Future of weather and climate forecasting

Public-Private Engagement Publication No. 3

WMO-No. 1263

WMO Open Consultative Platform White Paper #1

Future of weather and climate forecasting

WEATHER CLIMATE WATER

Cover photo credits:

© iStock

© World Meteorological Organization, 2021

The right of publication in print, electronic and any other form and in any language is reserved by WMO. Short extracts from WMO publications may be reproduced without authorization, provided that the complete source is clearly indicated.

Editorial correspondence and requests to publish, reproduce or translate this publication in part or in whole should be addressed to:

Chairperson, Publications Board

World Meteorological Organization (WMO) 7 bis, avenue de la Paix

P.O. Box 2300

CH-1211 Geneva 2, Switzerland

Tel.: +41 (0) 22 730 84 03 Fax: +41 (0) 22 730 81 17 Email: publications@wmo.int

World Meteorological Organization, 2020: Origin, Impact and Aftermath of WMO Resolution 40. Geneva.

ISBN 978-92-63-11263-7

NOTE

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

The mention of specific companies or products does not imply that they are endorsed or recommended by WMO in preference to others of a similar nature which are not mentioned or advertised.

The findings, interpretations and conclusions expressed in WMO publications with named authors are those of the authors alone and do not necessarily reflect those of WMO or its Members

WMO-No. 1263

Public-Private Engagement Publication No. 3

WMO Open Consultative Platform White Paper #1

Future of weather and climate forecasting

WEATHER CLIMATE WATER

CONTENTS

FOREWORD . . . V

ACKNOWLEDGEMENTS . . . 1

1. INTRODUCTION . . . 3

1.1 The need for a vision for climate forecasting and weather prediction . . . 3

1.2 Objective and scope of this White Paper . . . 4

2. WEATHER AND CLIMATE FORECASTING: SETTING THE SCENE . . . 6

2.1 Brief history . . . 6

2.2 WMO coordination role . . . 7

2.3 Baseline 2020 . . . 11

3. CHALLENGES AND OPPORTUNITIES IN THE COMING DECADE . . . 12

3.1 Infrastructure for forecasting . . . 12

3.1.1 Observational ecosystem . . . 13

3.1.2 High-performance computing ecosystem . . . 14

3.1.3 Changing landscape: advances in infrastructure through public–private engagement . . . 15

3.2 Science and technology driving advancement of numerical prediction . . . 16

3.2.1 Evolution of numerical Earth-system and weather-to-climate prediction . . . 17

3.2.2 High-resolution global ensembles . . . 19

3.2.3 Quality and diversity of models . . . 19

3.2.4  Innovation through artificial intelligence and machine learning

. . . 19

3.2.5 Advancing together: leveraging through public–private engagement . . . 20

3.3 Operational forecasting: from global to local and urban prediction . . . 21

3.3.1 Computational challenges and cloud technology . . . 22

3.3.2  Verification and quality assurance . . . 23

3.3.3 Further automation of post-processing systems and the evolving role of human forecasters . . . 24

3.3.4 Leveraging through public–private engagement . . . 25

3.4 Acquiring value through weather and climate services . . . 25

3.4.1 User perspective . . . 26

3.4.2 Forecasts for decision support . . . 26

3.4.3 Bridging between high-impact weather and climate services . . . 26

3.4.4 Education and training for future operational meteorologists/forecasters . . . 27

4. CONCLUSIONS . . . 28

4.1 Towards improved systems for forecasting: global, regional and local approaches . . . 28

4.2 Progressing together with developing countries . . . 30

REFERENCES . . . 32

BIBLIOGRAPHY . . . 33

Amazing supercell in Colorado

© iStock

FOREWORD

The advancement of our ability to predict the weather and climate has been the core aspiration of a global community of scientists and practitioners, in the almost 150 years of international cooperation in meteorology and related Earth system sciences.

The demand for weather and climate forecast information in support of critical decision-making has grown rapidly during the last decade, and will grow even faster in the coming years. Great advances have been made in the utilization of predictions in many areas of human activities.

Nevertheless, further improvements in accuracy and precision, higher spatial and temporal resolution, and better description of uncertainty are needed for realizing the full potential of forecasts as enablers of a new level of weather- and climate-informed decision-making.

In June 2019, WMO launched the Open Consultative Platform (OCP), Partnership and Innovation for the Next Generation of Weather and Climate Intelligence, in recognition that the progress in weather and climate services to the society will require a community-wide approach with participation of the stakeholders from the public and private sectors, as well as academia and civil society. The OCP is expected to serve as a vehicle for sustainable and constructive dialogue among sectors, to help articulate a common vision for the future of the weather and climate enterprise in the coming decade and beyond.

Undoubtedly, the 2020s will bring significant changes to the weather, climate and water community: on the one hand through rapid advancement of science and technology, and on the other hand through a swiftly changing landscape of stakeholders with evolving capabilities and roles. Such changes will affect the way weather and climate forecasts are produced and used. This is the reason the OCP selected the theme of “Forecasting and forecasters” as one of the “grand challenges” of the coming decade, which will require collective analytics to identify opportunities and risks and provide advice to planners and decision makers of relevant stakeholder organizations.

This White Paper on the Future of Weather and Climate Forecasting is a collective endeavour of more than 30 lead scientists and experts to analyse trends, challenges and opportunities in a very dynamic environment. The main purpose of the paper is to set directions and recommendations for scheduled progress, avoiding potential disruptions and leveraging opportunities through public–private engagement over the coming decade. This is done through description of three overarching components of the innovation cycle: infrastructure, research and development, and operation. The paper presents the converging views of the contributors, but also accounts for some variations of those views in areas where different options exist for advancing our capacity to predict weather and climate.

Thus, it informs and provides for intelligent choices based on local circumstances and resources.

I am pleased to present the White Paper on the Future of Weather and Climate Forecasting to the global audience and to encourage the use of its findings and recommendations by decision makers, practitioners and scientists from all sectors of the weather and climate enterprise. I would like to acknowledge, with much appreciation, the work done by Dr Gilbert Brunet, Chair of the WMO Scientific Advisory Panel, as the lead author and coordinator of the group of more than 30 prominent scientists and experts worldwide who contributed to the paper. I would like also to express my sincere thanks to all the contributing authors and reviewers for devoting their time and sharing their knowledge and foresight for the benefit of the whole enterprise.

