The effect of univariate bias adjustment on multivariate hazard estimates

Earth System Dynamics, Journal Year: 2019, Volume and Issue: 10(1), P. 31 - 43

Published: Jan. 7, 2019

Abstract. Bias adjustment is often a necessity in estimating climate impacts because impact models usually rely on unbiased information, requirement that model outputs rarely fulfil. Most currently used statistical bias-adjustment methods adjust each variable separately, even though depend multiple potentially dependent variables. Human heat stress, for instance, depends temperature and relative humidity, two variables are strongly correlated. Whether univariate effectively improve estimates of drivers largely unknown, the lack long-term data prevents direct comparison between observations many climate-related impacts. Here we use hazard indicators, stress simple fire risk indicator, as proxies more sophisticated models. We show such quantile mapping cannot reduce biases multivariate estimates. In some cases, it increases biases. These cases typically occur (i) when hazards equally than one climatic driver, (ii) exhibit dependence structure (iii) relatively small. Using perfect approach, further quantify uncertainty bias-adjusted indicators due to internal variability how imperfect bias can amplify this uncertainty. Both issues be addressed successfully with corrects addition marginal distributions drivers. Our results suggest modeled associated uncertainties related choice adjustment. conclude where these reduced using approaches correct variables' structure.

Language: Английский

---

Crop production losses associated with anthropogenic climate change for 1981-2010 compared with preindustrial levels

International Journal of Climatology, Journal Year: 2018, Volume and Issue: 38(14), P. 5405 - 5417

Published: Aug. 20, 2018

The accumulated evidence indicates that agricultural production is being affected by climate change. However, most of the available at a global scale based on statistical regressions. Corroboration using independent methods, specifically process‐based modelling, important for improving our confidence in evidence. Here, we estimate impacts change average yields maize, rice, wheat and soybeans 1981–2010, relative to preindustrial climate. We use results factual non‐warming counterfactual simulations performed with an atmospheric general circulation model do not include anthropogenic forcings systems, respectively, as inputs into gridded crop model. 100‐member ensemble simulation suggest has decreased mean 4.1, 1.8 4.5%, (preindustrial climate), even when carbon dioxide (CO 2 ) fertilization agronomic adjustments are considered. For no significant (−1.8%) detected. uncertainties estimated yield represented 90% probability interval derived from members −8.5 +0.5% −8.4 −0.5% soybeans, −9.6 +12.4% rice − 7.5 +4.3% wheat. Based impacts, estimates annual losses throughout world recent years study (2005–2009) account 22.3 billion USD (B$) 6.5 B$ 0.8 13.6 Our assessment confirms modulated led losses, adaptations date have been sufficient offset negative change, particularly lower latitudes.

Language: Английский

---

Climate Change Impact Studies: Should We Bias Correct Climate Model Outputs or Post‐Process Impact Model Outputs?

Water Resources Research, Journal Year: 2021, Volume and Issue: 57(5)

Published: April 28, 2021

Abstract The inter‐variable dependence of climate variables is usually not considered in many bias correction methods, even though it has been deemed important for various impact studies. Another possible approach to forgo the model outputs, and instead, post‐process outputs model. This advantage circumventing difficulties associated with correcting variables. Using a hydrological study as an example, this investigates feasibility by comparing performance pre‐processing post‐processing simulations when using methods. over calibration validation periods was used assess transferability both approaches. results show that procedures are capable significantly reducing simulated streamflow time series most global models (GCMs), their performances depend on GCM simulations, models, metrics watersheds. Both approaches were likely perform badly period factors have strong seasonal variability therefore sensitive nonstationarity and/or between periods. problem found be more acute method because streamflows often pattern abrupt changes than precipitation temperature. For reason, recommended less suffer from problem.

Language: Английский

---

Compound Events under Global Warming: A Dependence Perspective

Journal of Hydrologic Engineering, Journal Year: 2020, Volume and Issue: 25(9)

Published: July 2, 2020

Due to enhanced impacts of compound events, the importance assessing climate change on extremes from a multivariate perspective has recently been receiving considerable attention. This study provides state-of-the-art review events dependence multiple contributing variables, based both synthetic data sets and observations. The cause dependence, relationship between likelihoods changes in risks associated with are reviewed, illustration two typical examples dry–hot flooding events. Also discussed related topics, including sample sizes, bias correction separating driving factors event changes. Insights provided by this will be useful for building resilience cope under changing climate.

