographic growth in widespread precipitation. Quart. J. Stoffhaushalt landwirtschaftlich genutzter Einzugsgebiete R. Met. Soc. 120, pp. 1231-1254. unter besonderer Berücksichtigung von Substrataufbau, Klämt, A. (1988): Konzipierung eines nutzerorientierten Modells Relief und Nutzungsform, Teilvorhaben Gebietsverdunstung. zur Berechnung aktueller Monatssummen der Selbstverlag des Lehrstuhls für Physische Geographie und Gebietsverdunstung unter Berücksichtigung der Art der Landschaftsökologie der TU Braunschweig, pp. 110-170. Landnutzung. Acta hydophysica Berlin 32, pp. 237-250. Hoyningen-Heune, J. von (1983b): Die Interzeption des Klatt, R. (1983): Vorhersage von Hochwasser aus Niederschlags in landwirtschaftlichen Pflanzenbeständen. radargemessenem und prognostiziertem Niederschlag. Schriftenreibe des DVWK 57, pp. 1-53. Schriftenr. Hydrologie und Wasserwirtschaft der Ruhr-Univ. Hoyningen-Heune, J. von, H. Braden and F. -J. Löpmeier ( 1986): Bochum. Methoden zur Bestimmung der Verdunstung. Promet 2/3, Kleeberg, H.-B. (1992): Methodische Aspekte der pp. 14-20, Offenbach. Regionalisierung. In: Regionalisierung in der Hydrologie. ICID (1971): Fachwörterbuch für Bewässerung und DFG-Mitt. XI der Senats-kommission für Wasserforschung, Entwässerung. Verlagsbuchhandlung Stuttgart, 948 pp. VCH-Verlag, Weinheim. 68 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

Klein, W.H., B.M. Lewis and I. Enger (1959): Objective predic- Malitz, G. and T. Günther (2001): Vermutlich größte tion of five-day mean temperature during winter. J. Meteor. Gebietsniederschlagshöhen in den Flussgebieten 16, pp. 672-682. Deutschlands. Korrespondenz Abwasser (in print). Knauf, D. (1980): Die Berechnung des Abflusses aus einer Martinec, J. (1960): The degree-day factor for snowmelt-runoff Schneedecke. DVWK Schriften 46, pp. 95-135. forecasting. IAHS Publ. 51, pp. 468-477. Knüpffer, K. (1996): Methodical and Predictability Aspects of Matjeka, T.J., R.A. Houze and P. Hobbs (1981): Microphysics and MOS Systems. Preprint Volume, 13th AMS Conference on dynamics of clouds associated with mesoscale cyclones. Probability and Statistics in the Atmospheric Sciences, 21-23 Quart. J. Royal Meteorol. Soc. 106, pp. 29-56. Feb. 1996, San Francisco CA, pp. 19-23. Mauser, W. and M. Wegenhenkel (1989): Statusbericht zum Knüpffer, K. (1998): Statistical Weather Forecasting as a Link Landoberflächenklimaprogramm über das Between NWP, Synoptics, and the Customer, Preprint Forschungsvorhaben: Modellierung von Wasserhaushalt- Volume, 16th AMS Conference on Weather Analysis and skomponenten mit Hilfe von Satellitendaten im Forecasting, 11-16 Jan. 1998, Phoenix AZ. Hydrologischen Versuchsgebiet Ostkaiserstuhl. Krahe, P. (1998): Water-level and discharge forecasts in River Klimaforschungsprogramm (BMFT, GSF), pp. 195-198. Moselle basin on the basis of measured and forecasted mete- Meetschen, D. (1999): Erkennung, Nutzung, und Entfernung von orological data. Proceedings of the Workshop on Flood Clutter zur Verbesserung der Niederschlagsmessung mit dem Forecasting, 4-5 May, 1998, Prague, Czech Hydrometeo- Bonner Radar. Diplomarbeit, Meteorologisches Institut der rological Institute, pp. 108-118. Universität Bonn. Kremser, U. and Reich, T.