Forecasting climate variables using a mixed-effect state-space model

Several statistical downscaling models have been developed in the past couple of decades to assess the hydrologic impacts of climate change by projecting the station-scale hydrological variables from large-scale atmospheric variables simulated by general circulation models (GCMs). This paper presents and compares different statistical downscaling models that use multiple linear regression (MLR), positive coefficient regression (PCR), stepwise regression (SR), and support vector machine (SVM) techniques for estimating monthly rainfall amounts in the state of Florida. Mean sea level pressure, air temperature, geopotential height, specific humidity, U wind, and V wind are used as the explanatory variables/predictors in the downscaling models. Data for these variables are obtained from the National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis dataset and the Canadian Centre for Climate Modelling and Analysis (CCCma) Coupled Global Climate Model, version 3 (CGCM3) GCM simulations. The principal component analysis (PCA) and fuzzy c-means clustering method (FCM) are used as part of downscaling model to reduce the dimensionality of the dataset and identify the clusters in the data, respectively. Evaluation of the performances of the models using different error and statistical measures indicates that the SVM-based model performed better than all the other models in reproducing most monthly rainfall statistics at 18 sites. Output from the third-generation CGCM3 GCM for the A1B scenario was used for future projections. For the projection period 2001–10, MLR was used to relate variables at the GCM and NCEP grid scales. Use of MLR in linking the predictor variables at the GCM and NCEP grid scales yielded better reproduction of monthly rainfall statistics at most of the stations (12 out of 18) compared to those by spatial interpolation technique used in earlier studies.

Empirical-statistical downscaling (ESD) models use statistical relationships to infer local climate information from large-scale climate information produced by global climate models (GCMs), as an alternative to the dynamical downscaling provided by regional climate models (RCMs). Among various statistical downscaling approaches, the nonlinear methods are mainly used to construct downscaling models for local variables that strongly deviate from linearity and normality, such as daily precipitation. These approaches are also appropriate to handle downscaling of extreme rainfall. There are nonlinear downscaling techniques of various complexities. The simplest one is represented by the analog method that originated in the late 1960s from the need to obtain local details of short-term weather forecasting for various variables (air temperature, precipitation, wind, etc.). Its first application as a statistical downscaling approach in climate science was carried out in the late 1990s. More sophisticated statistical downscaling models have been developed based on a wide range of nonlinear functions. Among them, the artificial neural network (ANN) was the first nonlinear regression–type method used as a statistical downscaling technique in climate science in the late 1990s. The ANN was inspired by the human brain, and it was used early in artificial intelligence and robotics. The impressive development of machine learning algorithms that can automatically extract information from a vast amount of data, usually through nonlinear multivariate models, contributed to improvements of ANN downscaling models and the development of other new, machine learning-based downscaling models to overcome some ANN drawbacks, such as support vector machine and random forest techniques. The mixed models combining various machine learning downscaling approaches maximize the downscaling skill in local climate change applications, especially for extreme rainfall indices. Other nonlinear statistical downscaling approaches refer to conditional weather generators, combining a standard weather generator (WG) with a separate statistical downscaling model by conditioning the WG parameters on large-scale predictors via a nonlinear approach. The most popular ways to condition the WG parameters are the weather-type approach and generalized linear models. This article discusses various aspects of nonlinear statistical downscaling approaches, their strengths and weaknesses, as well as comparison with linear statistical downscaling models. A proper validation of the nonlinear statistical downscaling models is an important issue, allowing selection of an appropriate model to obtain credible information on local climate change. Selection of large-scale predictors, the model’s ability to reproduce historical trends, extreme events, and the uncertainty related to future downscaled changes are important issues to be addressed. A better estimation of the uncertainty related to downscaled climate change projections can be achieved by using ensembles of more GCMs as drivers, including their ability to simulate the input in downscaling models. Comparison between more future statistical downscaled climate change signals and those derived from dynamical downscaling driven by the same global model, including a complex validation of the RCMs, gives a measure of the reliability of downscaled regional climate changes.