Prof. Petteri Taalas Secretary-General

ACKNOWLEDGEMENTS

This paper has been prepared by a drafting team led by Gilbert Brunet, Chief Scientist and Group Executive Science and Innovation, Bureau of Meteorology, Chair of the Science Advisory Panel, World Meteorological Organization.

The team of contributing authors includes (in alphabetical order):

Peter Bauer Deputy Director, Research Department, European Centre for Medium-Range Weather Forecasts

Natacha Bernier Director, Meteorological Research Division, Environment and Climate Change Canada

Veronique Bouchet Acting Director General, Canadian Centre for Meteorological and Environmental Prediction, Meteorological Service of Canada, Environment and Climate Change Canada

Andy Brown Director of Research, European Centre for Medium-Range Weather Forecasts Antonio Busalacchi President, University Corporation for Atmospheric Research, USA

Georgina Campbell

& Rei Goffer

Executive Director, ClimaCell.org; CSO and Co-Founder, ClimaCell

Paul Davies Principal Fellow of Meteorology and Chief Meteorologist, Met Office, UK Beth Ebert Senior Professional Research Scientist, Weather and Environmental Prediction,

Bureau of Meteorology, Australia Karl Gutbrod CEO, Meteoblue, Switzerland

Songyou Hong Fellow, Korean Academy of Science and Technology, Republic of Korea PK Kenabatho Associate Professor, Department of Environmental Science, University

of Botswana, Botswana

Hans-Joachim Koppert Director, Business Area “Weather Forecasting Services”, Deutscher Wetterdienst, Germany

David Lesolle Lecturer (Climatologist), Department of Environmental Science, University of Botswana, Botswana

Amanda Lynch Lindemann Professor, Institute for Environment and Society, Department of Earth, Environmental and Planetary Sciences, Brown University, USA

Jean-François Mahfouf Ingénieur Général des Ponts, Eaux et des Forêts, Météo-France, Toulouse, France Laban Ogallo* Professor, University of Nairobi, Kenya

* The contributors to this White Paper express their great sadness of the demise of Prof. Laban A. Ogallo who passed away in November 2020. Prof.

Ogallo was one of the pioneers of climate science in Africa and he provided a significant input to the White Paper.

Tim Palmer Royal Society Research Professor of Climate Physics, Professorial Fellow, Jesus College Oxford, UK

David Parsons President’s Associates Presidential Professor, Director Emeritus, School of Meteorology, University of Oklahoma, USA

Kevin Petty Director, Science and Forecast Operations and Public-Private Partnerships, The Weather Company, an IBM Business

Dennis Schulze Managing Director, MeteoIQ, Chairman of PRIMET, Chairman of Verband Deutscher Wetterdienstleister e.V. (VDW)

Ted Shepherd Grantham Professor of Climate Science, University of Reading, UK

Thomas Stocker Professor, Head of Division Climate and Environmental Physics, Physics Institute, University of Bern, Switzerland; President of the Oeschger Centre for Climate Change Research, Switzerland

Alan Thorpe Visiting Professor, University of Reading, UK

Rucong Yu Deputy Administrator, China Meteorological Administration

The group of reviewers who provided valuable comments and proposals for improving the narrative of the paper included:

V Balaji Head, Modeling Systems Group, Princeton University, USA Brian Day Vice-President, Campbell Scientific, Canada

Andrew Eccleston General Secretary, PRIMET

Roger Pulwarty Physical Scientist at National Oceanic and Atmospheric Administration, USA Julia Slingo Retired, former Chief Scientist of the UK Met Office (2009-2016)

The work of the drafting team was supported by Dimitar Ivanov and Boram Lee from the Secretariat of the World Meteorological Organization.

1. INTRODUCTION

1 The Earth system encompasses the atmosphere and its chemical composition, the oceans, land/sea ice and other cryosphere components as well as the land surface, including surface hydrology and wetlands, lakes and human activities. On short timescales, it includes phenomena that result from the interaction between one or more components, such as ocean waves and storm surges. On longer timescales for climate applications, it includes terrestrial and ocean ecosystems, encompassing the carbon and nitrogen cycles and slowly varying cryosphere components such as large continental ice sheets and permafrost.

2 The term “weather and climate enterprise” is used to describe the multitude of systems and entities participating in the production and provision of meteorological, climatological, hydrological, marine and related environmental information and services. The enterprise includes public-sector entities (NMHSs and other governmental agencies), private-sector entities (equipment manufacturers, service-provider companies, private media companies, and so forth), academic institutions, and civil society entities (community-based entities, NGOs, national meteorological societies, scientific associations, etc.). The weather and climate enterprise has global, regional, national and local dimensions.

1.1 The need for a vision for weather and climate forecasting

Weather and climate forecasting is a leading environmental and socioeconomic challenge – whether on an urban or planetary scale, or covering a few hours or a few seasons. Significant progress has been achieved in numerical Earth- system

and weather-to-climate prediction (NEWP) over the past six decades, through collaborative efforts by many institutions from the public, private and academic sectors at national and international levels. As the new decade 2021–2030 begins, vigorous NEWP and high-performance computing (HPC) programmes of multidisciplinary research and development (R&D) worldwide are making innovative contributions to this ongoing challenge.

Earth-system models are developing in complexity, incorporating additional processes and needing more observations of diverse elements of the environment.

Thus, observational and HPC infrastructures are central to future advancement of NEWP systems. Numerical modelling and prediction were among the main motivations behind the first computer applications 70 years ago, and they are still a major use case for HPC today. Likewise, advances in satellite-based observations and telecommunications utilized in NEWP are at the forefront of technological innovations. Computational power and high-quality observations drive improvements in weather and climate models such as refined space–

time resolution, better representation of the physical processes and enhanced data-assimilation techniques.

They also help to quantify forecasting and modelling uncertainties, although trade-offs are often required among these. The achievements and improvements are remarkable; for instance, the mid-latitude 5-day weather forecast today is as accurate as the 1-day forecast 40 years ago. More accurate and reliable forecasts are produced by advances in science and technology. These

forecasts provide major support for life-saving decisions through mitigation of the risk of weather and climate hazards. In addition, improved forecasts create tangible socioeconomic benefits in many economic sectors (for example, energy, transport and agriculture), through avoided losses, better management of resources and enhanced opportunities for revenue.

Policy debates around the future of the planet and society are intense in a world with significant global technological transformations and environmental risks.