(1991): Die Bestimmung der Meinardus, W. (1900): Eine einfache Methode zur Berechnung Gebietsniederschlagshöhen mit der Methode der optimalen von klimatologischen Mittelwerten von Flächen. Interpolation am Beispiel des Selke-Einzugsgebietes, Z. Meteorologische Zeitschrift 6, pp. 241-275. Meteorologie 41 2, S. 122-129. Mellor, G.L. and T. Yamada (1974): A hierarchy of turbulence Krüger, L.R. (1983): Ermittlung konvektiver Niederschläge mittels closure models for planetary boundary layers. J. Atmos. Sci. METEOSAT Infrarotdaten. Meteorologische Rundschau 36, 31, pp. 1791-1806. pp. 50-58.. Mendel, H.G. (1977): Die Berechnung von Lattermann, A. (1983): Die zeitliche Niederschlagsschätzung Gebietsniederschlägen. DGM 21/6, pp. 129-141. mittels stochastischer Prozesse. Mit. Inst. f. Wasserwirtsch., Mendel, H.G. (1979): Zur Berechnung von Gebietsniederschlägen Hydrol. u. Iandwirtschaftl. Wasserbau d. Univ. Hannover 53, aus Punktmessungen. Mitteilung der Versuchsanstalt für pp. 9-133. Wasserbau, Hydrologie und Glaziologie der ETH Zürich 41, Lettenmaier, D.P. (1995): Stochastic modeling of precipitation pp. 187-214. with applications to climate model downscaling. In H. von Moch, H. (1989): Mündliche Mitteilungen in Offenbach anläßlich Storch and A. Navarra, editors, Analysis of Climate eines Abstimmungsgespräches mit dem Projekt REKLIP. Variability: Applications of Statistical Techniques, pp. 197- Molteni, F., R. Buizza, T.N. Palmer and T. Petroliagis, 1996: The 212. Springer Verlag. ECMWF ensemble prediction system: methodology and vali- LfU BW/DWD (1975): Relative Klimatische Verteilungskurven dation. Quart. J. Roy. Meteor. Soc. 122, pp. 37-120. der Niederschläge in Baden Württemberg. Karlsruhe. Monteith, J.L. (1965): Evaporation and environment. Symposium Löpmeier, F.-J. (1987): Verdunstung und Energiehaushalt. Verh. d. of the Society for Experimental Biology 19, Cambridge Deutschen Geographentages 45, pp. 434-438. (University Press), pp. 205-235. Löpmeier, F.-J. (1991): Der Einsatz der Fernerkundung zur Monteith, J.L. (1976): Vegetation and the atmosphere. Symposium Bestimmung der realen Evapotranspiration. Sonderheft of the Society for Experimental Biology, London-N.Y.-S. Bayerisches Landwirtschaftliches Jahrbuch, pp. 237-244. Francisco, Academic Press, 439 pp. Lott, F. and M.J. Miller (1997): A new sub-grid scale orographic Morgenschweis, G. (1999): Personal communication, Ruhr River drag parameterization: its formulation and testing. Quart. J. Association, Essen. R. Meteor. Soc. 123, pp. 101-128. Morton, F. I. (1983): Operational estimates of areal evapotranspi- Louis, J.-F. (1979): A parametric model of vertical eddy fluxes in ration and their significance to the science and practice of the atmosphere. Boundary-Layer Meteor. 17, pp. 187-202. hydrology. J. of Hydrology 66, pp. 1-76. Lützke, R. (1965): Über die Tauglichkeit der Lysimetermethode National Academy of Sciences (1988): Atmospheric prediction. für Wasserhaushaltsuntersuchungen. Besondere Mitt. zum Prediction and research problems. Committee on Gewässerkdl. Jb der DDR Nr. 4. Atmospheric Sciences, Panel on Precipitation Process. Lynch, P., D. Giard and V. Ivanovici (1997): Improving National Research Council, Wash. D. C. Nat. Academy Press, the Efficiency of a Digital Filtering Scheme for 63 pp. Diabatic Initialization. Monthly Weather Review 125, National Environment Research Council (1975): Flood Studies pp. 1976-1982. Report. Meteorological Studies, Vol. II, London. Makkink, G.F. (1957): Ekzameno de la formulo de Panman, Neil, L. and J. Spagnol (1987): Latest methods of inferring rainfall Netherlands J. Agr. Sci., No. 5. intensity and distribution from satellite data over the Eastern Malicki, M.A., R. Plagge, M. Renger u. R.T. Walczak (1992): Pacific. Preprints for a workshop on satellite and radar Application of Time Domain Reflectometry (TDR) soil mois- imagery interpretation, EUMETSAT, held at Reading, UK, ture microprobe for the determination of unsaturated soil pp. 437-457. water characteristics from undisturbed soil cores. Irrigation Ninomiya, K. (1987): Mesoscale numerical weather prediction- Sci. 13, 65-72. numerical prediction of mesoscale severe phenomena in REFERENCES 69