Estimating total net primary productivity of managed grasslands by a state-space modeling approach in a small catchment on the Loess Plateau, China

There is increasing concern that avoiding climate change impacts will require proactive adaptation, particularly for infrastructure systems with long lifespans. However, one challenge in adaptation is the uncertainty surrounding climate change projections generated by general circulation models (GCMs). This uncertainty has been addressed in different ways. For example, some researchers use ensembles of GCMs to generate probabilistic climate change projections, but these projections can be highly sensitive to assumptions about model independence and weighting schemes. Because of these issues, others argue that robustness-based approaches to climate adaptation are more appropriate, since they do not rely on a precise probabilistic representation of uncertainty. In this research, we present a new approach for characterizing climate change risks that leverages robust decision frameworks and probabilistic GCM ensembles. The scenario discovery process is used to search across a multi-dimensional space and identify climate scenarios most associated with system failure, and a Bayesian statistical model informed by GCM projections is then developed to estimate the probability of those scenarios. This provides an important advancement in that it can incorporate decision-relevant climate variables beyond mean temperature and precipitation and account for uncertainty in probabilistic estimates in a straightforward way. We also suggest several advancements building on prior approaches to Bayesian modeling of climate change projections to make them more broadly applicable. We demonstrate the methodology using proposed water resources infrastructure in Lake Tana, Ethiopia, where GCM disagreement on changes in future rainfall presents a major challenge for infrastructure planning.

<p>Selection of a suitable Global Climate Model (GCM) in hydrological research for basin-scale of monsoon affected regions under future climate projection scenarios is a great necessity. This study comprehensively evaluated the suitability of 25 available GCMs issued of  Coupled Model Intercomparison Project 6 (CMIP6) to choose the best performing GCMs in precipitation simulating skill over the whole main River Basin System in South Korea for the historical period of 1973–2014. Bilinear interpolation method was used for mapping the grid resolution of the simulated GCMs precipitation and observed precipitation with a 0.125<sup>0</sup> x 0.125<sup>0</sup> resolution. Where, the observed monthly precipitation at 56 automated weather stations from 1973 to 2014 were derived from the Korea Meteorological Administration (KMA). Multi-Criteria Decision Making (MCDM) approach based on four spatial metrics, Cramer’s V, Goodman-Kruskal (GK) Lambda, Mapcurves and TheilU were proposed to compare the simulated GCMs precipitation with the observed precipitation. To calculate the overall ranking of the GCMs and identify the best performing GCMs, this study applied Jenks Natural Break classification based on the Compromise Programming index. The results indicated that: 1) The GCMs performance was different with different spatial indices with the most suitable of GCMs ranking for each watershed. 2) The best performing GCMs well simulated the annual mean precipitation with a bias of less than 15% for southwestern watersheds and higher biases (30-50%) for remaining watersheds. 3) Majority of CMIP6 GCMs could be capture trends and the spatial distribution of annual, seasonal precipitation over South Korea. However, the result was also found that most GCMs underestimated summer precipitation and overestimated spring precipitation. Therefore, the selected GCMs with corrected biases can be usefully employed for analyzing future changes of hydrological pattern associated with climate change projections.</p><p><strong>Keywords: </strong>Global Climate Models (GCMs), CMIP6, Bilinear interpolation, Multi-Criteria Decision Making, Jenks Natural Break classification.</p>

Replicability of Annual and Seasonal Precipitation by CMIP5 and CMIP6 GCMs over East Asia

The Australian eastern seaboard is a distinct climate entity from the interior of the continent, with different climatic influences on each side of the Great Dividing Range. Therefore, it is plausible that downscaling of global climate models could reveal meaningful regional detail, or ‘added value’, in the climate change signal of mean rainfall change in eastern Australia un-der future scenarios. However, because downscaling is typically done using a limited set of global climate models and downscaling methods, the results from a downscaling study may not represent the range of uncertainty in plausible projected change for a region suggested by the ensemble of host global climate models. A complete and unbiased representation of the plausible changes in the climate is essential in producing climate projections useful for future planning. As part of this aim it is important to quantify any differences in the change signal between global climate models and downscaling, and understand the cause of these differ-ences in terms of plausible added regional detail in the climate change signal, the impact of sub-sampling global climate models and the effect of the downscaling models themselves. Here we examine rainfall projections in eastern Australia under a high emissions scenario by late in the century from ensembles of global climate models, two dynamical downscaling models and one statistical downscaling model. We find no cases where all three downscaling methods show the same clear regional spatial detail in the change signal that is distinct from the host models. However, some downscaled projections suggest that the eastern seaboard could see little change in spring rainfall, in contrast to the substantial rainfall decrease inland. The change signal in the downscaled outputs is broadly similar at the large scale in the various model outputs, with a few notable exceptions. For example, the model median from dynamical downscaling projects a rainfall increase over the entirety of eastern Australia in autumn that is greater than the global models. Also, there are some instances where a downscaling method produces changes outside the range of host models over eastern Australia as a whole, thus ex-panding the projected range of uncertainty. Results are particularly uncertain for summer, where no two downscaling studies clearly agree. There are also some confounding factors from the model configuration used in downscaling, where the particular zones used for statis-tical models and the model components used in dynamical models have an influence on results and produce additional uncertainty.

