Editing IPCC:AR6/WGI/Chapter-10 (section)

=== 10.2.2 Challenges for Regional Climate Change Assessment ===

<div id="h2-12-siblings" class="h2-siblings"></div>

<div id="10.2.2.1" class="h3-container"></div>

<span id="quality-control"></span>
==== 10.2.2.1 Quality Control ====

<div id="h3-8-siblings" class="h3-siblings"></div>

The usefulness of any observational dataset is conditioned by the availability and outcome of a quality control (QC) process. The objective of the QC is to verify that data are representative of the measured variable and to what degree the value could be contaminated by unrelated or conflicting factors ( [[#WMO--2017a|WMO, 2017a]] ). Data quality assessment is key for ensuring that the data are credible and to establish trusted relationships between the data provider and the users ( [[#Nightingale--2019|Nightingale et al., 2019]] ). QC is performed for all relevant global climate datasets (e.g., [[#Menne--2018|Menne et al., 2018]] ). For instance, QC informs users that old reanalysis datasets can be inconsistent in the long term because they assimilated inhomogeneous observations over the reanalyses period ( [[#Kobayashi--2015|Kobayashi et al., 2015]] ). As a consequence, the evaluation against independent observations suggests that reanalyses should not be automatically regarded as climate-quality products for monitoring long-term trends at the regional level ( [[#Manzanas--2014|Manzanas et al., 2014]] ; [[#Torralba--2017|Torralba et al., 2017]] ). QC needs to be systematically carried out by the institutions responsible for handling the data (e.g., [[#Cao--2016b|Cao et al., 2016b]] ).

The QC procedure depends strongly on the specific nature of the dataset. It focuses on aspects such as the correct identification of sensor, time and location, detection of unfeasible or inconsistent data, error estimation, assessment of the adequacy of the uncertainty information and the adequacy of the documentation (e.g., [[#Heaney--2016|Heaney et al., 2016]] ). QC principles also apply to model data ( [[#Tapiador--2017|Tapiador et al., 2017]] ). An important piece of information provided is the representativeness error ( [[#10.2.1.1|Section 10.2.1.1]] ; [[#Gervais--2014|Gervais et al., 2014]] ). When problems in the data representativeness are identified, observational datasets are provided with a quality mask ( [[#Contractor--2020|Contractor et al., 2020]] ), or the problematic dataare either removed or corrected ( [[#Ashcroft--2018|Ashcroft et al., 2018]] ). These are factors often taken into account in constructing regional climate information ( [[#Kotlarski--2019|Kotlarski et al., 2019]] ).

Quality-controlled data are now produced widely at the regional level, as in the case of sub-daily precipitation records in the United Kingdom ( [[#Blenkinsop--2017|Blenkinsop et al., 2017]] ) and the USA ( [[#Nelson--2016|Nelson et al., 2016]] ). However, many more datasets and variables lack the same level of scrutiny ( [[#Alexander--2016|Alexander, 2016]] ). Quality-controlled, high-resolution observational datasets are especially needed at regional and local scales to assess models as their resolution increases ( [[#Di%20Luca--2016|Di Luca et al., 2016]] ; [[#Zittis--2017|Zittis and Hadjinicolaou, 2017]] ), although the awareness and appropriate use of the QC information is challenging ( [[#Tapiador--2017|Tapiador et al., 2017]] ) when generating regional climate information ( ''high confidence'' ).