Language: Английский

---

Ensemble climate-impact modelling: extreme impacts from moderate meteorological conditions

Environmental Research Letters, Journal Year: 2020, Volume and Issue: 15(3), P. 034050 - 034050

Published: Feb. 14, 2020

Abstract The investigation of risk due to weather and climate events is an example policy relevant science. Risk the result complex interactions between physical environment (geophysical or conditions, including but not limited events) societal factors (vulnerability exposure). impact two similar meteorological at different times locations may therefore vary widely. Despite relation conditions impacts, most research focused on occurrence severity extreme events, often undersamples climatological natural variability. Here we argue that approach ensemble climate-impact modelling required adequately investigate relationship meteorology events. We demonstrate do always lead impacts; in contrast, impacts from (coinciding) moderate conditions. Explicit using complete distribution realisations, thus necessary ensure are identified. allows for high-impact provides higher accuracy consequent estimates risk.

Language: Английский

---

On deep learning-based bias correction and downscaling of multiple climate models simulations

Climate Dynamics, Journal Year: 2022, Volume and Issue: 59(11-12), P. 3451 - 3468

Published: April 15, 2022

Language: Английский

---

Historical changes and projected trends of extreme climate events in Xinjiang, China

Climate Dynamics, Journal Year: 2022, Volume and Issue: 59(5-6), P. 1753 - 1774

Published: Jan. 7, 2022

Language: Английский

---

Multivariate Bias‐Correction of High‐Resolution Regional Climate Change Simulations for West Africa: Performance and Climate Change Implications

Journal of Geophysical Research Atmospheres, Journal Year: 2022, Volume and Issue: 127(5)

Published: Feb. 13, 2022

Abstract A multivariate bias correction based on N‐dimensional probability density function transform (MBCn) technique is applied to four different high‐resolution regional climate change simulations and key meteorological variables, namely precipitation, mean near‐surface air temperature, maximum minimum surface downwelling solar radiation, relative humidity, wind speed. The impact of bias‐correction the historical (1980–2005) period, inter‐variable relationships, measures spatio‐temporal consistency are investigated. focus discrepancies between original bias‐corrected results over five agro‐ecological zones. We also evaluate relevant indices for agricultural applications such as extreme indices, under current future (2020–2050) conditions RCP4.5. Results show that MBCn successfully corrects seasonal biases in spatial patterns intensities all their intervariable correlation, distributions most analyzed variables. Relatively large reductions during period give indication possible benefits when scenarios. Although models do not agree same positive/negative sign seven variables grid points, model ensemble shows a statistically significant rainfall, humidity Northern zone speed Coastal West Africa increasing summer temperature up 2°C Sahara.

Language: Английский

---

Effects of Climate change on temperature and precipitation in the Lake Toba region, Indonesia, based on ERA5-land data with quantile mapping bias correction

Scientific Reports, Journal Year: 2023, Volume and Issue: 13(1)

Published: Feb. 13, 2023

Abstract Climate change is a serious problem that can cause global variations in temperature and rainfall patterns. This variation affect the water availability of lakes. In this study, trends Lake Toba area for 40 years (1981–2020) were analyzed using ERA5-Land data corrected with observation station utilizing quantile mapping bias correction method. Corrected used study to show spatial patterns trends. The Mann–Kendall Sen slope tests carried out see magnitude trend. A comparison against their baseline period (1951–1980) was also investigated. results climate has affected trend increasing area, an increase 0.006 °C per year average 0.71 mm year. general, significant changes occurred last decade, 0.24 22%. impact expected be useful policymakers managing resources area.

Language: Английский

---

Multivariate bias correction of regional climate model boundary conditions

Climate Dynamics, Journal Year: 2023, Volume and Issue: 61(7-8), P. 3253 - 3269

Published: March 10, 2023

Abstract Improving modeling capacities requires a better understanding of both the physical relationship between variables and climate models with higher degree skill than is currently achieved by Global Climate Models (GCMs). Although Regional (RCMs) are commonly used to resolve finer scales, their application restricted inherent systematic biases within GCM datasets that can be propagated into RCM simulation through model input boundaries. Hence, it advisable remove in simulations prior downscaling, forming improved boundary conditions for RCMs. Various mathematical approaches have been formulated correct such biases. Most techniques, however, each variable independently leading inconsistencies across dynamically linked fields. Here, we investigate bias corrections ranging from simple more complex techniques conditions. The results show substantial improvements performance after applying correction boundaries RCM. This work identifies effectiveness increasingly sophisticated able improve simulated rainfall characteristics. An multivariate correction, which corrects temporal persistence inter-variable relationships, represents extreme events relative univariate do not account variables.

Language: Английский

---