The paper presents a broad general review of the state space approach to time series analysis. It begins with an introduction to the linear Gaussian state space model. Applications to problems in practical time series analysis are considered. The state space approach is briefly compared with the Box–Jenkins approach. The Kalman filter and smoother and the simulation smoother are described. Missing observations, forecasting and initialisation are considered. A representation of a multivariate series as a univariate series is displayed. The construction and maximisation of the likelihood function are discussed. An application to real data is presented. The treatment is extended to non-Gaussian and nonlinear state space models. A simulation technique based on importance sampling is described for analysing these models. The use of antithetic variables in the simulation is considered. Bayesian analysis of the models is developed based on an extension of the importance sampling technique. Classical and Bayesian methods are applied to a real time series.

Seasonal snow directly affects New Zealand’s economy through the energy, agriculture and tourism sectors. In New Zealand, little is known about the long-term variability of the snow cover and the expected impacts of climate change on snow cover. The lack of systematic historical snow observations in New Zealand means that information on interannual variability, trends and projections of future seasonal snow must be generated using simulation models. We use a temperature index snow model to calculate the accumulation and ablation of the current (1980–1999) snowpack for more than 37,000 third-order river basins with 100 m contour intervals, resulting in over 200,000 individual model elements in New Zealand. Using this model, which captures the gross features of snow under the current climate, we assess the range of likely effects of climate change on seasonal snow in New Zealand using downscaled temperature and precipitation changes from the middle of the road (A1B) climate change projections from 12 general circulation models (GCMs). For each of the 12 GCMs, we consider two future time periods 2030–2049 (mid-point reference 2040) and 2080–2099 (mid-point reference 2090). These future time periods are compared to simulations of current, 1980–1999 (mid-point reference 1990), seasonal snow. Our results show that on average at a national scale, at nearly all elevations, the 2040s and 2090s result in a decrease in snow as described by all of our summary statistics: snow duration, percentage of precipitation that is snow and peak snow accumulation in each year. This decrease in snow is more marked at elevations below 1,000 m but is evident at all but the very highest elevations. Relative to snow simulations for average peak snow accumulation for the present, we observe that by the 2040s, depending on the GCM used, there is a reduction of between 3 and 44 % at 1,000 m, and an increase of 8 % through to a reduction of 22 % at 2,000 m. By the 2090s, the average reduction is greater, with a decrease of between 32 and 79 % at 1,000 m and between 6 and 51 % at 2,000 m. More substantial reductions are observed below these elevations. When we consider the elevation where snow duration exceeds 3 months, we see a rise in this elevation from 1,550 m in the 1990s to between 1,550 and 1,750 m by the 2040s and 1,700 and 2,000 m by the 2090s, depending on the GCM used. The results of this work are consistent with our understanding of snow processes in general and with work from other similar mid-latitude locations.

Regional climate modeling is a dynamical downscaling technique applied to the results of global climate models (GCMs) in order to acquire more information on climate simulations and climate change projections. GCMs and regional climate models (RCMs) have undergone considerable development over the past few decades, and both have increased in resolution. The higher‐resolution edge of RCMs compared to GCMs still remains, however. This has been demonstrated in a number of specific studies. As GCMs operate on relatively coarse resolutions, they do not resolve more variable land forms and similar features that shape regional‐scale climates. RCMs operate on higher resolutions than GCMs, by a factor of 2–10. Some RCMs now explore resolutions down to 1–5 km. This adds value in regions with variable orography, land–sea and other contrasts, as well as in capturing sharp, short‐duration and extreme events. In contrast, large‐scale and time‐averaged fields, not least over smooth terrain and on scales that have been already skillfully resolved in GCMs, are not much affected. RCMs also generate additional detail compared to GCMs when in climate projection mode. Compared to the present‐day climate for which observations exist, here the added value aspect is more complex to evaluate. Nevertheless, added value is meaningfully underlined when there is a clear physical context for it to appear in. In addition to climate modeling and model evaluation‐related added value considerations, a significant relevant aspect of added value is the provision of regional scale information, including climate change projections, for climate impact, adaptation, and vulnerability research. WIREs Clim Change 2016, 7:145–159. doi: 10.1002/wcc.378This article is categorized under: Climate Models and Modeling > Knowledge Generation with Models