<div id="10.2.2.2" class="h3-container"></div>

<span id="homogenization"></span>
==== 10.2.2.2 Homogenization ====

<div id="h3-9-siblings" class="h3-siblings"></div>

Homogenization aims to make data spatially and temporally ‘homogeneous’. Changes in a homogeneous time series are solely due to large-scale climatic changes (whether forced or due to internal variability). Station data are influenced by factors that act at regional scales, from the mesoscale and local scale down to the microscale ( [[#WMO--2019|WMO, 2019]] ). Station time series contain inhomogeneities such as artificial jumps or trends, which hamper assessments of regional long-term trends. Typical reasons for this are the urbanization of a station’s surroundings, which can lead to warming ( [[#Hamdi--2010|Hamdi, 2010]] ; [[#Hansen--2010|Hansen et al., 2010]] ; [[#Adachi--2012|Adachi et al., 2012]] ; [[#Jones--2016|Jones, 2016]] ; Y. [[#Sun--2016|]] [[#Sun--2016|Sun et al., 2016]] ), or relocations outside of the urban area, which could lead to cooling ( [[#Tuomenvirta--2001|Tuomenvirta, 2001]] ; [[#Yan--2010|Yan et al., 2010]] ; [[#Xu--2013|Xu et al., 2013]] ; [[#Dienst--2017|Dienst et al., 2017]] , 2019). Another potential source of inhomogeneity is a change in measurement methods that affect most instruments of an observational network over a limited time span, such as the transition to Stevenson screens ( [[#Parker--1994|Parker, 1994]] ; [[#Böhm--2010|Böhm et al., 2010]] ; [[#Brunet--2011|Brunet et al., 2011]] ; [[#Auchmann--2012|Auchmann and Brönnimann, 2012]] ) or to automatic weather stations ( [[#WMO--2017b|WMO, 2017b]] ).

The above examples have been selected as they are present in many stations and without going through homogenization they could potentially have influenced global land warming estimates ( [[IPCC:Wg1:Chapter:Chapter-1#1.5.1|Section 1.5.1]] ). Single-break inhomogeneities tend to have a magnitude comparable to global climate change ( [[#Tuomenvirta--2001|Tuomenvirta, 2001]] ; [[#Venema--2012|Venema et al., 2012]] ) and are thus important for analyses of small regions. Also station records in national networks often have similar changes, making them important for national climate change estimates, but many of these influences are averaged out at the global scale ( [[#Jones--2016|Jones, 2016]] ).

The main approach to reduce the influence of inhomogeneities in station observations is statistical homogenization by comparing the data from a candidate station with those of neighbouring reference stations in conjunction with the use of metadata ( [[#Trewin--2010|Trewin, 2010]] ). This is a challenging task because both reference and candidate records normally have multiple inhomogeneities. Three challenges should be considered. First, most of our understanding of statistical homogenization stems from the homogenization of temperature observations from dense networks. Recent studies suggest that our ability to remove biases quickly diminishes for sparse networks ( [[#Gubler--2017|Gubler et al., 2017]] ; [[#Lindau--2018a|Lindau and Venema, 2018a]] ). This affects early instrumental data and observations that are not strongly correlated between stations, such as wind and humidity ( [[#Chimani--2018|Chimani et al., 2018]] ).

Second, in addition to systematic errors, homogenized data also suffer from random errors, introduced by the homogenization process. These errors are largest at the station level but are also present in network-averaged signals ( [[#Lindau--2018b|Lindau and Venema, 2018b]] ). These errors are determined by the break time series, as well as the noise series and the performance of the homogenization method, are spatially correlated, and have an impact on activities such as interpolation and statistical post-processing of climate simulations ( [[#10.2.3.1|Section 10.2.3.1]] ). Third, the above discussion pertains to the homogenization of monthly and annual means. Homogenization of daily variability around the mean is more difficult. For daily data, specific correction methods are used ( [[#Della-Marta--2006|Della-Marta and Wanner, 2006]] ; [[#Mestre--2011|Mestre et al., 2011]] ; [[#Trewin--2013|Trewin, 2013]] ; [[#Zhou--2021|]] [[#Zhou--2021|C. Zhou et al., 2021]] ) that are able to improve the homogeneity of test cases, although recent independent validation efforts were not able to show much improvement ( [[#Chimani--2018|Chimani et al., 2018]] ). The difference with homogenization methods of monthly and annual means may stem from assumptions on the nature of inhomogeneities for daily data, which are not yet well understood ( [[#Chimani--2018|Chimani et al., 2018]] ).

It is ''virtually certain'' that statistical homogenization methods reduce the uncertainties of long-term estimates. Considering a decomposition of the long-term warming error into a bias and a noise uncertainty around the bias, the (trend) bias especially will be reduced, but also most of the noise uncertainty. This conclusion is based on our understanding of the causes of inhomogeneities and their statistical nature combined with the design principles of statistical homogenization methods, as well as on analytical ( [[#Lindau--2018b|Lindau and Venema, 2018b]] ), numerical ( [[#Venema--2012|Venema et al., 2012]] ; [[#Williams--2012|Williams et al., 2012]] ) and empirical validation studies ( [[#Hausfather--2016|Hausfather et al., 2016]] ; [[#Gubler--2017|Gubler et al., 2017]] ; [[#Killick--2020|Killick et al., 2020]] ).

