Model Grid Scale Research Articles

Improvement in the Forecasting of Low Visibility over Guizhou, China, Based on a Multi-Variable Deep Learning Model

High-quality visibility forecasting benefits traffic transportation safety, public services, and tourism. For a more accurate forecast of the visibility in the Guizhou region of China, we constructed several visibility forecasting models via progressive refinements in different compositions of input observational variables and the adoption of the Unet architecture to perform hourly visibility forecasts with lead times ranging from 0 to 72 h over Guizhou, China. Three Unet-based visibility forecasting models were constructed according to different inputs of meteorological variables. The model training via multiple observational variables and visibility forecasts of a high-spatiotemporal-resolution numerical weather prediction model (China Meteorological Administration, Guangdong, CMA-GD) produced a higher threat score (TS), which led to substantial improvements for different thresholds of visibility compared to CMA-GD. However, the Unet-based models had a larger bias score (BS) than the CMA-GD model. By introducing the U2net architecture, there was a further improvement in the TS of the model by approximately a factor of two compared to the Unet model, along with a significant reduction in the BS, which enhanced the stability of the model forecast. In particular, the U2net-based model performed the best in terms of the TS below the visibility threshold of 200 m, with a more than eightfold increase over the CMA-GD model. Furthermore, the U2net-based model had some improvements in the TS, BS, and RMSE (root-mean-square error) compared to the LSTM_Attention model. The spatial distribution of the TS showed that the U2net-based model performed better at the model grid scale of 3 km than at the scale of individual weather stations. In summary, the visibility forecasting model based on the U2net algorithm, multiple observational variables, and visibility data from the CMA-GD model performed the best. The compositions of input observational variables were the key factor in improving the deep learning model’s forecasting capability, and these improvements could improve the value of forecasts and support the socioeconomic needs of sectors reliant on visibility forecasting.

Evapotranspiration prediction for European forest sites does not improve with assimilation of in situ soil water content data

Abstract. Land surface models (LSMs) are an important tool for advancing our knowledge of the Earth system. LSMs are constantly improved to represent the various terrestrial processes in more detail. High-quality data, freely available from various observation networks, are being used to improve the prediction of terrestrial states and fluxes of water and energy. To optimize LSMs with observations, data assimilation methods and tools have been developed in the past decades. We apply the coupled Community Land Model version 5 (CLM5) and Parallel Data Assimilation Framework (PDAF) system (CLM5-PDAF) for 13 forest field sites throughout Europe covering different climate zones. The goal of this study is to assimilate in situ soil moisture measurements into CLM5 to improve the modeled evapotranspiration fluxes. The modeled fluxes will be evaluated using the predicted evapotranspiration fluxes with eddy covariance (EC) systems. Most of the sites use point-scale measurements from sensors placed in the ground; however, for three of the forest sites we use soil water content data from cosmic-ray neutron sensors, which have a measurement scale closer to the typical land surface model grid scale and EC footprint. Our results show that while data assimilation reduced the root-mean-square error for soil water content on average by 56 % to 64 %, the root-mean-square error for the evapotranspiration estimation is increased by 4 %. This finding indicates that only improving the soil water content (SWC) estimation of state-of-the-art LSMs such as CLM5 is not sufficient to improve evapotranspiration estimates for forest sites. To improve evapotranspiration estimates, it is also necessary to consider the representation of leaf area index (LAI) in magnitude and timing, as well as uncertainties in water uptake by roots and vegetation parameters.