Global climate models (GCMs) represent invaluable instrument for various purposes, most prominently studying climate system dynamics, evolution of past climates and climate change projections. The newest set of GCM simulations has been produced under CMIP6 initiative coordinated by the the World Climate Research Programme’s (WCRP) Working Group on Coupled Modelling. Due to their coarse spatial resolution, some kind of downscaling is necessary for applications of GCM outputs on regional and local scales, with dynamical downscaling using regional climate models (RCMs) being a common solution of this issue. Clearly, the outputs of RCMs are influenced by the boundary conditions provided at the lateral boundaries of the integration domain. The magnitude of this influence is subject of ongoing research and depends on various aspects including the geographical region, temporal scale, climatic variable etc. Nevertheless, previous studies proved that an analysis of boundary conditions is needed for for proper RCM evaluation. Especially with regard to potential error propagation.              In present study we evaluate the simulation of CNRM-ESM2-1, one of the CMIP6 GCMs, and compare it to other CMIP6 ensemble members. The CNRM-ESM2-1 is being used as driving model for convective-permitting simulation of Aladin-CLIMATE RCM within Czech national project PERUN aimed at creation of updated climate change scenarios for the Czech Republic. We concentrate on historical GCM simulations in the period of 1990-2014, and use reanalysis ERA5 as the reference. The analysis is conducted over the boundaries of Aladin-CLIMATE/CZ intergration domain, which covers Central Europe. The deviations of the driving GCM (CNRM-ESM2-1) from reanalysis are compared to other CMIP6 GCMs. We also evaluate uncertainty arising from natural climate variability using perturbed initial conditions ensemble of CNRM-ESM2-1.

General Circulation Models (GCMs) enable one to understand how climate changes on a regional scale and are of great importance in planning for our habitat. Statistical downscaling is the commonly used downscaling method mainly due to its computational efficiency. Even though studies utilizing GCM outputs for watershed level studies for certain basins in India are available, only very few are reported for peninsular region of India. The present study is aimed at estimating the better performing GCM for a river basin of Kerala State, India, which is bounded by the orographic features on one side and sea on the other side. Such type of studies is of increasing importance presently as the rainfall patterns and climate are seen to have a striking change in the recent past in the state. Multivariate Linear Regression with Principal Component Analysis was performed in this study for statistical downscaling of rainfall in the study region. Two GCMs namely CGCM3 and BCM2 were used for development of the downscaling models. The model was calibrated and tested with 50 years of NCEP reanalysis data and was used for projecting the precipitation in the study region. The performance of these GCMs was compared for arriving at a better model for water resources planning in the region. It was seen that BCM2 performed better in projecting the rainfall in the area.

Are the Projections of Future Climate Change Reliable in the IPCC Reports?