Such debates shape high demands for better weather and climate information and for services addressing the risks and socioeconomic impacts of the weather, climate and water hazards. The importance of climate risk-based decision-making is increasing substantially with population growth. This is particularly so in major cities, often on coasts, where more people and assets are exposed and vulnerable to weather, climate, water, ocean and even space hazards. Essential services (for example, power, water, transport, telecommunications, the Internet and finance) are also exposed to these hazards.

Meeting the demands for highly localized and accurate information with frequent updates, as well as tailored services for informed decision-making over multiple timescales, will require a new level of collaboration within the weather and climate enterprise². Working with user communities in the co-design of fit-for-purpose information and services will also be important.

Traditional risk assessment and management strategies are increasingly challenged by systemic risks that connect local conditions to broader global systems. These systemic risks are unconstrained and include the potential for thresholds and surprises, along with the need to account for the evolution of weather and climate high-impact events, variability, and change across time and space. Addressing such complex risks requires analytical, technical and deliberative capacity, as well as consideration of equity and broader participation to consider implications beyond a single project or decision context. Thus, when considering the future of weather and climate forecasting, the need for an international multidisciplinary research agenda,

covering both applications and services, and providing for a systematic link between NEWP science and policy/

decision-making, should be recognized.

Over the coming decade, these developments will drive many innovations to satisfy diverse socioeconomic needs:

• Higher-resolution and more localized and relevant NEWP forecasts, updated frequently (hourly or even sub-hourly) for cities and other areas of interest. These will be combined with nowcasting tools optimized to provide users with enhanced decision support based on more timely forecast updates (on a minutes scale) before and during high-impact weather.

• Enhanced quality of observational data for analyses and for assimilation into NEWP systems, as well as increased number of Earth-system observations of all types done in an economic and sustainable way.

• Transition to a full Earth-system numerical prediction capability with coupled subcomponents, to deliver a wider breadth of information-rich data that are consistent across the atmosphere, land and ocean, including waves, sea ice and hydrological elements.

Aligned with the Earth-system framework and approach, these NEWP systems will enable prediction of multi- hazard events in a fully consistent manner, providing more precise, accurate and reliable information.

• Seamless weather and climate risk-based services will be further developed, providing insights from minutes to seasons, to enable improved decision-making and risk reduction. This will include the integration of historical observations and forecasts with a full characterization of uncertainty.

1.2 Objective and scope of this White Paper

The main objective of this paper is to provide a basis for informed decision-making by weather and climate enterprise stakeholders in planning their activities and investments in NEWP and operational forecasting during the coming decade. This decade, often referred to as the

“decade of digital transformation”, will bring profound impacts on organizations of all types. The weather and climate enterprise will also undergo significant changes since it is highly driven by data and information technology (IT). The High-level Round Table on the launch of the Open Consultative Platform (OCP) Partnership and Innovation for the Next Generation of Weather and Climate Intelligence (5–6 June 2019, Geneva) highlighted this expectation, and included “Forecasting and … forecasters” among the five themes on key challenges for the next decade (WMO, 2019a).

This reflects the recognition that the innovation cycle (see Figure 1) for weather and climate forecasts includes various stakeholders from public, private and academic sectors. The important drivers of the innovation cycle are computational and observational infrastructures (in the middle of the figure), and increasing stakeholder and customer demand (on the circumference of the figure) for tailored and seamless weather and climate forecasting (localized, timely, precise and accurate). Figure 1 shows that stakeholders and customers can push clockwise new initiatives at different positions in the innovation cycle:

R&D, operation and services. The structure of this paper is aligned along three components of the innovation cycle:

infrastructure, R&D and operation. Stakeholders engaged in all three components will have to make strategic choices in the coming years, and some will struggle to keep up as technologies continue to combine and advance, and new ways of doing business appear quickly.

© iStock

Macedonia

Figure 1. The innovation cycle: the public–private engagement challenge This paper aims to help decision makers, researchers

and even users in the rapidly changing landscape of the weather and climate enterprise, by compiling views, knowledge and expertise of a group of prominent scientists and practitioners from the public, private and academic sectors. It does not attempt to provide unique solutions on the many open questions of the future of weather and climate forecasting. Instead, it serves to improve the understanding of ongoing R&D, and to identify technological trends and sometimes possible impediments to progress such as the lack of data sharing. In this way, risks and opportunities for each player can be better assessed, and decisions made on future organizational plans and investment can be better informed.

The scope of this paper is purposefully restricted to the process of NEWP innovation and production of weather and climate forecasts, and also to climate insight when there is a close relationship with NEWP and climate change science issues. The production value chain in the operation (see Figure 1) is increasingly developing towards seamless interfaces among its elements.

Thus, the paper also partly treats elements at the input side (observational data), as well as at the output side (generation of products for services) of this chain.

Science and research that form the basis for forecasting and determine its foreseen advances are also discussed.

Technology is another key factor in the discussion of the future with many exciting developments in IT and computing that bring enormous opportunities for improved quality and efficiency.

The many contributors to this paper were all people dealing with Earth-system weather and climate numerical prediction. However, for the purposes of this paper, they were asked to try to forecast the future of their enterprise.

Engaging 27 such contributors may be seen as applying the ensemble prediction method, which highlights uncertainties and potential different trajectories of development. Therefore, the individual views and inputs of each contributor are available at the following weblink:

https://library.wmo.int/doc_num.php?explnum_id=10552.

The bibliography at the end of this white paper also provides an extensive list of further reading.

INFRASTRUCTURE

2. WEATHER AND CLIMATE FORECASTING:

SETTING THE SCENE

2.1 Brief history

Operational weather forecasting and climate predictions started long before numerical modelling using computers became possible. There have always been attempts to understand weather and climate patterns and eventually foresee their future states, due to the impact on humans and their activities. In the absence of theories and knowledge of the forces driving weather behaviour, such attempts have been part of astrology or folklore for centuries. There were several important theoretical advances in the early nineteenth century, including a growing understanding of the nature of storms. The efforts for organized systematic collection of observational data and using these data for predicting weather events started later in that century. A common reference point for the start of “weather forecasting” is the work of Admiral FitzRoy during the 1850s and 1860s.

FitzRoy started issuing storm warnings for sailors in 1860, and, one year later, general weather forecasts (the first such forecast appeared in The Times on 1 August 1861).