The above section is about the homogenization of land stations. Satellite data has its own issues and methods for homogenization ( [[#Brinckmann--2013|Brinckmann et al., 2013]] ; [[#Huang--2015|Huang et al., 2015]] ; [[#Brogniez--2016|Brogniez et al., 2016]] ). The homogenization of radiosonde data and land station data use similar methods ( [[#Haimberger--2012|Haimberger et al., 2012]] ; [[#Jovanovic--2017|Jovanovic et al., 2017]] ).

<div id="10.2.2.3" class="h3-container"></div>

<span id="data-scarcity"></span>
==== 10.2.2.3 Data Scarcity ====

<div id="h3-10-siblings" class="h3-siblings"></div>

Data scarcity arises largely due to the lack of maintenance of observing stations, inaccessibility of the data held in national networks, and uneven spatial distribution of stations that lead to a low density in many regions. This is particularly problematic when trying to assess regional climate change, for which a high density of observational data is desirable. Although in several regions numerous stations provide (monthly) data covering more than 100 years for both temperature and precipitation ( [[#GCOS--2015|GCOS, 2015]] ), large areas of the world remain sparsely covered. The post-1990 decline in the total number of stations contributing to the Global Precipitation Climatology Centre (GPCC) monthly product may be related to delays in data acquisition and not paucity of data ( [[#GCOS--2015|GCOS, 2015]] ). This is because GPCC is the result of a single time scale, single Essential Climate Variable (ECV) and single data collection centre. There is no similar drop-off of the rainfall reports in the Global Historical Climatology Network Daily database (GHCNd, [[#Menne--2012|Menne et al., 2012]] ) or the Integrated Surface Database (ISD) at the sub-daily time scale.

[[#Kidd--2017|Kidd et al. (2017)]] made some assumptions about GPCC-available gauges and indicated that only 1.6% of Earth’s surface lies within 10 km of a rain gauge, and many areas of the world are beyond 100 km from the nearest rain gauge. Data scarcity is especially critical over Africa ( [[#Nikulin--2012|Nikulin et al., 2012]] ; [[#Dike--2018|Dike et al., 2018]] ) but the apparent data scarcity could be due to reasons other than actual paucity of data, as stated earlier. For instance, over South Africa, the number of weather stations collecting daily temperature used in the fourth version of the Climatic Research Unit Temperature dataset (CRUTEM4, [[#Osborn--2014|Osborn and Jones, 2014]] ) has significantly declined since 1980 ( [[#Archer--2018|Archer et al., 2018]] ). Although CRUTEM4 has now been replaced by CRUTEM5 ( [[#Osborn--2021|Osborn et al., 2021]] ) it has yet to take advantage of the significant international efforts to curate and make available improved global holdings ( [[#Rennie--2014|Rennie et al., 2014]] ) which increased the global available station count for monthly mean temperatures. This includes additional stations from many African countries. The apparent decline in stations since the 1980s could also be due to countries not contributing their data to the SYNOP/CLIMAT networks for reasons other than having non-operational stations.

Even in Europe, precipitation station density in the widely used E-OBS gridded dataset varies largely in space and time across regions ( [[#Prein--2017|Prein and Gobiet, 2017]] ). This variability is partly due to the reluctance of some data owners to share their data with an international effort. Regardless of the reason, low station density is a major source of uncertainty ( [[#Isotta--2015|Isotta et al., 2015]] ). [[#Kirchengast--2014|Kirchengast et al. (2014)]] and [[#O--2019|O and Foelsche (2019)]] found that at least 2 to 5 (12) stations are required for capturing the area-averaged precipitation amount of heavy summer precipitation events on a daily (hourly) basis with a normalized root-mean-square error of less than 20%. Like the E-OBS dataset, gridded daily temperature and precipitation datasets are being developed for other regions of the world. Examples include south-east Asia (SA-OBS, [[#Van%20den%20Besselaar--2017|Van den Besselaar et al., 2017]] ), and Latin America and West Africa (ICA&amp;D, Van den [[#Besselaar--2015|Besselaar et al., 2015]] ). Despite the uneven distribution of stations in space and time, the value in these initiatives is illustrated by the large number of studies in which the data product is used. This is the case, for instance, in the work of [[#Condom--2020|Condom et al. (2020)]] over the Andes, a region with prominent data scarcity, and the African Monsoon Multidisciplinary Analyses project over West Africa (AMMA; e.g., [[#Lebel--2009|Lebel and Ali, 2009]] ). There have been efforts to reduce data scarcity through initiatives such as the International Surface Temperature Initiative (ISTI, [[#Thorne--2011|Thorne et al., 2011]] ), GHCND, and the Expanding Met Office Hadley Centre ISD with quality-controlled, sub-daily station data from 1931 (HadISD, [[#Dunn--2016|Dunn et al., 2016]] ).