Modeling Low Intensity Fires: Lessons Learned from 2012 RxCADRE

Coupled fire-atmosphere models are increasingly being used to study low-intensity fires, such as those that are used in prescribed fire applications. Thus, the need arises to evaluate these models for their ability to accurately represent fire spread in marginal burning conditions. In this study, wind and fuel data collected during the Prescribed Fire Combustion and Atmospheric Dynamics Research Experiments (RxCADRE) fire campaign were used to generate initial and boundary conditions for coupled fire-atmosphere simulations. We present a novel method to obtain fuels representation at the model grid scale using a combination of imagery, machine learning, and field sampling. Several methods to generate wind input conditions for the model from eight different anemometer measurements are explored. We find a strong sensitivity of fire outcomes to wind inputs. This result highlights the critical need to include variable wind fields as inputs in modeling marginal fire conditions. This work highlights the complexities of comparing physics-based model results against observations, which are more acute in marginal burning conditions, where stronger sensitivities to local variability in wind and fuels drive fire outcomes.

SERMeQ Model Produces a Realistic Upper Bound on Calving Retreat for 155 Greenland Outlet Glaciers

AbstractThe rate of land ice loss due to iceberg calving is a key source of variability among model projections of the 21st century sea level rise. It is especially challenging to account for mass loss due to iceberg calving in Greenland, where ice drains to the ocean through hundreds of outlet glaciers, many smaller than typical model grid scale. Here, we apply a numerically efficient network flowline model (SERMeQ) forced by surface mass balance to simulate an upper bound on decadal calving retreat of 155 grounded outlet glaciers of the Greenland Ice Sheet—resolving five times as many outlets as was previously possible. We show that the upper bound holds for 91% of glaciers examined and that simulated changes in terminus position correlate with observed changes. SERMeQ can provide a physically consistent constraint on forward projections of the dynamic mass loss from the Greenland Ice Sheet associated with different climate projections.

Importance of geological information for assessing drain flow in a Danish till landscape

AbstractThere has, in recent years, been an increasing interest in developing nutrient load mitigation measures focussing on tile drains. To plan the location of such tile drain measures, it is important to know where in the landscape drain flow is generated and to understand the key factors governing drain flow dynamics. In the present study, we test two approaches to assess spatial patterns in drain flow generation and thereby assess the importance of including geological information. The approaches are the widely used topographical wetness index (TWI), based solely on elevation data, and hydrological models that include the subsurface geology. We set‐up an ensemble of 20 hydrological models based on 20 stochastically generated geological models to predict drain flow dynamics in the clay till Norsminde catchment in Denmark and test the results against TWI. We find that the hydrological models predict observed daily drain flow reasonably well. High drain flow volumes were found in stream valleys and in wetlands and lower drain flow volumes in the more hilly parts of the catchment. In spite of the apparent connection to the landscape, there was no statistically significant correlation between TWI and drain flow at grid scale (100 × 100 m). TWI was therefore not found to be a sufficient index on its own to assess where drain flow is generated, especially in the highlands of the catchment. The geology below 3 m was found to have a large impact on the drain flow, and correlations between sand percentage in the subsurface geology and drain flow volume were found to be statistically significant. Geological uncertainty therefore give rise to uncertainty on simulated drain flow, and this uncertainty was found to be high at the model grid scale but decreasing with increasing scale.

Comparative analysis of low-Earth orbit (TROPOMI) and geostationary (GeoCARB, GEO-CAPE) satellite instruments for constraining methane emissions on fine regional scales: application to the Southeast US

Abstract. We conduct Observing System Simulation Experiments (OSSEs) to compare the ability of future satellite measurements of atmospheric methane columns (TROPOMI, GeoCARB, GEO-CAPE) for constraining methane emissions down to the 25 km scale through inverse analyses. The OSSE uses the GEOS-Chem chemical transport model (0.25∘×0.3125∘ grid resolution) in a 1-week simulation for the Southeast US with 216 emission elements to be optimized through inversion of synthetic satellite observations. Clouds contaminate 73 %–91 % of the viewing scenes depending on pixel size. Comparison of GEOS-Chem to Total Carbon Column Observing Network (TCCON) surface-based methane column observations indicates a model transport error standard deviation of 12 ppb, larger than the instrument errors when aggregated on the 25 km model grid scale, and with a temporal error correlation of 6 h. We find that TROPOMI (7×7 km2 pixels, daily return time) can provide a coarse regional optimization of methane emissions, comparable to results from an aircraft campaign (SEAC4RS), and is highly sensitive to cloud cover. The geostationary instruments can do much better and are less sensitive to cloud cover, reflecting both their finer pixel resolution and more frequent observations. The information content from GeoCARB toward constraining methane emissions increases by 20 %–25 % for each doubling of the GeoCARB measurement frequency. Temporal error correlation in the transport model moderates but does not cancel the benefit of more frequent measurements for geostationary instruments. We find that GeoCARB observing twice a day would provide 70 % of the information from the nominal GEO-CAPE mission preformulated by NASA in response to the Decadal Survey of the US National Research Council.

