Christian Merkenschlager
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Christian Merkenschlager has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
Elke Hertig
Regional Climate Change and Health, Faculty of Medicine, University of Augsburg, Universitätsstr. 2, 86159 Augsburg, Germany
5 shared publications- Elke Hertig
Faculty of Medicine, University of Augsburg, Augsburg, Germany
3 shared publications Freddy Bangelesa
Institute of Geography and Geology, Universtiy of Wuerzburg, Am Hubland, 97074 Wuerzburg, Germany
3 shared publicationsHeiko Paeth
Institute of Geography and Geology, Universtiy of Wuerzburg, Am Hubland, 97074 Wuerzburg, Germany
3 shared publications- Heiko Paeth
University of Würzburg
3 shared publications Christoph Beck
Institute of Geography, University of Augsburg, Augsburg, Germany
2 shared publications
Latest Publications
- Open Access
- 27 November 2025
- Published by Springer Nature in Parasites & Vectors
- Vol. 18 (1)
Changes in the recent habitat suitability of Euro-Mediterranean Anopheles species due to land-use and climate
Background: Habitat suitability of Anopheles mosquitoes depends on appropriate climate and land-use conditions. Anopheles mosquitoes are the main vectors for malaria transmission in the Euro-Mediterranean region, and there are major concerns that these species will expand and/or shift their range due to the expected changes in climate and land-use. This study aims to identify the main climate and land-use drivers of changes in the habitat suitability for six different Anopheles species between 2000 and 2020 within the Euro-Mediterranean region. Methods: Boosted regression trees were applied to establish the link between climate and land-use predictors and habitat suitability. An ensemble of 16 models, based on different methods of selecting background points and statistical predictors, was applied to each species. The ensemble was evaluated by means of model skill and transferability to identify the best model. Taking contribution, interactions and response range into account, the most important predictors and those responsible for changes were identified. Results: The model ensembles agreed on the direction of change for four Anopheles species within the study area, with two of these showing an overall increase ( An. atroparvus, An. sacharovi ) of areas with suitable conditions and two showing a decrease ( An. messeae, An. sergentii ). Climate change was found to be the main driver of shifts in habitat suitability, with only a few models attributing changes mainly to land-use. The limited influence of land-use changes may be due to the spatial resolution being too coarse. For most species, temperature-related bioclimatic variables (BIO4, BIO5, BIO8) were the most important predictors of changes in habitat suitability. A superior method for either the specific background points or predictor selection did not emerge because the relative ranking of the corresponding models is dependent on the species analyzed. Conclusions: Between 2000 and 2020, rising temperatures were the main driver of changes in the habitat suitability of Anopheles mosquitoes in the Euro-Mediterranean region, with land-use changes having a relatively minor impact. In particular, regions to the north of the respective distribution area were found to be characterized by an increasing habitat suitability, while regions to the south showed decreasing trends. These trends may also impact the risk of local malaria transmission in these regions. Graphical
- 27 November 2025
- Published by Springer Nature in Parasites & Vectors
- Vol. 18 (1)
Open AccessChanges in the recent habitat suitability of Euro-Mediterranean Anopheles species due to land-use and climate