FitzRoy’s work was enabled by the rapidly expanding use of electrical telegraphs, which allowed collection of observations from several stations, and some primitive situational analysis. It seems he also introduced the use of the terms “forecast” and “forecasting” in place of

“prognostication”, which had been used previously (BBC News, 2015).

These first attempts at weather forecasting were, understandably from today’s perspective, rather unsuccessful. Nevertheless, interest in developing knowledge and methods for meteorological analysis and prediction grew rapidly during the last decades of the nineteenth century and the early decades of the twentieth century. Collecting and exchanging (through telegraphs) data across national borders established one of the early cases of globalized infrastructure and an unprecedented international cooperation between scientists and practitioners. The “weather knows no borders” slogan called for a partnership that needed governance – to initiate a global standardization of methods and procedures for research and operations in each individual country. The formal start of such organized international cooperation was the first International Meteorological Congress in Vienna in August 1873. This event established a format of collaboration that WMO continues today.

Without going into the details of the pre-NEWP decades of weather forecast development, it is worth mentioning that the knowledge and methods improved slowly.

However, the number of incorrect forecasts (visible to the public, due to the popularity of the subject) led to a prevailing scepticism about the ability of science to deal with the challenge and to make operational forecasting possible with reliable day-to-day outcomes.

This may have been the reason for Margules to state, in the early twentieth century, that weather forecasting was “immoral and damaging to the character of a meteorologist” (Lynch, P., 2006).

However, developments at the beginning of the twentieth century quickly changed the pessimism of Margules into a much more optimistic scenario for the future of weather forecasting. Since the ground-breaking work of Abbe (1901), Bjerknes (1904) and Richardson (1922), the challenge of NEWP has been related to an initial value conditions problem of mathematical physics (based on the non-linear equations governing fluid flow), and has been approached using numerical techniques and algorithms.

The success of the first numerical prediction by Charney et al. (1950) launched a spectacular trend of innovations in NEWP over the following seven decades. Routine, real-time forecasting with NEWP started in the mid- 1950s and was introduced in operations in the 1960s.

Improved observational coverage, the advent of satellite observations, the steady growth of computer power and breakthroughs in the theory of Earth-system coupled processes all underpinned a successful story of weather forecasting in the NEWP era.

The high cost of NEWP, including the capital investment for computers and their running and maintenance costs, as well as resources needed in R&D, meant that the most developed nations had the highest concentration of major developments. Nonetheless, exemplary cooperation and knowledge-sharing with scientists from many countries and institutes has nurtured advancement of NEWP.

European countries undertook a strong collaborative move with the establishment of the European Centre for Medium-Range Weather Forecasts (ECMWF) in 1975 as an intergovernmental organization.

Progress in NEWP is often illustrated by the improvement in the horizontal and vertical resolution of operational models. There has been an almost 40 times increase in the horizontal resolution of global models (from about 400 km in the early 1960s, to less than 10 km in 2020);

in addition, regional fine-mesh models have reached a 1-km resolution. In the vertical direction, from the early one- and three-layered quasi-geostrophic models, today’s models utilize more than 130 levels, reaching an altitude of about 80 km (pressure of 0.01 hPa).

There are several excellent papers on the history of the highlights of NEWP developments (Pudykiewicz and Brunet, 2008; Benjamin et al., 2019; see also Box 2). For example, Benjamin et al. (2019) reviewed the progress in forecasting and NEWP applications over the 100-year period from 1919 to 2019, and divided the period into four ”eras” as follows:

• Era 1 (1919–39: maps only; observations and extrapolation/advection techniques)

• Era 2 (1939–56: increasing science understanding;

application especially to aviation; birth of computers)

• Era 3 (1956–85: advent of NEWP and dawn of remote-sensing)

• Era 4 (1985–2018: weather forecasting, and especially NEWP, matured and penetrated virtually all areas of human activity)

The same study also provided an outlook for Era 5, encompassing the next 30 years until 2050, which could well be named the era of “next generation of weather and climate Earth-system intelligence”.

2.2 WMO coordination role

It is important to highlight the role of WMO in the progress of and insight into weather and climate forecasting. The WMO technical commissions (for example, the Commission for Atmospheric Sciences, the Commission for Climatology, the Commission for Basic Systems, and the Joint Technical Commission for Oceanography and Marine Meteorology) were instrumental in facilitating international collaboration and knowledge-sharing. The World Weather Research Programme and the World Climate Research Programme were at the forefront of scientific efforts underpinning progress in NEWP development and in research-to- operation transition.

Establishment of the WWW programme was one of the main WMO contributions. This was initiated on 20 December 1961 with the adoption of Resolution 1721 (XVI) by the United Nations General Assembly (United Nations, 1961), which called upon WMO to undertake a comprehensive study of measures:

Box 1. Major milestones in weather and climate forecasting

• 1861: Met Office weather forecast services using telegraphs established by FitzRoy

• 1873: Working towards global meteorological observatories and international data sharing with the foundation of the International Meteorological Organization in Vienna

• 1900–1922: Birth of numerical weather prediction (NWP) with the work of Abbe (1901), Bjerknes (1904) and Richardson (1922)

• Early 1920s: Onset of statistical climate prediction and global atmospheric teleconnection insights pioneered by Walker

• 1950: First computer NWP forecast on ENIAC (Electronic Numerical Integrator and Computer) by Charney et al. (1950)

• 1960 onward: Satellite-based meteorological observations and telecommunications at the forefront of technological innovations since the launch of the first weather satellite TIROS-1

• 1960s onward: Emergence of general circulation models for climate research and forecasting

• 1962: Establishment of the World Weather Watch (WWW) programme with its three main components (Global Observing System, Global Telecommunication System and Global Data-Processing System)

• 1963: Lorenz’s seminal work on chaos initiated atmospheric predictability theory and paved the way to numerical ensemble prediction in the 1980 and 1990s

• 1969: Launch of the Global Atmospheric Research Program (GARP) led by Charney

“(a) To advance the state of atmospheric science and technology so as to provide greater knowledge of basic physical forces affecting climate and the possibility of large-scale weather modification;

(b) To develop existing weather forecasting capabilities and to help Member States make effective use of such capabilities through regional meteorological centres”

It is interesting to note the emphasis of “large-scale weather modification”, which was hoped would mitigate the unfavourable weather impacts on human activities.

This hope proved over-optimistic, as became clear in the following decades, and weather modification research and operational activities have not developed much.