Input data processing tools for the integrated hydrologic model GSFLOW

Integrated hydrologic modeling (IHM) encompasses a vast number of processes and specifications, variable in time and space, and development of models can be arduous. Model input construction techniques have not been formalized or made easily reproducible. Creating the input files for integrated hydrologic models requires complex GIS processing of raster and vector datasets from various sources. Developing stream network topology that is consistent with the model grid-scale digital elevation model (DEM) is important for robust simulation of surface water and groundwater exchanges. Distribution of meteorological data over the model domain is difficult in complex terrain at the model-grid scale, but is necessary for realistic simulations. As model development requires extensive GIS and computer programming expertise, the use of IHMs has mostly been limited to research groups with available financial, human, and technical resources. Here we present a series of open-source Python scripts that are combined with ESRI ArcGIS to provide a formalized technique for the parameterization and development of inputs for the readily available IHM called GSFLOW. This Python toolkit automates many of the necessary and laborious processes of parameterization, including stream network development, land coverages, and meteorological distribution over the model domain. The final products of the toolkit are PRMS ready Parameter Files, along with several input parameters for a MODFLOW model, including input for the Streamflow Routing Package. A demonstration of the toolkit is provided to illustrate its capabilities.

Quantifying the Effects of Horizontal Grid Length and Parameterized Convection on the Degree of Convective Organization Using a Metric of the Potential for Convective Interaction

Abstract The organization of deep convection and its misrepresentation in many global models is the focus of much current interest. A new method is presented for quantifying convective organization based on the identification of convective objects and subsequent derivation of object numbers, areas, and separation distances to describe the degree of convective organization. These parameters are combined into a “convection organization potential” based on the physical principle of an interaction potential between pairs of convective objects. This technique is applied to simulated and observed fields of outgoing longwave radiation (OLR) over the West African monsoon region using data from Met Office Unified Model simulations and satellite observations made by the Geostationary Earth Radiation Budget (GERB) instrument. The method is evaluated by using it to quantify differences between models with different horizontal grid lengths and representations of convection. Distributions of OLR, precipitation and organization parameters, the diurnal cycle of convection, and relationships between the meteorology in different states of organization are compared. Switching from a configuration with parameterized convection to one that allows the model to resolve convective processes at the model grid scale is the leading-order factor improving some aspects of model performance, while increased model resolution is the dominant factor for others. However, no single model configuration performs best compared to observations, indicating underlying deficiencies in both model scaling and process understanding.

Surface-atmosphere exchange in a box: Making the control volume a suitable representation for in-situ observations