Background: Habitat suitability of Anopheles mosquitoes depends on appropriate climate and land-use conditions. Anopheles mosquitoes are the main vectors for malaria transmission in the Euro-Mediterranean region, and there are major concerns that these species will expand and/or shift their range due to the expected changes in climate and land-use. This study aims to identify the main climate and land-use drivers of changes in the habitat suitability for six different Anopheles species between 2000 and 2020 within the Euro-Mediterranean region. Methods: Boosted regression trees were applied to establish the link between climate and land-use predictors and habitat suitability. An ensemble of 16 models, based on different methods of selecting background points and statistical predictors, was applied to each species. The ensemble was evaluated by means of model skill and transferability to identify the best model. Taking contribution, interactions and response range into account, the most important predictors and those responsible for changes were identified. Results: The model ensembles agreed on the direction of change for four Anopheles species within the study area, with two of these showing an overall increase ( An. atroparvus, An. sacharovi ) of areas with suitable conditions and two showing a decrease ( An. messeae, An. sergentii ). Climate change was found to be the main driver of shifts in habitat suitability, with only a few models attributing changes mainly to land-use. The limited influence of land-use changes may be due to the spatial resolution being too coarse. For most species, temperature-related bioclimatic variables (BIO4, BIO5, BIO8) were the most important predictors of changes in habitat suitability. A superior method for either the specific background points or predictor selection did not emerge because the relative ranking of the corresponding models is dependent on the species analyzed. Conclusions: Between 2000 and 2020, rising temperatures were the main driver of changes in the habitat suitability of Anopheles mosquitoes in the Euro-Mediterranean region, with land-use changes having a relatively minor impact. In particular, regions to the north of the respective distribution area were found to be characterized by an increasing habitat suitability, while regions to the south showed decreasing trends. These trends may also impact the risk of local malaria transmission in these regions. Graphical
- Open Access
- 17 July 2025
- Published by Springer Nature in Research Square
Changes in the recent habitat suitability of Euro-Mediterranean Anopheles species due to land-use and climate
Background: Habitat suitability of Anopheles mosquitoes depends on appropriate climate and land-use conditions. Anopheles are the main vectors for malaria transmissions in the Euro-Mediterranean region. There is major concern that there will be a spread or shift of Anopheles species due to the expected climate and land-use change. This study aims to identify the main climate and land-use drivers for changes in the habitat suitability for six different Anopheles species between 2000 and 2020 within the Euro-Mediterranean region. Methods: Boosted Regression Trees were applied to establish the link between climate and land-use predictors and habitat suitability. An ensemble of 16 models, based on different methods of selecting background points and statistical predictors, was applied to each species. The ensemble was evaluated by means of model skill and transferability to identify the best model. Under consideration of contribution, interactions and response range, the most important predictors and those responsible for changes were identified. Results: The model ensembles agree on the direction of change for four species within the study area, with two showing an overall increase ( An. atroparvus, An. sacharovi ) of areas with suitable conditions and two showing a decrease ( An. messeae, An. sergentii ). Climate change is mainly responsible for shifts in the habitat suitability. Only a few models attribute changes mainly to land-use. The limited influence of land-use changes may be due to the too coarse spatial resolution. For most species, temperature-related bioclimatic variables (BIO4, BIO5, BIO8) are the most important predictors for changes in habitat suitability. A superior method for either the specific background points or predictor selection did not emerge, because they depend on the species analyzed. Conclusions: Between 2000 and 2020, rising temperatures were the main driver of changes in the habitat suitability of the Anopheles species in the Euro-Mediterranean region, with land-use changes having a relatively minor impact. Especially regions to the north of the distribution area are characterized by a greater habitat suitability, while regions to the south show decreasing trends. This may also impact the risk of local malaria transmission in these regions.