However, those early intentions for human control on weather and climate may be revived to a certain extent due to recent geoengineering ideas to mitigate climate change. However, the gains of geoengineering relative to reduced greenhouse gas emissions and against the hazards it could bring to the environment must be balanced rigorously.

Paragraph (b) above of Resolution 1721 is significant for the scope of this White Paper. In cooperation with partners, WMO established the WWW programme

composed of three main components: the Global Observing System, the Global Telecommunication System and the Global Data-Processing and Forecasting System (GDPFS), coupled with the Meteorological Applications Programme. Thus, the output of the WWW system was a global set of observational and forecast data that were shared among WMO Member States and Territories, and served as input for development of the whole spectrum of user-oriented applications and services.

Today, GDPFS is an elaborate system of global and regional centres, including nine World Meteorological Centres (WMCs) and 11 Regional Specialized Meteorological Centres (RSMCs), with geographical specialization (see Figures 2 and 3). Various centres are tasked with production of: global deterministic and ensemble NWP; limited-area deterministic and ensemble NWP; nowcasting; various specialized forecasting activities, like tropical cyclone forecasting;

atmospheric transport and dispersion modelling (nuclear and non-nuclear); atmospheric sandstorm and duststorm forecasting; numerical ocean wave prediction; aviation forecasting; and so forth. In addition, 13 centres are designated as Global Producing Centres for Long-range Prediction (monthly to seasonal), and four centres as Global Producing Centres for Annual to Decadal Climate Prediction.

• 1969 onward: Global NWP innovations since the first global NWP simulation by Robert

• 1975: Federation of global NWP R&D effort in Europe with the foundation of the European Centre for Medium-range Weather Forecasts (ECMWF)

• 1979: First GARP Global Experiment, to gather the most detailed observations ever of the global atmosphere

• 1980s onward: Development of coupled ocean–

atmosphere climate models

• 1992: Operational implementation of ensemble prediction systems at the ECMWF and the National Centers for Environmental Prediction (NCEP)

• 1997: Ground-breaking numerical prediction advances in the use of multiple sources of Earth-system observations with the introduction at ECMWF of four- dimensional data assimilation

• 2002: Earth Simulator, Japan – a landmark supercomputer investment for climate, weather and geophysical research

• 2007: A great step forward for weather and climate Earth-system forecasting with 3 000 Argo oceanic floats in global operation

• 2015 onward: Dealing with prediction uncertainty in data assimilation with ensemble–variational data- assimilation techniques

Figure 2. WMO-designated GDPFS centres (nowcasting and weather forecasting, up to 30 days) Source: WMO (2019)

La Reunion

* World Global Centres are also Global Producing Centres for a) Deterministic Numerical Weather Prediction, b) Ensemble Numerical Weather Prediction, and c) Long-Range Forecasts.

** RSMC for nuclear and non-nuclear emergency response have Atmospheric Transport and Dispersion Modelling (ATDM) capabilities.

*** NRT stands for Non-Real-Time DESIGNATIONS USED

The depiction and use of boundaries, geographic names and related data shown on maps and included in lists, tables, documents, and database on this website are not warranted to be error free nor do they necessarily impy official endorsement or acceptance by the WMO.

Winnipeg

Callao

Valparaiso

Legend

Buenos Aires Niteroi Anchorage

Brasilia Miami

Washington Montreal Ottawa

Tunis

Tromso Offenbach St Petersburg Moscow Obninsk

Dakar

Pretoria

Dar es Salaam Nairobi Jeddah Karachi

New Delhi Tashkent

Novosibirsk

Beijing Tokyio Vladivostok Khabarovsk

Hong Kong

Honolulu

Darwin

Melbourne

Wellingtone Nadi Cairo

Athens Rome Vienna

Vacoas Casablanca

Barcelona

Algier

World Meteorological Centres (WMCs)* (9) RSMCs Geographic Specialization (12)

RSMCs (NRT***) Lead Centre for Coordination of Wave Forecast (1) RSMCs (NRT***) Lead Centre for Coordination of EPS Verification (1) RSMCs (NRT***) Lead Centre for Coordination of DNV (1)

RSMCs Numerical Ocean Wave Prediction (4) RSMCs Tropical Cyclone Forecasting (6) RSMCs Severe Weather Forecasting (5) RSMCs Severe Weather Forecasting (24)

RSMCs Nuclear Emergency Response** (10) RSMCs Non-Nuclear Emergency Response** (3) RSMCs Sand and Duststorm Forecasts (2) RSMCs Nowcasting (3)

RSMCs Limited Area Ensemble NWP (2) RSMCs Global Ensemble NWP (7)

RSMCs Limited Area Deterministic NWP (6) RSMCs Global Deterministic NWP (8)

ICAO designated Volcanic Ash Advisory Centres (9) Edmonton

Exeter ECMWF

Toulouse

La Reunion

* World Global Centres are also Global Producing Centres for a) Deterministic Numerical Weather Prediction, b) Ensemble Numerical Weather Prediction, and c) Long-Range Forecasts.

** RSMC for nuclear and non-nuclear emergency response have Atmospheric Transport and Dispersion Modelling (ATDM) capabilities.

*** NRT stands for Non-Real-Time DESIGNATIONS USED

The depiction and use of boundaries, geographic names and related data shown on maps and included in lists, tables, documents, and database on this website are not warranted to be error free nor do they necessarily impy official endorsement or acceptance by the WMO.