Using the eddy-covariance (EC) method to determine net surface-atmosphere exchange relies on extensive simplifications of the mass balance concept. Among others, it is assumed that the 3-D flux field within a control volume is divergence-free, which is shown to be violated e.g. from large-eddy simulations. To practically evaluate the severity of these assumptions, case studies have monitored the surrounding of an EC tower, so the control volume can be represented more explicitly. Alternatively, diagnostic tests during data processing can be used to subset the EC data for periods that more likely fulfill the underlying assumptions. However, these existing methods are constrained either by their degree of realism, resource demand, temporal coverage, varying spatial representativeness, or combinations thereof.It is hypothesized that these deficiencies can be overcome by using the environmental response function (ERF) technique: Relating flux observations at very high spatio-temporal resolution to meteorological forcings and surface properties, and utilizing the extracted relationships to map a single, regular and stationary control volume explicitly in 3-D space and time. Here, the novel ERF virtual control volume (VCV) concept and its implications are derived, and Xu et al. (this issue) are presenting its first practical application.Initial results show that even from a single EC tower, ERF-VCV reduces advective errors by at least one order of magnitude, and incorporates net low-frequency flux contributions. In the same process tower location bias is treated through attaining a fixed-frame, thus equitable and time-invariant representation of the net surface-atmosphere exchange across a target domain. With regard to the frequently observed non-closure of the surface energy balance, this offers the potential for reconciling “spatial heterogeneity” and “storage term” theories. In extension, ERF promises a rectifying observational operator for unbiased model-data comparison, assimilation, and process representation at model grid scale.

Surface-atmosphere exchange in a box: Space-time resolved storage and net vertical fluxes from tower-based eddy covariance

Systematic bias in eddy-covariance flux measurements are pervasive. These arise both from unmeasured terms such as advection, and sampling bias in representativeness of the footprint for both turbulent and storage fluxes. As a result, the majority of eddy-covariance towers suffer from unaccounted bias when comparing to gridded earth system models and fail to close the surface energy balance. We hypothesize that one cause for these two problems is a mismatch between mass and energy fluxes measured within a time-varying source area and the actual storage and net vertical flux over a presumed “control volume”, a novel concept derived theoretically in Metzger (this issue). Here, we practically implement this theory to estimate the true net surface-atmosphere exchange (NSAE) over such control volume, thus resolving “storage flux” and “vertical advection” issues by applying the environmental response function (ERF) technique to a virtual control volume (VCV). In this method, flux observations are related at high spatio-temporal resolution to meteorological forcings and surface properties within the estimated flux footprint, and these relationships are utilized to map the control volume explicitly in 3-D over space and time. Volume integration then allows, for the first time, retrieval of the NSAE. When ERF was applied to eddy covariance and profile observations in July and August 2014 from the AmeriFlux Park Falls WLEF tower in Wisconsin, USA, heat emission integrated over the target domain increased substantially over the tower observations by +18.2 Wm−2 (+20.6%). Storage flux contributes up to 30% of NSAE at hourly timescale. The systematic uncertainty of ERF-VCV method applied for vertical flux and storage flux is within 15% and 20%, respectively. This systematic uncertainty is effectively corrected in projections. Volume-controlled NSAE provides improvements for mapping unbiased surface-atmosphere exchange for model-data comparison, assimilation and model building at model grid scale. These advances also present a promising direction for reconciling energy balance non-closure.

Accounting for CO2 variability over East Asia with a regional joint inversion system and its preliminary evaluation

A regional surface carbon dioxide (CO2) flux inversion system, the Tan-Tracker-Region, was developed by incorporating an assimilation scheme into the Community Multiscale Air Quality (CMAQ) regional chemical transport model to resolve fine-scale CO2 variability over East Asia. The proper orthogonal decomposition-based ensemble four-dimensional variational data assimilation approach (POD-4DVar) is the core algorithm for the joint assimilation framework, and simultaneous assimilations of CO2 concentrations and surface CO2 fluxes are applied to help reduce the uncertainty in initial CO2 concentrations. A persistence dynamical model was developed to describe the evolution of the surface CO2 fluxes and help avoid the “signal-to-noise” problem; thus, CO2 fluxes could be estimated as a whole at the model grid scale, with better use of observation information. The performance of the regional inversion system was evaluated through a group of single-observation-based observing system simulation experiments (OSSEs). The results of the experiments suggest that a reliable performance of Tan-Tracker-Region is dependent on certain assimilation parameter choices, for example, an optimized window length of approximately 3 h, an ensemble size of approximately 100, and a covariance localization radius of approximately 320 km. This is probably due to the strong diurnal variation and spatial heterogeneity in the fine-scale CMAQ simulation, which could affect the performance of the regional inversion system. In addition, because all observations can be artificially obtained in OSSEs, the performance of Tan-Tracker-Region was further evaluated through different densities of the artificial observation network in different CO2 flux situations. The results indicate that more observation sites would be useful to systematically improve the estimation of CO2 concentration and flux in large areas over the model domain. The work presented here forms a foundation for future research in which a thorough estimation of CO2 flux variability over East Asia could be performed with the regional inversion system.

