Forecasting of the urban weather and climate is of great importance as our cities become more populated and considering the combined effects of global warming and local land use changes which make urban inhabitants more vulnerable to e.g. heat waves and flash floods. In meso/global scale models, urban parameterization schemes are used to represent the urban effects. However, these schemes require a large set of input parameters related to urban morphological and thermal properties. Obtaining all these parameters through direct measurements are usually not feasible.  A number of studies have reported on parameter estimation and sensitivity analysis to adjust and determine the most influential parameters for land surface schemes in non-urban areas. Similar work for urban areas is scarce, in particular studies on urban parameterization schemes in tropical cities have so far not been reported.

In order to address above issues, the town energy balance (TEB) urban parameterization scheme (part of the SURFEX land surface modeling system) was subjected to a sensitivity and optimization/parameter estimation experiment at a suburban site in, tropical Singapore. The sensitivity analysis was carried out as a screening test to identify the most sensitive or influential parameters. Thereafter, an optimization/parameter estimation experiment was performed to calibrate the input parameter. The sensitivity experiment was based on the “improved Sobol’s global variance decomposition method” . The analysis showed that parameters related to road, roof and soil moisture have significant influence on the performance of the model.   The optimization/parameter estimation experiment was performed using the AMALGM (a multi-algorithm genetically adaptive multi-objective method) evolutionary algorithm. The experiment showed a remarkable improvement compared to the simulations using the default parameter set. The calibrated parameters from this optimization experiment can be used for further model validation studies to identify inherent deficiencies in model physics.

Models are an essential tool for studying how our surroundings influence us and how we, intentionally or inadvertently, influence our surroundings. The Surface Urban Energy and Water balance Scheme (SUEWS) uses a basic meteorological forcing dataset and information about the surface cover to model components of the energy and water balance. The model was initially developed based on studies in North America and is now being run for multiple locations around the world. Here, we evaluate the model at two locations in the UK.

A network of micrometeorological observations exists across London, enabling comparisons between the city centre and suburbs. The central London study site is one of the most highly urbanised and densely populated to date. 120 km to the west is the typical suburban town of Swindon. At both of these locations, extensive observational datasets spanning several years have been collected, and work has been undertaken to classify the surface characteristics. However, as detailed land cover and socio-economic information may not always be available, we consider the impact on model performance of using only easily accessible data to provide the required inputs. SUEWS is evaluated against observations of energy and water balance components (including turbulent heat fluxes from eddy covariance and scintillometry techniques). SUEWS estimates evaporation using an adapted Penman-Monteith formulation with a variable surface conductance. Analysis of observed surface conductances suggests adjustments to improve model performance.

CO2 fluxes, closely linked to the surface conductance, are also examined. The central London and suburban Swindon sites behave differently, in terms of both the magnitude and temporal variability of CO2 exchanges. These differences are almost entirely a result of land use and land cover, and associated patterns of human behaviour. Simple models based on anthropogenic emissions inventories provide an indication of the magnitude of the CO2 release, however, at the suburban site vegetation plays an important role in CO2 uptake and must be incorporated too.

With improved modelling capability, the exposure of the population to risks such as thermal stress or flooding can be better estimated. Having validated the model, the impact of policy decisions and future climate scenarios on the wellbeing of the citizens can be assessed.

An offline comparison of three urban land-surface models (CLMU, SUEWS and SURFEX) was undertaken in Helsinki, Finland. The modeled net all-wave radiation, turbulent fluxes of sensible and latent heat and snow depth were compared against observations at two sites. One of the sites was highly built-up European city center with less than 10% of vegetation cover (Torni) whereas the other was semi-urban (Kumpula) with the plan area fraction of vegetation being over 50%. The model evaluation was made for 2012 (following a 6 month spin-up period) with particularly focusing on the seasonality, especially in snow which is a frequent feature of cold climate cities.

All models simulated the net all-wave radiation well and the largest uncertainties were related to the snow-melting period in spring when the fraction of snow on surfaces causes a bias to the outgoing short- and long-wave radiation. The largest uncertainties in the sensible heat flux seem to relate to the estimation of surface temperatures. Similarly to previous studies, the latent heat flux performance was most problematic for all models with a clear underestimation at both sites particularly in summer. Energy partitioning of the turbulent fluxes was better during the growing season than outside it.

All models simulated the snow depth well. However SUEWS and SURFEX delayed the complete snowmelt for Torni (> ten days) longer than for a vegetated surface. This had only a minor impact on the turbulent fluxes given the small fraction of vegetated surfaces at the site. No models outperformed the others, but rather the performances were season, site and flux dependent.

Atmospheric stability, an important parameter in applications like air quality forecasts, were compared against observations. Winter-time stability classes varied between the models. However, they were better simulated at the suburban site than at downtown. There, CLMU is unable to simulate stable atmosphere whereas SUEWS and SURFEX simulate more stable and neutral cases than the observations indicate. This emphasizes a need for correct description of the storage and anthropogenic heat fluxes.

Validation of an urban micro-climate estimation model requires the acquisition of a significant amount of accurate meteorological data. The BUBBLE experimental effort is undoubtedly one of the most significant scientific contribution in the field of urban climate data acquisition. For this reason, we decided to employ this dataset to demonstrate with which precision urban temperature can be approximated in a cold and humid environment using a lumped parameter urban canopy model coupled with an EnergyPlus model. Based on measurements of the Sperrstrasse (Basel, Switzerland) from November to December 2001, we calculated the frequency distribution and the average diurnal cycle of temperature. The similarity between these statistical outcomes and the ones assessed from the coupled scheme was computed in terms of Kolmogorov-Smirnov distance between two non-parametric distributions and root mean square error between two average hourly cycles. As a result, the coupled scheme estimates the urban temperature frequency distribution and average diurnal cycle of the Sperrstrasse with an accuracy of 45 in terms of Kolmogorov-Smirnov distance and 1.8 degrees Celsius in terms of root mean square error, respectively. A comparison between these results and the ones obtained in a complete different climatic environment (i.e. Masdar City, UAE) will be developed in a future study.

The importance of anthropogenic heat flux to the Urban Heat Island effect is recognised in literature, but understanding the true nature of anthropogenic influence on urban climatology is generally limited by a static representation of these sources when applied in urban climate models. Capturing spatial and temporal variability is often limited to use of daily profiles of behaviour that can be represented as some function of local meteorological conditions. This approach being a limited (if at all) representation of dynamic response to localised (spatial and temporal) conditions, with no way of determining how localised conditions may affect anthropogenic heat flux patterns. For example, flooding of roads causing a diversion in traffic that leads to a changed spatial pattern in transport related heat flux, and a possible change in overall heat flux due to increased traffic congestion. This type of human response to weather conditions seems intuitively to be a system feedback, but the significance of these anthropogenic responses to urban climate have not been robustly investigated.

To investigate the relevance of human response to climate conditions, SUEWS (Surface Urban Energy and Water Balance Scheme) has been developed further such that the anthropogenic heat flux component can be linked to an external model that calculates critical parameter values such as population density, metabolic rate, and activity type (commercial, residential, transport). To capture these values the movement of people and their response to local environmental conditions is modelled using an agent-based modelling approach. The decision-making behaviour of modelled agents will be dependent on their classification (e.g. groups of people classified by work, age, or economic status), which is yet to be fully explored. This paper presents a comparison between existing anthropogenic heat flux models and the early stage development of a more dynamically responsive model: the new approach is shown to remain representative in orders of magnitude of heat flux, whilst daily and annual profiles differ to existing model output.