Data scarcity arising from changing coverage in observation station networks results in substantial problems for climate monitoring (e.g., trend analysis of extreme events requires high temporal and spatial resolutions) or model evaluation ( [[#10.3.3.1|Section 10.3.3.1]] ). It is ''virtually certain'' that the scarcity and decline of observational availability in some regions (but not necessarily globally), increase the uncertainty of the long-term global temperature and precipitation estimates. As an example, [[#Lin--2019|Lin and Huybers (2019)]] found that changes in the number of rain gauges after 1975 resulted in spurious trends in extremes of Indian rainfall in a 0.25° gridded dataset spanning the 20th century. In fact, the number of stations used to construct the gridded dataset dropped by half after 1990, leading to inhomogeneity and spurious trends ( [[#10.6.3|Section 10.6.3]] ). Over the southern part of the Mediterranean, which is an area sparsely covered by meteorological stations, data scarcity can lead to large uncertainties in the different gridded datasets and strongly affect model evaluation ( [[#10.6.4|Section 10.6.4]] ). Satellite observations can compensate the ground-based precipitation radar data sparsity to prevent an oversight of significant climate change signals ( [[#Yokoyama--2019|Yokoyama et al., 2019]] ).

There are techniques for estimating and reconstructing missing data. The methods depend on the variable of interest, the temporal resolution (e.g., daily or monthly), and the type of climate (wet or dry), among others. There has been very little evaluation of the performance of classical and data mining methods (e.g., [[#Sattari--2017|Sattari et al., 2017]] ). The classical methods include the arithmetic mean, inverse distance weighting method, multiple regression analysis, multiple imputation, and single best estimator, while the data-mining methods include multilayer perceptron artificial neural network, support vector machine, adaptive neuro-fuzzy inference system, gene expression programming method, and K-nearest neighbour. Crowd-sourced data (individuals contribute their own data points to create a dataset for others to use) could play a role in minimizing data scarcity ( [[#10.2.4|Section 10.2.4]] ).

<div id="10.2.2.4" class="h3-container"></div>

<span id="gridding"></span>
==== 10.2.2.4 Gridding ====

<div id="h3-11-siblings" class="h3-siblings"></div>

Derived gridded datasets require merging data from different sources of observations and/or reanalysis data on a regular grid ( [[#10.2.1.2|Section 10.2.1.2]] ; e.g., [[#Xie--1997|Xie and Arkin, 1997]] ). However, in situ observations are distributed irregularly, especially over sparsely populated areas. This leads to an interpolation challenge. Gridded products of climate variables, including temperature and precipitation, are strongly affected ( ''high confidence'' ) by the interpolation method over complex orography and data scarce regions ( [[#Hofstra--2008|Hofstra et al., 2008]] ; [[#Herrera--2016|Herrera et al., 2016]] ).

There are two main approaches to produce gridded datasets: (i) based on in situ observations only, and (ii) combining in situ observations with remote-sensing data and/or reanalysis data. The first approach has been widely employed in regions with high station density using interpolation techniques, such as inverse-distance weighting, optimal interpolation, and kriging ( [[#Chen--2008|Chen et al., 2008]] ; [[#Haylock--2008|Haylock et al., 2008]] ; [[#Frei--2014|Frei, 2014]] ; [[#Isotta--2014|Isotta et al., 2014]] ; Masson and [[#Frei--2014|Frei, 2014]] ; [[#Hiebl--2016|Hiebl and Frei, 2016]] ; [[#Nguyen-Xuan--2016|Nguyen-Xuan et al., 2016]] ). The second approach has been mainly applied in data-sparse regions with low station density, using simple bias adjustment, quantile mapping, and kriging techniques with in situ observations, remote-sensing and reanalysis data ( [[#Cheema--2012|Cheema and Bastiaanssen, 2012]] ; [[#Erdin--2012|Erdin et al., 2012]] ; Dinku et al., 2014; [[#Abera--2016|Abera et al., 2016]] ; [[#Krähenmann--2018|Krähenmann et al., 2018]] ).