Derivation of fractional urban signals in better capturing urbanization process

Rapid urbanization has changed land use and urban structure in China and therefore greatly modified land surface properties and land–atmosphere interactions, causing further local and regional climate change. Climate model simulation and urbanization process analysis are usually limited by poor accuracy of coarse-resolution land use/cover products employed in regional climate models. This study sought to identify better urban representation from Landsat images and monitor urban expansion by change detection of spatial patterns and urban fractions in southeastern coastal region of China. We used the improved normalized indices-based method to classify urban and built-up areas from Landsat images in Jiading District, Shanghai. Classification results were evaluated at both the pixel scale and the model grid scale, with overall accuracy of 88% and k coefficient of 0.76. Moreover, urbanization process over the Guangzhou–Foshan–Dongguan area was examined from 2000 to 2009. We aggregated the original results of urban classification data from Landsat images as fractional cover information in 1-km grids. The total fractional urban change in 2000–2005 (10.65%) was approximately three times greater than in 2005–2009 (3.38%). We also compared the fractional cover of urban expansion with the corresponding period of MODIS land cover products. It showed that existing land cover products in models had deviations and could not capture well the underlying conditions and urbanization process. Different fractional covers of urban scenarios were expected to provide better inputs for accurate modeling of critical environmental feedbacks over expanding urban clusters.

Where are the limits of model predictive capabilities?

AbstractDistributed hydrological models can make predictions with much finer spatial resolution than the supporting field data. They will, however, usually not have a predictive capability at model grid scale due to limitations of data availability and uncertainty of model conceptualizations. In previous publications, we have introduced the Representative Elementary Scale (RES) concept as the theoretically minimum scale at which a model with a given conceptualization has a potential for obtaining a predictive accuracy corresponding to a given acceptable accuracy. The new RES concept has similarities to the 25‐year‐old Representative Elementary Area concept, but it differs in the sense that while Representative Elementary Area addresses similarity between subcatchments by sampling within the catchment, RES focuses on effects of data or conceptualization uncertainty by Monte Carlo simulations followed by a scale analysis. In the present paper, we extend and generalize the RES concept to a framework for assessing the minimum scale of potential predictability of a distributed model applicable also for analyses of different model structures and data availabilities. We present three examples with RES analyses and discuss our findings in relation to Beven's alternative blueprint and environmental modeling philosophy from 2002. While Beven here addresses model structural and parameter uncertainties, he does not provide a thorough methodology for assessing to which extent model predictions for variables that are not measured possess opportunities to have meaningful predictive accuracies, or whether this is impossible due to limitations in data and models. This shortcoming is addressed by the RES framework through its analysis of the relationship between aggregation scale of model results and prediction uncertainties and for considering how alternative model structures and alternative data availability affects the results. We suggest that RES analysis should be applied in all modeling studies that aim to use simulation results at spatial scales smaller than the support scale of the calibration data.