Japan. In short- and medium-range numerical weather predic- Wasserflächen auf statischer Grundlage. Abh. Meteor. Dienst tion. Collection of papers presented at the WMO/IUGG NWP der DDR 16, pp. 119. Symosium, Tokyo, Special Volume of J. of the Richter, D. (1995): Ergebnisse langjähriger methodischer Meteorological Society of Japan, pp. 517-531. Untersuchungen zum systematischen Fehler NOAA (1972): National weather service river forecast system, konventioneller Niederschlagsmessungen. Ber. Dt. forecast procedure. Technical memorandum NWS HYDRO- Wetterdienst, Nr. 194, 93 p. 14, 250 pp. Ritter, B. and J.-F. Geleyn (1992): A comprehensive radiation Olbrisch, H.-D. (1975): Beitrag zur Methodik der Datenanalyse scheme for numerical weather prediction models with poten- bei Lysimeteruntersuchungen. Mitt. Inst. Wasserbau u. tial applications in climate simulations. Mon. Wea. Rev. 120, Wasserwirtschaft, Rhein-Westf. Techn. Hochschule Aachen, pp. 303-325. Heft 8. Rosenfeld, D., E. Amitai and D. B. Wolff (1995): Improved Pander, R. von (1982): Ortsbezogene statistische accuracy of radar WPMM-estimated rainfall upon application Modellinterpretation. Promet 1/2, pp. 49-53, Offenbach. of objective classification criteria. J. Appl. Met. 34, pp. 212- Paul, E. (1988): Die räumlichen und zeitlichen Variationen der 223. Niederschlagsverteilung im Rhein-Main Gebiet. Rohrer, M.B. (1992): Die Schneedecke im Schweizer Alpenraum Diplomarbeit am Institut für Meteorologie der TH Darmstadt. und ihre Modellierung. Zürcher Geogr. Schriften 49. Penman, H. L. (1948): Natural evaporation from open water, bare Roth, R. (1989): Klimaänderungen und ihre Wechselwirkungen soil and grass. Proceedings of the Royal Meteorological mit den Landoberflächen. Statusseminar Klimaforschung- Society, Serie A 193, London, pp. 12-145. sprogramm, GSF-Forschungszentrum für Umwelt und Petznick, M. (1988): Bestimmung der Gebietsverdunstung am Gesundheit GmbH, pp. 143-149. Beispiel des Naab Gebietes. Diplomarbeit am Ruhrverband (1994): Ruhrwassermenge 1994 (Annual Report on Meteorologischen Institut der Universität BONN, 143 pp. Water Management of the Ruhr River 1994). 71 pp. Pfützner, B., Kaden, St., Krone, A., Flacke, W. (1992): Rutter, A.J.; A. J. Mortan, P.C. Robins (1975): A predictive Flächendifferenzierte hydrologische Einzugsgebiets- model of rainfall interception in forests.II. Generalization modellierung bei Anwendung eines Geographsichen of the model and comparison with observations Informationssystems (GIS). Deutsche Gewässerkundliche in some coniferous and hardwood stands. J. Appl. Ecol. 12, Mitteilungen 36-2, pp. 48-56. 367-380. Priestley, C.H.B. u. R.J. Taylor (1972): On the assessment of Sadourny, R., A. Arakawa and Y. Mintz (1968): Integration of the surface heat flux and evaporation using large scale parame- nondivergent barotropic vorticity equation with an icosahe- ters. Monthly Weather Review 100 H. 2, 81-92. dral-hexagonal grid on the sphere. Mon. Wea. Rev. 96, Rachner, M. (1996): Ermittlung und Vorhersage von pp. 351-356. Gebietswerten der Schmelzwasser-abgabe aus der Schädler, B. (1980): Die Variabilitat der Evapotranspiration im Schneedecke im Echtzeitregime. Projekt SNOW. XVII. Einzugsgebiet Rietholzbach bestimmt mit Energiebilan- Konferenz der Donauländer über hydrol. Vorhersagen und zmethoden. Mitteilung Nr. 46 der VAW der ETH Zurich, hydrologisch-wasserwirtschaftl. Grundlagen. 