Gridding of station data is affected by uncertainties stemming from measurement errors, inhomogeneities, the distribution of the underlying stations and the interpolation error, with station density being the dominant factor ( [[#Herrera--2019|Herrera et al., 2019]] ). Uncertainty due to interpolation is typically small for temperature but substantial for precipitation and its derivatives, such as drought indices ( [[#Chubb--2015|Chubb et al., 2015]] ; [[#Hellwig--2018|Hellwig et al., 2018]] ). The largest uncertainties typically occur in sparsely sampled mountain areas ( [[#10.2.2.5|Section 10.2.2.5]] ). Interpolation generally give rise to smoothing effects, such as low variability of the derived dataset with respect to the in situ observations ( [[#Chen--2019|Chen et al., 2019]] ). As a result, the effective resolution of gridded data is typically much lower than its nominal resolution. For instance, a 5 km gridded precipitation dataset for the European Alps has an effective resolution of about 10 to 25 km ( [[#Isotta--2014|Isotta et al., 2014]] ). In an example for precipitation in Spain, the effective resolution converged to the nominal resolution only when at least 6 to 7 stations were inside the gridcell ( [[#Herrera--2019|Herrera et al., 2019]] ). To account for the smoothing errors, new stochastic ensemble observation datasets have been introduced ( [[#Von%20Clarmann--2014|Von Clarmann, 2014]] ).

<div id="10.2.2.5" class="h3-container"></div>

<span id="observations-in-mountain-areas"></span>
==== 10.2.2.5 Observations in Mountain Areas ====

<div id="h3-12-siblings" class="h3-siblings"></div>

Spatiotemporal variability of meteorological parameters observed over mountainous areas is often large, indicating strong control exerted by local topography on meteorological parameters ( [[#Gultepe--2014|Gultepe et al., 2014]] ). Difficult access, harsh climatic conditions as well as instrumental issues make meteorological measurements extremely challenging at higher elevations ( [[#Azam--2018|Azam et al., 2018]] ; [[#Beniston--2018|Beniston et al., 2018]] ). Measurements of wind speed, temperature, relative humidity and radiative fluxes are critical for climate model evaluation, but difficult to handle due to their point-scale representativeness and small-scale spatiotemporal variability over mountainous terrain, and often need adjustment ( [[#Gultepe--2015|Gultepe, 2015]] ). High-altitude (&gt;3000 metres) permanent meteorological stations are limited and current knowledge is mainly based on valley-bottom or low-elevation meteorological stations ( [[#Qin--2009|Qin et al., 2009]] ; [[#Lawrimore--2011|Lawrimore et al., 2011]] ; [[#Gultepe--2015|Gultepe, 2015]] ; [[#Condom--2020|Condom et al., 2020]] ), which, generally do not represent the higher elevation climate ( [[#Immerzeel--2015|Immerzeel et al., 2015]] ; [[#Shea--2015|Shea et al., 2015]] ).

Measuring precipitation amounts, especially of solid precipitation, in mountainous areas is particularly challenging due to the presence of orographic barriers, strong vertical and horizontal precipitation rate variability, and the difficulty in finding representative sites for precipitation measurements ( [[#Barry--2012|Barry, 2012]] ). However, the precipitation amounts can be indirectly estimated by the observed point mass balances at glacier accumulation areas representing net snow accumulation ( [[#Haimberger--2012|Haimberger et al., 2012]] ; [[#Immerzeel--2015|Immerzeel et al., 2015]] ; [[#Sakai--2015|Sakai et al., 2015]] ; [[#Azam--2018|Azam et al., 2018]] ). There is ''very high confidence'' that precipitation measurements, especially solid precipitation, in mountainous areas are strongly affected by the gauge location and setup. Precipitation measurements are also affected by the type of measurement method, presence/absence of shielding, presence/absence of a heating system and operating meteorological conditions ( [[#Nitu--2018|Nitu et al., 2018]] ). Solid precipitation measurements may have errors ranging from 20% to 50%, largely due to under-catch in windy, icing and riming conditions ( [[#Rasmussen--2012|Rasmussen et al., 2012]] ), and therefore require corrections by applying transfer functions developed mainly from collected wind speed and temperature data ( [[#Kochendorfer--2017|Kochendorfer et al., 2017]] ). The latest Solid Precipitation Intercomparison Experiment (SPICE) report recommends measurements of wind speed, wind direction and temperature as the minimum standard ancillary data for solid precipitation monitoring ( [[#Nitu--2018|Nitu et al., 2018]] ).