Bayesian Analysis of Uncertainty in the GlobCover 2009 Land Cover Product at Climate Model Grid Scale

Land cover data derived from satellites are commonly used to prescribe inputs to models of the land surface. Since such data inevitably contains errors, quantifying how uncertainties in the data affect a model’s output is important. To do so, a spatial distribution of possible land cover values is required to propagate through the model’s simulation. However, at large scales, such as those required for climate models, such spatial modelling can be difficult. Also, computer models often require land cover proportions at sites larger than the original map scale as inputs, and it is the uncertainty in these proportions that this article discusses. This paper describes a Monte Carlo sampling scheme that generates realisations of land cover proportions from the posterior distribution as implied by a Bayesian analysis that combines spatial information in the land cover map and its associated confusion matrix. The technique is computationally simple and has been applied previously to the Land Cover Map 2000 for the region of England and Wales. This article demonstrates the ability of the technique to scale up to large (global) satellite derived land cover maps and reports its application to the GlobCover 2009 data product. The results show that, in general, the GlobCover data possesses only small biases, with the largest belonging to non–vegetated surfaces. In vegetated surfaces, the most prominent area of uncertainty is Southern Africa, which represents a complex heterogeneous landscape. It is also clear from this study that greater resources need to be devoted to the construction of comprehensive confusion matrices.

InSAR Water Vapor Data Assimilation into Mesoscale Model MM5: Technique and Pilot Study

In this study, a technique developed to retrieve integrated water vapor from interferometric synthetic aperture radar (InSAR) data is described, and a three-dimensional variational assimilation experiment of the retrieved precipitable water vapor into the mesoscale weather prediction model MM5 is carried out. The InSAR measurements were available in the framework of the European Space Agency (ESA) project for the “Mitigation of electromagnetic transmission errors induced by atmospheric water vapor effects” (METAWAVE), whose goal was to analyze and possibly predict the phase delay induced by atmospheric water vapor on the spaceborne radar signal. The impact of the assimilation on the model forecast is investigated in terms of temperature, water vapor, wind, and precipitation forecast. Changes in the modeled dynamics and an impact on the precipitation forecast are found. A positive effect on the forecast of the precipitation is found for structures at the model grid scale or larger (1 km), whereas a negative effect is found on convective cells at the subgrid scale that develops within 1 h time intervals. The computation of statistical indices shows that the InSAR assimilation improves the forecast of weak to moderate precipitation ( ${ ).

Influence of wind speed on the global variability of burned fraction: a global fire model’s perspective

Understanding of fire behaviour, especially fire spread, is mostly based on local-scale observations but the same equations are applied in global models on a much coarser scale. Most model formulations include the effect of wind speed with a positive influence on fire spread. Availability of global datasets offers new possibilities to evaluate these approaches based on local-scale observations at the global scale. Here, we analyse the relation between wind speed derived from three datasets and remotely sensed burned fraction (burned area divided by grid cell area) on a climate model grid scale. The bivariate relationship between burned fraction and wind speed is characterised by an initial increase in burned fraction and a decrease in burned fraction for wind speeds higher than 2–3 ms–1. In a multivariate analysis we additionally included the effect of tree cover, precipitation or atmospheric moisture, temperature, vegetation net primary productivity and population density on burned fraction. This analysis confirmed the lack of an increase in burned fraction for high wind speeds on annual and daily time scale. From the observation-based analysis we conclude that a positive response of burned fraction for high wind speed should not be applied in coarse-scale global fire models.

Spatial variability of liquid cloud and rain: observations and microphysical effects

AbstractSpatial variability of liquid cloud water content and rainwater content is analysed from three different observational platforms: in situ measurements from research aircraft, land‐based remote sensing techniques using radar and lidar, and spaceborne remote sensing from CloudSat. The variance is found to increase with spatial scale, but also depends strongly on the cloud or rain fraction regime, with overcast regions containing less variability than broken cloud fields. This variability is shown to lead to large biases, up to a factor of 4, in both the autoconversion and accretion rates estimated at a model grid scale of ≈︁40 km by a typical microphysical parametrization using in‐cloud mean values. A parametrization for the subgrid variability of liquid cloud and rainwater content is developed, based on the observations, which varies with both the grid scale and cloud or rain fraction, and is applicable for all model grid scales. It is then shown that if this parametrization of the variability is analytically incorporated into the autoconversion and accretion rate calculations, the bias is significantly reduced.