115 pp. Rachner, M. und Matthäus, H. (1986): Project Snow: operational esti- Schädler, B. (1992): Die Abflu§vorhersage in der Schweiz. mation of snowcover development in the mountains of the Symposium Hochwasservor-hersage Karlsruhe 6.11.1992. German Democratic Republic. IAHS-Publ. No. 155, pp. 71-81. Tagungsband, pp. 105-110. Rachner, M., H. Matthäus, and G. Schneider (1997): Schilling, W. (1983a): Operationelle Niederschlagsvorhersagen Echtzeitvorhersage der Schneedeckenentwicklung und der mit empirisch identifizierten Modellen und optimaler Wasserabgabe aus der Schneedecke. Erste Ergebnisse aus Zustandschätzung. Mitteilungen Nr. 51 des Instituts für dem Projekt SNOW-D. Deutsche Gewässerkundliche Wasserwirtschaft der Univ. Hannover. Mitteilungen 41-3, pp. 98-106. Schilling, W. (1983b): Der Einflu§ räumlich variabler Rajagopalan, B. and U. Lall (1999): A k-nearest neighbor simula- Niederschläge auf Kanalisationsabflüsse. Workshop “Die tor for daily precipitation and other weather variables. Water Bedeutung der räumlichen Niederschlagsverteilung für den Resources Research 35, pp. 3089-3101. Abflu§proze§ in kleinen Einzugsgebieten” an der Univ. Reich, T. (1998): Generierung homogener Tageswertreihen der Hannover. Niederschlagshöhe in einem mesoskaligen Gitternetz. Ann. Schirmer, H. and V. Vent-Schmidt (1979): Das Klima der D. Meteorologie 37, 1, pp. 289-290. Bundesrepublik Deutschland Zeitraum 1931Ð1960. Lieferung Renger, M., O. Strebel and W. Giesel (1974): Beurteilung 1: Mittlere Niederschlagshöhen für Monate und Jahr. bodenkundlicher, kulturtechnischer und hydrologischer Deutscher Wetterdienst, Offenbach. Fragen mit Hilfe von klimatischer Wasserbilanz und boden- Schrödter, H. (1985): Verdunstung — Anwendungsorientierte physikalischen Kennwerten (1. Bericht). Zeitschrift für Me§verfahren und Bestimmungsmethoden. Springer Berlin- Kulturtechnik und Flurbereinigung 15/3, Hamburg-Berlin, Hamburg, 186 pp. pp. 148-160. Schultz, G. (1987): Das räumlich-zeitlich variable Richter, D. (1969): Ein Beitrag zur Bestimmung der Verdunstung Niederschlagsdargebot als hydrologisches Problem. In: Fleer, von freien Wasserflächen, dargestellt am Beispiel des H. (Hrsg.): Möglichkeiten der Niederschlagsvorhersage für Stechlinsees. Abh. Metekor. Dienst d. DDR, 11, Nr. 88. Hydrologie und Wasserwirtschaft. Mitt. VIII d. Richter, D. (1977): Zur einheitlichen Berechnung der Senatskommission. für Wasserforschung der DFG, VCH- Wassertemperatur und der Verdunstung von freien Verlagsgesellschaft Weinheim, pp. 49-63. 70 METEOROLOGICAL SYSTEMS FOR HYDROLOGICAL PURPOSES

Scofield, R.A. and V.J. Oliver (1977): A scheme for estimating Tracton, S., and E. Kalney (1993): Ensemble forecasting at NMC: convective rainfall from satellite imagery. NOM Technical practical aspects. Wea. Forecasting 8, pp. 379-398. Memorandum, NESS 86. Turc, L. (1961): Evaluation des besoins en eau d’ irrigation, evap- Seguin, B. (1989): Use of surface temperature in agrometeorology. otranspiration potentielle, formule simplifiee et mise a jour. Proc. of a course held at the Joint Research Center of the Annales agronomiques, Institut National de la Recherche Commission of the EC in the framework of the ISPRA- Agronomique 12, Paris, pp. 13-9. Courses, Ispra, Varese, Italy, 1987, pp. 221-241. Turpeinen, O.M., A. Abidi and W. Belhaouane (1986): La valida- Seguin, B. and B. Itier (1983): Using midday surface temperature tion de l’indice de precipitation. ESA Joumal 10/3. to estimate daily evaporation from satellite thermal IR data. U.S. Army Corps of Engineers (1956): Snow hydrology - Int. J. of Remote Sensing, 4/2, pp. 371-383. Summary report of the snow investigations. US Arrny North Simmer, C. (1996): Retrieval of precipitation from satellites. In: Pacific Div., Portland OR. Radiation and Water in the Climate System, Springer Berlin, U.S. Weather Bureau (1947): Thunderstorm rainfall. Hydromet. Heidelberg, New York, Barcelona, Budapest, Hongkong, Rep. 5, 331 pp. London, Milan, Paris, Santa Clara, Singapore, Tokyo, Uhlig, S. (1954): Berechnung der Verdunstung aus klimatologis- pp. 249-276. chen Daten. Mitteilungen des DWD 6, Offenbach. Smith, C.S. (1986): The reduction of errors caused by bright- Verworn, H.R. and W. Flender (1986): Niederschlagshöhen- bands in quantitative rainfall measurements made using radar. Dauer-Flächenbeziehungen. Promet 2/3, pp. 40-45, J. Atmos. Ocean. Technol. 