Recent advances in remote-sensing methods provide an alternative, but they also have limitations over mountainous areas. Different versions of the Tropical Rainfall Measuring Mission (TRMM) products were found to perform differently over mountainous areas ( [[#Zulkafli--2014|Zulkafli et al., 2014]] ). Orographic heavy rainfall associated with Typhoon Morakot in 2009 was severely underestimated in all microwave products including TRMM 3B42 ( [[#Shige--2013|Shige et al., 2013]] ). The underestimation has been mitigated in the Global Satellite Mapping of Precipitation (GSMaP) product by considering the orographic effects ( [[#Shige--2013|Shige et al., 2013]] ). Studies have suggested a high accuracy of passive optical satellite (e.g., MODIS, Landsat) snow products under clear skies when compared with the field observations. However, cloud masking and sub-pixel cloud heterogeneity in these snow-cover products considerably restrict their applications ( [[#Kahn--2011|Kahn et al., 2011]] ; [[#Brun--2015|Brun et al., 2015]] ; [[#Tang--2017|Tang et al., 2017]] ; [[#Stillinger--2019|Stillinger et al., 2019]] ). Gridded datasets (e.g., CRU, GPCC Full Data Product, GPCC Monitoring Product, ERA-Interim, ERA5, ERA5-land, MERRA-2, MERRA-2 bias adjusted, PERSIANN-CDR) are of paramount importance, yet they often lack enough in situ observations to improve the temporal and spatial distribution of meteorological parameters over complex mountain terrain ( [[#Zandler--2019|Zandler et al., 2019]] ).

<div id="10.2.2.6" class="h3-container"></div>

<span id="structural-uncertainty"></span>
==== 10.2.2.6 Structural Uncertainty ====

<div id="h3-13-siblings" class="h3-siblings"></div>

Beyond climate monitoring, the quality and availability of multiple observational reference datasets play a central role in model evaluation. In fact, when using observations for model evaluation, there are multiple examples where inter-observational uncertainty is as large as the inter-model variability. This has been shown for various aspects of the Indian monsoon ( [[#10.6.3|Section 10.6.3]] ; [[#Collins--2013a|Collins et al., 2013a]] ) and for precipitation uncertainties over Africa ( [[#10.6.4|Section 10.6.4]] ; [[#Nikulin--2012|Nikulin et al., 2012]] ; [[#Sylla--2013|Sylla et al., 2013]] ; [[#Dosio--2015|Dosio et al., 2015]] ; [[#Bador--2020|Bador et al., 2020]] ) and Europe ( [[#Prein--2017|Prein and Gobiet, 2017]] ). [[#Kotlarski--2019|Kotlarski et al. (2019)]] compared three high-resolution observational temperature and precipitation datasets (E-OBS, a compilation of national/regional high-resolution gridded datasets, and the EURO4M-MESAN 0.22° reanalysis based on a high-resolution limited-area model) with five EURO-CORDEX RCMs driven by ERA-Interim. Generally, the differences between RCMs are larger than those between observation datasets, but for individual regions and performance metrics, observational uncertainty can dominate. They also showed that the choice of reference dataset can have an influence on the RCM performance score. Over the high mountain Asia region and East Asia, differences among gridded precipitation datasets can generate significant uncertainties in deriving precipitation characteristics (J. [[#Kim--2015|]] [[#Kim--2015|Kim et al., 2015]] ; [[#Kim--2016|Kim and Park, 2016]] ; [[#Guo--2017|Guo et al., 2017]] ). Over western North America, observational uncertainty induces differences in multi-decadal precipitation trends ( [[#Lehner--2018|Lehner et al., 2018]] ). Taking a very different perspective, the agreement between model simulations may be used to estimate the uncertainty and quality of observations ( [[#Massonnet--2016|Massonnet et al., 2016]] ). There is ''high confidence'' that an ensemble of multiple observational references at a regional scale is fundamental for model performance assessment. The uncertainties vary according to region, season, and statistical properties (Cross-Chapter Box 10.2).

<div id="10.2.3" class="h2-container"></div>

<span id="other-uses-of-observations-at-regional-scale"></span>