Winnipeg

Callao

Valparaiso

Legend

Buenos Aires Niteroi Anchorage

Brasilia Miami

Washington Montreal Ottawa

Tunis

Tromso Offenbach St Petersburg Moscow Obninsk

Dakar

Pretoria

Dar es Salaam Nairobi Jeddah Karachi

New Delhi Tashkent

Novosibirsk

Beijing Tokyio Vladivostok Khabarovsk

Hong Kong

Honolulu

Darwin

Melbourne

Wellingtone Nadi Cairo

Athens Rome Vienna

Vacoas Casablanca

Barcelona

Algier

World Meteorological Centres (WMCs)* (9) RSMCs Geographic Specialization (12)

RSMCs (NRT***) Lead Centre for Coordination of Wave Forecast (1) RSMCs (NRT***) Lead Centre for Coordination of EPS Verification (1) RSMCs (NRT***) Lead Centre for Coordination of DNV (1)

RSMCs Numerical Ocean Wave Prediction (4) RSMCs Tropical Cyclone Forecasting (6) RSMCs Severe Weather Forecasting (5) RSMCs Severe Weather Forecasting (24)

RSMCs Nuclear Emergency Response** (10) RSMCs Non-Nuclear Emergency Response** (3) RSMCs Sand and Duststorm Forecasts (2) RSMCs Nowcasting (3)

RSMCs Limited Area Ensemble NWP (2) RSMCs Global Ensemble NWP (7) RSMCs Limited Area Deterministic NWP (6) RSMCs Global Deterministic NWP (8)

ICAO designated Volcanic Ash Advisory Centres (9) Edmonton

Exeter ECMWF

Toulouse

Figure 3. WMO-designated GDPFS centres (long range and climate forecasting, over 30 days) Source: WMO (2019)

* World Global Centres are also Global Producing Centres for a) Deterministic Numerical Weather Prediction, b) Ensemble Numerical Weather Prediction, and c) Long-Range Forecasts.

** NRT stands for Non-Real-Time.

*** ADCP stands for Annual to Decadal Climate Prediction

**** LRFMME stands for Long-Range Forecast Multi-Model Ensemble DESIGNATIONS USED

The depiction and use of boundaries, geographic names and related data shown on maps and included in lists, tables, documents, and database on this website are not warranted to be error free nor do they necessarily impy official endorsement or acceptance by the WMO.

Legend

Buenos Aires Brasilia CPTEC Washington

Bridgetown

Guayaquil

Montreal

Tunis

Tripoli

Offenbach Moscow

Pretoria Nairobi

Pune Beijing

Tokyio Seoul

Melbourne Cairo

Casablanca Barcelona

Algier

Niamey

World Meteorological Centres (WMCs)* (9)

RSMCs (NRT***) Lead Centre for Coordination of ADCP*** (1) RSMCs (NRT***) Lead Centre for Coordination of LRFMME**** (2) RSMCs (NRT***) Lead Centre for Coordination of LRF verification (2)

RCC - Networks Regional Climate Prediction and Monitoring NODEs (11) RCC Regional Climate Prediction and Monitoring (9)

GPC for ADCP*** (4)

GPC for Long-Range Forecasting (13) Exeter

ECMWF

Toulouse De Bilt

* World Global Centres are also Global Producing Centres for a) Deterministic Numerical Weather Prediction, b) Ensemble Numerical Weather Prediction, and c) Long-Range Forecasts.

** NRT stands for Non-Real-Time.

*** ADCP stands for Annual to Decadal Climate Prediction

**** LRFMME stands for Long-Range Forecast Multi-Model Ensemble DESIGNATIONS USED

The depiction and use of boundaries, geographic names and related data shown on maps and included in lists, tables, documents, and database on this website are not warranted to be error free nor do they necessarily impy official endorsement or acceptance by the WMO.

Legend

Buenos Aires Brasilia CPTEC Washington

Bridgetown

Guayaquil

Montreal

Tunis

Tripoli

Offenbach Moscow

Pretoria Nairobi

Pune Beijing

Tokyio Seoul

Melbourne Cairo

Casablanca Barcelona

Algier

Niamey

World Meteorological Centres (WMCs)* (9)

RSMCs (NRT***) Lead Centre for Coordination of ADCP*** (1) RSMCs (NRT***) Lead Centre for Coordination of LRFMME**** (2) RSMCs (NRT***) Lead Centre for Coordination of LRF verification (2)

RCC - Networks Regional Climate Prediction and Monitoring NODEs (11) RCC Regional Climate Prediction and Monitoring (9)

GPC for ADCP*** (4)

GPC for Long-Range Forecasting (13) Exeter

ECMWF

Toulouse De Bilt

2.3 Baseline 2020

To provide a vision for developments in weather forecasting and climate predictions over the next 10 years (Vision 2030) and beyond, it is important to set up a baseline: the present situation in year 2020. The main elements of the ”current state” – baseline 2020 – are as follows:

• High-resolution global deterministic models for the medium range operate at horizontal resolutions of

~10 km, with 50–140 vertical layers and ~10 prognostic variables. These models are usually run for 10–15 days with an update cycle of 6 h (four times a day).

• Ensemble prediction systems for the medium range use ~50 ensemble members and the horizontal resolution is ~20 km. For an extended range of up to 45 days, the horizontal resolution is ~35–40 km.

• As these systems are extended beyond the medium range towards the seasonal range, the horizontal resolution is usually downgraded to 40–100 km, while vertical levels and ensemble size are kept constant. Major updates in these systems occur less frequently, typically every 5 years or so, with a rate of improvement closer to a week of extra lead time per decade of development for the Madden–Julian oscillation (Kim et al., 2018).

3. CHALLENGES AND OPPORTUNITIES IN THE COMING DECADE

Operational weather forecasting based on numerical prediction systems has continuously improved over the past few decades. The usefulness of NEWP forecasts has been pushed to lead times beyond 10 days for some high- impact weather phenomena such as mid-latitude snowstorms in North America. However, the steady progress has been at a slower pace for some forecasted elements, like quantitative precipitation, where more efforts are needed.

By 2050, it is envisaged that NEWP will approach the theoretical limit of mid-latitude predictability of the chaotic atmosphere – a century after the first numerical weather forecasts were produced by Charney and his team. Several factors have steered progress, including: advances in NWP underpinned by increasing HPC capacity; improved observational instrumentation providing more accurate data with higher temporal and spatial resolutions; better representation of complex physical processes; better model initialization through the utilization of expanding satellite observations and more effective data-assimilation methods; and use of ensembles to represent uncertainty in the initial state and model processes. Furthermore, scientific insight across fields ranging from meteorology to computer science has provided a growing suite of tools, catalysing innovations in numerical prediction system design. On the policy side, prevailing free and open data sharing among countries and institutions has provided access to observational data for operational and research purposes, which has facilitated progress. However, in some areas, policies implying commercial or other conditions in accessing important data sets have slowed the progress.

The increased availability and adoption of forecast- driven tools for weather- and climate-informed decision- making, especially by the commercial sector, have also facilitated major progress. The demand for such decision-support tools by many industry sectors is growing rapidly when striving to mitigate weather and climate impacts on operations and profits. This presents challenges and opportunities for further advancing weather and climate forecasting, which is yet to reach its full potential.