3, pp. 129-141. Offenbach. Spencer, R.W. (1984): Passive microwave rain estimation over Vidal, A., Y. Kerr, J.P. Lagourdena and B. Seguin (1987): Teledetection land and ocean at 37 GHz. Reprints Conf. Sat./Remote et bilan hydrique: utilisation combinee d’un modele Sensing and Applic., Public. Amer. Met.Soc., pp. 64-67. agrometeorologique et des donees de l’IR thermique du satellite Sponagel, H. (1980): Zur Bestimmung der realen NOAA-AVHRR. Agric. For. Meteor. 39, pp. 155-175. Evapotranspiration landwirtschaftlicher Kulturpflanzen. Wales-Smith, B.G. and J.A. Arnott (1985): Casebook on opera- Geologsiches Jahrbuch, Reihe F, Heft 9. tional assessment of areal evaporation. OHR No. 22 Stewart, J.B. (1987): On the use of Penman-Monteith equation for (WMO-No. 635), pp. 50-79. determining areal evaporation. IAHS-Publ. 177, pp. 3-12. Wankiewicz, A. (1979): A review of water movement in snow. Proc. Sverdrup, H.U. (1936): Das maritime Verdunstungsproblem. Ann. Hanover Workshop, CRREL, Hanover NH, pp. 222- 252. Hydrogr. U. Marit. Meteorol. 64, H. 2. Wendling, U. (1995): Berechnung der Gras-Referenzverdunstung Swinbank, W.C. (1951): The measurement of vertical transfer of mit der FAO Penman-Monteith-Beziehung. Wasserwirtschaft heat and water vapor by eddies in the lower atmosphere, J. 85 H. 12, pp. 602-604. Meteorol. 8, 135-145. Wendling, U., J. Müller u. K. Schwede (1984): Ergebnisse von Swinbank, W.C. (1955): An experimental study of eddy transports Verdunstungsmessungen über Gras mit einem Off-line in the lower atmosphere. CSIRO, Vol. 2.. Datenerfassungssystem (2. Mitteilung). Z. Meteorol 34, Szabo, G. (1985): Casebook on operational assessment of areal pp. 190-202. evaporation. OHR No. 22 (WMO-No. 635), pp. 149-158. Wendling, U., H.-G. Schellin and M. Thomä (1991): Szeicz, G. (1970): Comments on the value of some wetting experi- Bereitstellung von täglichen Informationen zum ments in studies of crop evaporation. Water Recources Wasserhaushalt des Bodens für die Zwecke der agrarmeteo- Research 6, pp. 347-348. rologischen Beratung. Z. Meteorol. 41 (1991), pp. 468-475. Tetzlaff, G. (1987): Niederschlag aus meteorologischer Sicht. In: Wergen, W. (1984): Datenassimilation. Promet 1, pp. 7-16, Fleer, H. (ed.): Möglichkeiten der Niederschlagsvorhersage Offenbach. für Hydrologie und Wasserwirtschaft. Mitt. VIII der Werner, J. (1987): Energie- und Massenbilanzstudien an Wald und Senatskommission. für Wasserforschung der DFG,VCH- Wasser in St. Arnold bei Rheine 1981 und 1982. Münstersche Verlagsgesellschaft Weinheim, pp. 1-10. Geograph. Arbeiten 26, pp. 183-190. Thiessen, A.H. (1911): Precipitation of large areas. Monthly Wessolek, G. et al. (1990): Aspekte zum Wasser- und Weather Review 39, pp. 1082-1084. Wärmehaushalt versiegelter Standorte. Mitt. Dtsch. Thom, A.S. (1975): Momentum, mass and theat exchange. In: Bodenkdl. Gesellsch. 61, pp. 153-159. Monteith. J.L.: Vegetation and the Atmosphere. Academic Wicke, W. (1988): Studien zu einem Verdunstungsmodell für Press. London Vol. 1, pp. 57-109. einen Wald. Diplomarbeit am Institut für physische Thompson, N., I.A. Barrie and M. Ayles (1981): The Geographie der Universitat Freiburg, 56 pp.. Meteorological office rainfall and evaporation calculation Williamson, D.L. (1968): Integration for the barotropic vorticity equa- system MORECS. Met. Office, Bracknell UK, 69 pp. tion on a spherical geodesic grid. Tellus, 20, pp. 642-653. Thornthwaite, C.W. (1948): An approach toward a rational classi- World Meteorological Organization (1966a): Manual on fication on climate. Geographical Review 38, Burlington, pp. Meteorological Observing in Transport Aircraft (WMO-No. 55Ð94. 