Future climate change is expected to impact the extent, frequency, and magnitude of soil erosion in a variety of ways. The most direct of these impacts is the projected increase in the erosive power of rainfall owing to an increase in the moisture-holding capacity of the atmosphere, but other more indirect impacts include changes in plant biomass and shifts in land use to accommodate the new climatic regime. Given the potential for climate change to increase soil erosion and its associated adverse impacts, modelling future rates of erosion is a crucial step in its assessment as a potential future environmental problem, and as a basis to help advise future conservation strategies. In this study, the Water Erosion Prediction Project (WEPP) model is used to simulate the impacts of climate change on future rates of soil erosion for a case study hillslope in Northern Ireland for three future time periods centered on the 2020s, 2050s and 2080s. Despite the wide range of previous modelling studies, in the majority of cases a number of limitations are apparent with respect to their treatment of the direct impacts (changed climate data), and their failure to factor in the indirect impacts (changing land use and management). In addressing the need for site-specific climate change impacts, for example, many previous studies have attempted to downscale future climate change output from general circulation models (GCMs). The most popular downscaling approach in future soil erosion studies is the change factor method, yet this approach possesses severe limitations with respect to modelling future erosion rates since it incorporates only changes in the mean climate and fails to account for climate variability. In order to address this limitation, statistical downscaling methods are used in this study to downscale future climate change projections using three GCMs and two emissions scenarios, providing daily site-specific climate inputs to WEPP in a manner that incorporates both changes in the mean climate and its variability. The temporal scale of climate change projections is also a key limitation with respect to modelling future erosion rates. The most severe soil losses often occur in high intensity rainfall events that occur over very short time intervals, yet input to soil erosion models tends to be at a daily resolution. Given the decreasing confidence of future climate change projections at a sub-daily temporal resolution, a sensitivity analysis approach is used in this study to perturb the sub-daily rainfall intensity parameter in WEPP. In addition, most previous studies fail to account for the indirect impacts of climate change on soil erosion, with no change in land use and management often assumed. Here, a scenarios-based approach is employed to examine the impacts of changing crop cover and management on future rates of soil erosion. Results indicate a mix of soil erosion increases and decreases, depending on which scenarios are considered. Downscaled climate change projections in isolation generally result in erosion decreases, whereas large increases are projected when land use is changed from the current cover of grass to a row crop which requires annual tillage, and/or where large changes in sub-daily rainfall intensity are applied. The overall findings illustrate the potential for increased soil erosion under future climate change, and illuminate the need to address key limitations in previous studies with respect to the treatment of future climate change projections, and crucially, the factoring in of future land use and management.

Climate change is a well-known phenomenon all over the globe. The influence of projected climate change on agricultural production, either positive or negative, can be assessed for various locations. The present study was conducted to investigate the impact of projected climate change on rice’s production, water demand and phenology for the state of Bihar, India. Furthermore, this study assessed the irrigation water requirement to increase the rice production by 60%, for the existing current climate scenario and all the four IPCC climate change scenarios (RCP 2.6, RCP 4.5, RCP 6.0 and RCP 8.5) by the 2050s (2050–2059). Various management practices were used as adaptation methods to analyze the requirement of irrigation water for a 60% increase in rice production. The climate data obtained from the four General Circulation Models (GCMs) (bcc_csm1.1, csiro_mk3_6_0, ipsl_cm5a_mr and miroc_miroc5) were used in the crop growth model, with the Decision Support System for Agrotechnology Transfer (DSSAT) used to simulate the rice yield, phenological days and water demand under all four climate change scenarios. The results obtained from the CERES-Rice model in the DSSAT, corresponding to all four GCMs, were ensembled together to obtain the overall change in yield, phenology and water demand for 10 years of interval from 2020 to 2059. We investigated several strategies: increasing the rice’s yield by 60% with current agronomic practice; increasing the yield by 60% with conservation agricultural practice; and increasing the rice yield by 30% with current agronomic practice as well as with conservation agricultural practices (assuming that the other 30% increase in yield would be achieved by reducing post-harvest losses by 30%). The average increase in precipitation between 2020 and 2059 was observed to be 5.23%, 13.96%, 9.30% and 9.29%, respectively, for RCP 2.6, RCP 4.5, RCP 6.0 and RCP 8.5. The decrease in yield during the 2050s, from the baseline period (1980–2004), was observed to be 2.94%, 3.87%, 4.02% and 5.84% for RCP 2.6, RCP 4.5, RCP 6.0 and RCP 8.5, respectively. The irrigation requirement was predicted to increase by a range of 39% to 45% for a 60% increase in yield using the current agronomic practice in current climate scenario and by 2050s with all the four climate change scenarios from the baseline period (1980–2004). We found that if we combine both conservation agriculture and removal of 30% of the post-harvest losses, the irrigation requirement would be reduced by 26% (45 to 19%), 20% (44 to 24%), 21% (43 to 22%), 22% (39 to 17%) and 20% (41 to 21%) with current climate scenario, RCP 2.6, RCP 4.5, RCP 6.0 and RCP 8.5 conditions, respectively. This combination of conservation practices suggests that the irrigation water requirement can be reduced by a large percentage, even if we produce 60% more food under the projected climate change conditions.