3.1 Infrastructure for forecasting

Two main infrastructural elements define the performance of NEWP systems: the observational ecosystem that provides the input data and the IT ecosystem including communication, computers and storage, with all internal and external interfaces.

The steady improvements in the skill of NEWP are based, in large part, on the performance of the global observing system of systems, which has advanced significantly in the past few decades. Recent examples of such improvements include the development of space- based measurements for wind and clouds/precipitation using lidar and radar technologies, respectively. Remote- sensing technologies such as infrared and microwave sounders/imagers in all-sky conditions, combined with advanced ground-based observational networks as the bed-rock, have provided accurate initial conditions that are a key factor for improved synoptic-scale forecast skill.

In addition to atmospheric measurements, the evolving capabilities of other Earth-system observations has made progression possible towards integrated Earth- system modelling and forecasting. For example, operational oceanography has increased the availability of observations necessary to improve ocean state estimation, including its mesoscale variability. This has brought rapid improvement in the accuracy of oceanic forecasts. Starting in the 1990s, oceanic measurements, like Argo floats and Tropical Ocean Global Atmosphere arrays, permitted operationalization of forecasts of storm surges, waves and sea ice for use by operational centres. Progress has also been made in land-surface hydrology, but much more is needed to advance terrestrial hydrology observations and integrate these observations into NEWP systems at all timescales.

In contrast to the advances in remote-sensing, there has been alarming evidence that in situ, high-quality observation systems have decreased in number over the past 20 years in some regions of the world. Such negative effects are notable in developing countries due to insufficient public funding for operating and maintaining observing networks. The in situ networks remain foundational for monitoring climate variations and change by serving as reference stations, even with the rapid growth of satellite and other remote observations. They are also important to climate and weather simulations as a reference for the accuracy of

remote-sensing observations, and for identifying forecast errors. Local observations such as weather radars are an important part of early warning systems, which need accurate short-term forecasts of convection and other hazards. Various capacity-development projects have attempted to fill these observational gaps in the developing countries, but the success of these efforts has been undermined by the lack of sustainability and continuity of the operations after the expiration of the project period.

On the IT side, mid-range HPC systems, which nowadays are more affordable and accessible, permit effective operations and research. This could allow for a wider range of forecasting centres to operate regional NEWP systems in partnership with global forecasting providers, by enabling demanding computational processes with higher space–time resolution in complex settings.

A significant computational challenge continues to be assimilating the ever-increasing volume and variety of observational data, particularly from satellites.

3.1.1 Observational ecosystem

Availability of observational data is key to reaching the desired model performance, even with the best NEWP model. Thus, discussion about the refinement/

development of future NEWP models should go together with that of future observing capabilities. Several factors of the observational ecosystem need to be considered:

• Overcoming the lack of observational data and data quality issues is critical for continuous improvement.

For example, poor instrumentation, particularly in developing countries, limits the ground-truthing and application of NEWP systems especially at catchment/basin/watershed levels, where most water management decisions are usually made.

• Monitoring the Earth’s surface at high temporal frequency and high spatial resolution will improve the description of kilometre and sub-kilometre scales associated with convective systems, boundary layer processes and new surface types (for example, towns, lakes and rivers). Meeting this observational challenge will be demanding as numerical models move towards convective-permitting scales. Boundary layer observations and also observations in data-sparse regions would advance forecasting considerably.

• The evolution of satellite programmes for operational prediction undertaken by governmental space agencies is stable but takes place over timescales of decades. The development of satellite remote- sensing for the research community has a more rapid response. In parallel, the private sector has started

investing in low-cost technology, often built upon research advances, to build short-lifetime missions (for example, constellations of nanosatellites). The availability, quality, interest and methods to pay for these observations have yet to be evaluated.

Public–private arrangements will be needed for improved coordination of the short- and long-term delivery schedules of these different space-based observations and for identifying possible synergies, especially where the private sector could fill some observational gaps. Efforts should be made to exploit new satellite observations, and to better utilize the data already available. Since many of the advances in operational prediction are built upon refining and improving research breakthroughs, access of the research community to these private sector satellite- based observations is also critical.

• Significant challenges remain in the access to and exploitation of data from observing systems owned and operated by various non-State stakeholders. For example, many underutilized in situ weather stations exist, often used for academic purposes, but with potential to contribute to operational forecasting.

Many municipalities, farms, road agencies and other industries maintain regular observations with their own networks of instruments. Such observations may be of substandard quality compared with those operated by National Meteorological and Hydrological Services (NMHSs), but through sharing arrangements and innovative quality control, they could add significantly to the overall observing ecosystem, especially in remote areas, where operation and maintenance of ground stations poses challenges.

• The growing availability of “non-conventional”

observations will offer major new opportunities for augmenting the classical approaches and filling existing observational data gaps. There is a plethora of such new data, many available as by-products of systems or devices not intended for meteorological or similar purposes. These include: estimating rainfall from attenuation of signals between cell phone towers, commercial surface sensors purchased and deployed by citizens, virtual sensors, “Internet of Things” devices, smartphone sensors and military- grade weather stations. The data provided by these new systems or devices offer unprecedented sources of information, but can also present challenges in terms of observational quality, data access and volumes, and privacy and ethical concerns when data are owned by individuals or commercial companies.

With these concerns addressed appropriately, and with proper quality control, such non-conventional data could deliver observations in sparsely covered domains like urban areas, tropical land surfaces, oceans, the upper atmosphere and polar regions.

International collection and sharing of such weather observations is already happening with websites like the Met Office Weather Observation Website.

However, their systematic use in NEWP should be cautious since the long-term availability and reliability of such data provision cannot be guaranteed.

• Supplementary information based on indigenous and traditional knowledge and citizen science is yet to be explored as a potential source for improved forecasts and insights. However, these forms of information remain challenging across several dimensions, such as frequency and distribution of collection, mapping between epistemological domains and quality control.

These challenges can be addressed only through more systematic and grounded research partnerships.

• Future weather and climate observational data should be interoperable with socioeconomic, biophysical and other data, especially at the local and urban levels, to expand knowledge generation and to provide informative forecasting results to end users.