197), Geneva. Thornthwaite, C.W. and B. Holzmann (1942): Measurement of World Meteorological Organization (1966b): Measurement and evaporation from land and water surfaces. USDA Technical Estimation of Evaporation and Evapotranspiration. Bulletin 817, Washington. Technical Note No. 83, Geneva, WMO-No. 201, Geneva. Tiedtke, M. (1989): A Comprehensive Mass Flux Scheme for World Meteorological Organization (1971): Climatological Cumulus Parameterization in Large-Scale Models. Mon. Wea. Normals (CLINO) for CLIMAT and CLIMAT SHIP stations Rev. 117, pp. 1779-1800. for the period 1931Ð1960 (WMO-No. 117), Geneva. REFERENCES 71

World Meteorological Organization (1975): Intercomparison of World Meteorological Organization (1988a): Guidance Document Conceptual Models Used in Operational Hydrological on Satellite Capabilities to 2000 for Meteorology and Forecasting. OHR No. 7 (WMO-No. 429), Geneva. Hydrology (WMO/TD-No. 276), Geneva. World Meteorological Organization (1976): Use of WWW World Meteorological Organization (1988b): The World Weather Facilities for Hydrology. WMO-No. 451, Geneva. Watch — 25th Anniversary 1963Ð1988 (WMO-No. 709), World Meteorological Organization (1978): International Geneva. Operations Handbook for Measurement of Background World Meteorological Organization (1988c): Technical Atmospheric Pollution (WMO-No. 491), Geneva. Regulations (WMO-No. 49), Geneva. World Meteorological Organization (1981a): Guide to World Meteorological Organization (1988d): Validation of Agricultural Meteorological Practices (WMO-No. 134), Satellite Precipitation Measurements for the Global Geneva. Precipitation Climatology Project: Report of an International World Meteorological Organization (1981b): Manual on the Workshop, Washington, D.C., 17Ð21 November 1986. Global Observing System. Vol. I: Global Aspects, Vol. II: WCRP-1 (WMO/TD-No. 203), Geneva. Regional Aspects (WMO-No. 544), Geneva. World Meteorological Organization (1989): Guide on the Global World Meteorological Organization (1981c): Selection of Observing System (WMO-No. 488), Geneva. Distribution Types for Extremes of Precipitation. OHR World Meteorological Organization (1990): Fifth Planning No. 15, WMO-No. 560, Geneva. Meeting on World Climate ProgrammeÐWater (WMO/TD- World Meteorological Organization (1982a): Methods of No. 374), Geneva, WCAP-11. Correction for Systematic Error in Point Precipitation World Meteorological Organization (1993): Guide on the Global Measurement for Operational Use. OHR No. 21, Data-processing System. WMO-No. 305, Geneva. WMO-No. 589, Geneva. World Meteorological Organization (1994a): Guide to World Meteorological Organization (1982b): Satellites in Hydrological Practices (WMO-No. 168), Geneva. Meteorology, Oceanography and Hydrology. WMO-No. 585, World Meteorological Organization (1994b): Hydrological Geneva. Information Referral Service — INFOHYDRO Manual. World Meteorological Organization (1985a): INFOCLIMA OHR No. 28 (WMO-No. 683), Geneva. Catalogue of Climate System Data Sets. WCP-101, World Meteorological Organization (1994c): Meteorological WMO/TD-No. 55, Geneva. Services of the World. WMO-No. 2, Geneva. World Meteorological Organization (1985b): Use of Radar in World Meteorological Organization (1995): Manual on Codes. Meteorology. WMO-No. 625, Technical Note. No. 181, Volumes I.1, I.2 and II (WMO-No. 306), Geneva. Geneva. World Meteorological Organization (1996): Guide to World Meteorological Organization (1986a): Intercomparison of Meteorological Instruments and Methods of Observation. Models of Snowmelt Runoff. OHR No. 23, WMO-No. 646, WMO-No. 8, Geneva. Geneva. World Meteorological Organization (1999): Numerical Weather World Meteorological Organization (1986b): Manual for Prediction Progress Report for 1998 (WMO/TD-No. 968), Estimation of Probable Maximum Precipitation. OHR No. 1, Geneva, 1999 pp. 261. WMO-No. 332, Geneva. Ziehmann, C. (2000): Comparison of a single-model EPS with a World Meteorological Organization (1987): The World Weather multi-model ensemble consisting of a few operational Watch Programme 1988-1997. 