Impact of Climate Change on Extreme Rainfall across the United States

Changes in climate may alter extreme rainfall intensity in most regions. In order to incorporate potential future changes, tools for planning and design should be capable of considering nonstationary climate conditions. In this study, potential changes in intensity-duration-frequency (IDF) curves, which are often used for assessment of extreme rainfall events, were explored using climate model historic runs, reanalysis runs, and future climate projections. The concept of the adjustment factor, which represents the general ability of climate models in representing the reanalysis data that is a reasonable estimate of actual gridded historical climate at the model grid scale, was applied to investigate potential impacts of climate change on IDF curves. For most of the study regions, future climate projections suggested an increase in the intensity of extreme storms for a given return period and duration with strong regional variations. The results also revealed the dependence of the adjustment factor values on return period and storm durations as related to the characteristics of extreme rainfalls, as it varied significantly with region, storm duration, and return period.

How Processing Digital Elevation Models Can Affect Simulated Water Budgets

For regional models, the shallow water table surface is often used as a source/sink boundary condition, as model grid scale precludes simulation of the water table aquifer. This approach is appropriate when the water table surface is relatively stationary. Since water table surface maps are not readily available, the elevation of the water table used in model cells is estimated via a two-step process. First, a regression equation is developed using existing land and water table elevations from wells in the area. This equation is then used to predict the water table surface for each model cell using land surface elevation available from digital elevation models (DEM). Two methods of processing DEM for estimating the land surface for each cell are commonly used (value nearest the cell centroid or mean value in the cell). This article demonstrates how these two methods of DEM processing can affect the simulated water budget. For the example presented, approximately 20% more total flow through the aquifer system is simulated if the centroid value rather than the mean value is used. This is due to the one-third greater average ground water gradients associated with the centroid value than the mean value. The results will vary depending on the particular model area topography and cell size. The use of the mean DEM value in each model cell will result in a more conservative water budget and is more appropriate because the model cell water table value should be representative of the entire cell area, not the centroid of the model cell.

Modelling the impacts of weather and climate variability on crop productivity over a large area: A new process-based model development, optimization, and uncertainties analysis

Process-based crop models are increasingly being used to investigate the impacts of weather and climate variability (change) on crop growth and production, especially at a large scale. Crop models that account for the key impact mechanisms of climate variability and are accurate over a large area must be developed. Here, we present a new process-based general Model to capture the Crop–Weather relationship over a Large Area (MCWLA). The MCWLA is optimized and tested for spring maize on the Northeast China Plain and summer maize on the North China Plain, respectively. We apply the Bayesian probability inversion and a Markov chain Monte Carlo (MCMC) technique to the MCWLA to analyze uncertainties in parameter estimation and model prediction and to optimize the model. Ensemble hindcasts (by perturbing model parameters) and deterministic hindcasts (using the optimal parameters set) were carried out and compared with the detrended long-term yields series both at the crop model grid (0.5° × 0.5°) and province scale. Agreement between observed and modelled yield was variable, with correlation coefficients ranging from 0.03 to 0.88 ( p < 0.01) at the model grid scale and from 0.45 to 0.82 ( p < 0.01) at the province scale. Ensemble hindcasts captured significantly the interannual variability in crop yield at all the four investigated provinces from 1985 to 2002. MCWLA includes the process-based representation of the coupled CO 2 and H 2O exchanges; its simulations on crop response to elevated CO 2 concentration agree well with the controlled-environment experiments, suggesting its validity also in future climate. We demonstrate that the MCWLA, together with the Bayesian probability inversion and a MCMC technique, is an effective tool to investigate the impacts of climate variability on crop productivity over a large area, as well as the uncertainties.