• Finally, when planning observational ecosystem improvements and optimization, it should be kept in mind that achievements and improvements in NEWP systems have permitted the same global forecast skills to be accomplished utilizing fewer observations, as demonstrated by reforecast experiments based on reanalyses. This allows the opportunity to consider optimal and cost-effective design of future operational observing systems better tailored to the capabilities of the forecasting systems. Furthermore, the skill of NEWP systems often depends more on the ability to properly assimilate existing observations, rather than on adding additional observations. Hence, rigorous forecast sensitivity studies are needed to understand the impact of observational data to inform and prioritize investments in observational and NEWP systems at all space–time scales. As an example, even with the phenomenal impact of the increase in satellite observations for NEWP, in situ observations will always be needed to provide a reference, such as for surface pressure. However, what the optimal investments in such in situ observations are to satisfy all user requirements is still an open question.

3.1.2 High-performance computing ecosystem

The evolution towards running higher-resolution and more complex NEWP systems on tight operational schedules poses significant challenges for HPC and “big data” handling. Computing and data must always be considered together since more sophisticated prediction systems create more diverse and more voluminous output data. Challenges include the following:

• Projects conducted by leading global weather prediction centres, and the climate projection community (for example, the Coupled Model Intercomparison Project (CMIP)), already struggle to afford the sustainable supercomputing infrastructures required for hosting R&D activities and upcoming prediction system upgrades, in terms of capital investment and running operational costs (for example, the cost of electrical power). To overcome these challenges, research organizations are under increasing pressure to find ways to join forces in operating the HPC infrastructure and gain efficiency through resource and cost sharing.

• The main technological breakthroughs linked to HPC are expected from the combined effects of several sources.

In the past, an exponential computing power growth rate was provided by increasing transistor density while maintaining overall power consumption on general- purpose chips. Today, new power-efficient processor technologies (for example, graphics processing units, tensor processing units, field programmable gate arrays and custom application-specific integrated circuits) are increasingly available and necessary to sustain that exponential growth. Their use requires code adaptation to different ways of mapping operations onto processor memory, parallelization and vectorization. It might be that some of the new processors targeting artificial intelligence (AI) will never be effective at solving partial differential equations, and it is necessary to seek radically new approaches, such as emulation by machine learning (ML). The implementation of such adaptation will require enough lead time to be effective and serve the entire community. Furthermore, there is a need to enhance the scope of expertise towards computational sciences in all programmes, which offers potential for attracting new talent and career development.

• As future architectures will be composed of a wider range of different technologies, mathematical methods and algorithms need to adapt so computations can be delegated to those parts of the architecture that deliver optimal performance for each task. Such specialization is not embodied in present-day codes and not delivered by the available compilers and programming standards. A breakthrough can be achieved only by a radical redesign of codes, likely to be carried out by the weather and climate community in partnership with computer scientists and hardware providers. This redesign will ensure the theoretically achievable performance gains are scalable from small to large machines and are transferable to even more advanced and novel technologies in the future without yet another redesign effort.

• The resulting combination of code adaptivity and algorithmic flexibility will require a community-wide effort; again, there are concerns for computing and

data handling. Future HPC workflows are likely to be distributed across specialized units for the heavy computing and data handling tasks, for operating software layers interfacing with observational input and prediction output data, and for providing a flexible and open interface to a large variety of users.

This will be a move away from single centres towards federated infrastructures, components of which will be operated on the cloud.

• Weather and climate data volumes are massive and are growing daily. The old model of storing data in centralized archives – ostensibly part of HPC systems, which likely produce the data – is not capable of scaling up to support the diverse user groups expected to want these data over the coming decade, for two reasons: (a) remote users likely cannot access them and (b) security concerns make it unwieldy to support large user bases on these systems. A solution is needed to make data open and accessible for diverse stakeholders from across the globe. A competitive market of cloud-powered data archives and support infrastructure is likely to be a key enabler.

3.1.3 Changing landscape: advances in infrastructure through public–private engagement

The new technological developments in the infrastructure domain have significantly increased the range of solutions available to the NEWP challenges, many of which are increasingly being offered by private sector stakeholders:

• Innovative technological solutions in observations and monitoring with possible application in weather and climate forecasting will continue to be delivered in part by the private sector. As in telecommunications, more private sector providers are likely to offer observational data services (“observations as a service”) rather than simply selling hardware as in the past. Further engagement of the private sector with affordable initial investments (including more

“start-up” businesses) will be possible due to the availability of low-cost weather stations. The miniaturization of satellites and their instruments also promises cost-effective, flexible and resilient options for augmenting critical components of the global observing system.

• The growing amount of data through private sector investments poses a question on the conditions for the utilization and sharing of such data by other sectors at national and international levels. For some data sets, it is likely that national and international space agencies will start to procure data services (from

the private sector) instead of satellite hardware. To ensure mutual benefits and avoid “data inequalities”, the WMO data policy should evolve to reflect the changing observational ecosystem with its economic foundation, while at the same time preserving the global basic infrastructure delivery as a public good for the benefit of all nations.

• Innovative remote observing platforms and systems (such as autonomous drones, crowdsourced observations or other emerging systems that can probe and measure atmospheric parameters in three dimensions) will have a significant impact on global forecast quality at higher resolutions as well as for extreme events. Understanding the precision and accuracy of these new information sources will be a critical research target to help increase their usefulness.

• The expansion of HPC ecosystems offers opportunities for much more comprehensive and prevalent public–

private partnerships. Indeed, such partnerships are crucial if the supercomputers of the future are to be designed to be best suited to the numerical simulation software used across the weather and climate enterprise.

• Transition to cloud solutions for archiving and computing will be a major trend in the coming decade. This opens another area of collaboration and partnership among sectors, as big data companies will continue to provide cloud services. Many NMHSs and other public sector agencies will gain long-term efficiency by using these services. However, the transfer of responsibility for data handling to those service providers should be based on strong and reliable relationships with guarantees for protection of data and continuity of service over long periods.

• A challenge for rapid advancement of weather and climate forecasting is to find the right balance of investment in remote, in situ and space-based observations. Today’s backbone observing system needs to be maintained with sufficient redundancy to fill potential gaps in case of individual mission failure. But the global observation network already shows significant resilience and coverage; therefore, questions about cost-effectiveness and affordability arise. The still existing data deficits in some regions, mostly due to insufficient public funding (or, in some cases, data sharing policy issues), may be offset through leveraging private sector data sources, which have been growing rapidly over the last decade. The diversity of the new data produced by innovative technologies, including “by-product” data derived from the Internet of Things, are expected to find their place in future data-assimilation systems, with special attention to their accuracy and reliability.