5LTP, Part II, Vol. 1, models. Tellus A52, pp. 280-299. WMO-No. 691, Geneva. Zingg, T. (1951): Beziehung zwischen Temperatur und Schmelzwasser und ihre Bedeutung für Niederschlags- und Abflussfragen. IAHS Publication No. 32, pp. 266-269. OPERATIONAL HYDROLOGY REPORTS

WMO Report No. No. 332 No. 1 — Manual for estimation of probable maximum precipitation (Second edition) 337 No. 2 — Automatic collection and transmission of hydrological observations* 341 No. 3 — Benefit and cost analysis of hydrological forecasts. A state-of-the-art report 356 No. 4 — Applications of hydrology to water resources management* 419 No. 5 — Meteorological and hydrological data required in planning the development of water resources* 425 No. 6 — Hydrological forecasting practices* 429 No. 7 — Intercomparison of conceptual models used in operational hydrological forecasting 433 No. 8 — Hydrological network design and information transfer 461 No. 9 — Casebook of examples of organization and operation of Hydrological Services 464 No. 10 — Statistical information on activities in operational hydrology* 476 No. 11 — Hydrological application of atmospheric vapour-flux analyses* 513 No. 12 — Applications of remote sensing to hydrology 519 No. 13 — Manual on stream gauging — Volume 1: Field work — Volume 2: Computation of discharge 559 No. 14 — Hydrological data transmission 560 No. 15 — Selection of distribution types for extremes of precipitation 561 No. 16 — Measurement of river sediments 576 No. 17 — Case studies of national hydrological data banks (planning, development and organization) 577 No. 18 — Flash flood forecasting 580 No. 19 — Concepts and techniques in hydrological network design 587 No. 20 — Long-range water-supply forecasting 589 No. 21 — Methods of correction for systematic error in point precipitation measurement for operational use 635 No. 22 — Casebook on operational assessment of areal evaporation 646 No. 23 — Intercomparison of models of snowmelt runoff 650 No. 24 — Level and discharge measurements under difficult conditions 655 No. 25 — Tropical hydrology 658 No. 26 — Methods of measurement and estimation of discharges at hydraulic structures 680 No. 27 — Manual on water-quality monitoring 683 No. 28 — Hydrological information referral service — INFOHYDRO Manual 686 No. 29 — Manual on operational methods for the measurement of sediment transport 704 No. 30 — Hydrological aspects of combined effects of storm surges and heavy rainfall on river flow 705 No. 31 — Management of groundwater observation programmes 717 No. 32 — Cost-benefit assessment techniques and user requirements for hydrological data 718 No. 33 — Statistical distributions for flood frequency analysis 740 No. 34 — Hydrological models for water-resources system design and operation 749 No. 35 — Snow cover measurements and areal assessment of precipitation and soil moisture 773 No. 36 — Remote sensing for hydrology — Progress and prospects 754 No. 37 — Hydrological aspects of accidental pollution of water bodies 779 No. 38 — Simulated real-time intercomparison of hydrological models 804 No. 39 — Applications of remote sensing by satellite, radar and other methods to hydrology 803 No. 40 — Land surface processes in large-scale hydrology 806 No. 41 — An overview of selected techniques for analysing surface-water data networks

* Out of print