- 17 July 2025
- Published by Springer Nature in Research Square
Open AccessChanges in the recent habitat suitability of Euro-Mediterranean Anopheles species due to land-use and climate
Background: Habitat suitability of Anopheles mosquitoes depends on appropriate climate and land-use conditions. Anopheles are the main vectors for malaria transmissions in the Euro-Mediterranean region. There is major concern that there will be a spread or shift of Anopheles species due to the expected climate and land-use change. This study aims to identify the main climate and land-use drivers for changes in the habitat suitability for six different Anopheles species between 2000 and 2020 within the Euro-Mediterranean region. Methods: Boosted Regression Trees were applied to establish the link between climate and land-use predictors and habitat suitability. An ensemble of 16 models, based on different methods of selecting background points and statistical predictors, was applied to each species. The ensemble was evaluated by means of model skill and transferability to identify the best model. Under consideration of contribution, interactions and response range, the most important predictors and those responsible for changes were identified. Results: The model ensembles agree on the direction of change for four species within the study area, with two showing an overall increase ( An. atroparvus, An. sacharovi ) of areas with suitable conditions and two showing a decrease ( An. messeae, An. sergentii ). Climate change is mainly responsible for shifts in the habitat suitability. Only a few models attribute changes mainly to land-use. The limited influence of land-use changes may be due to the too coarse spatial resolution. For most species, temperature-related bioclimatic variables (BIO4, BIO5, BIO8) are the most important predictors for changes in habitat suitability. A superior method for either the specific background points or predictor selection did not emerge, because they depend on the species analyzed. Conclusions: Between 2000 and 2020, rising temperatures were the main driver of changes in the habitat suitability of the Anopheles species in the Euro-Mediterranean region, with land-use changes having a relatively minor impact. Especially regions to the north of the distribution area are characterized by a greater habitat suitability, while regions to the south show decreasing trends. This may also impact the risk of local malaria transmission in these regions.
- Open Access
- 27 March 2025
- Published by Elsevier in Science of The Total Environment
- Vol. 974, 179202
Evolution of the recent habitat suitability area of Aedes albopictus in the extended Mediterranean area due to land-use and climate change
- 27 March 2025
- Published by Elsevier in Science of The Total Environment
- Vol. 974, 179202
Open AccessEvolution of the recent habitat suitability area of Aedes albopictus in the extended Mediterranean area due to land-use and climate change
- Open Access
- 28 September 2023
- Published by Wiley in Ecology and Evolution
- Vol. 13 (10) , e10553
Blessing and curse of bioclimatic variables: A comparison of different calculation schemes and datasets for species distribution modeling within the extended Mediterranean area
Bioclimatic variables (BCVs) are the most widely used predictors within the field of species distribution modeling, but recent studies imply that BCVs alone are not sufficient to describe these limits. Unfortunately, the most popular database, WorldClim, offers only a limited selection of bioclimatological predictors; thus, other climatological datasets should be considered, and, for data consistency, the BCVs should also be derived from the respective datasets. Here, we investigate how well the BCVs are represented by different datasets for the extended Mediterranean area within the period 1970–2020, how different calculation schemes affect the representation of BCVs, and how deviations among the datasets differ regionally. We consider different calculation schemes for quarters/months, the annual mean temperature (BCV‐1), and the maximum temperature of the warmest month (BCV‐5). Additionally, we analyzed the effect of different temporal resolutions for BCV‐1 and BCV‐5. Differences resulting from different calculation schemes are presented for ERA5‐Land. Selected BCVs are analyzed to show differences between WorldClim, ERA5‐Land, E‐OBS, and CRU. Our results show that (a) differences between the two calculation schemes for BCV‐1 diminish as the temporal resolution decreases, while the differences for BCV‐5 increase; (b) with respect to the definition of the respective month/quarter, intra‐annual shifts induced by the calculation schemes can have substantially different effects on the BCVs; (c) all datasets represent the different BCVs similarly, but with partly large differences in some subregions; and (d) the largest differences occur when specific month/quarters are defined by precipitation. In summary, (a) since the definition of BCVs matches different calculation schemes, transparent communication of the BCVs calculation schemes is required; (b) the calculation, integration, or elimination of BCVs has to be examined carefully for each dataset, region, period, or species; and (c) the evaluated datasets provide, except in some areas, a consistent representation of BCVs within the extended Mediterranean region.
- 28 September 2023
- Published by Wiley in Ecology and Evolution
- Vol. 13 (10) , e10553
Open AccessBlessing and curse of bioclimatic variables: A comparison of different calculation schemes and datasets for species distribution modeling within the extended Mediterranean area
Bioclimatic variables (BCVs) are the most widely used predictors within the field of species distribution modeling, but recent studies imply that BCVs alone are not sufficient to describe these limits. Unfortunately, the most popular database, WorldClim, offers only a limited selection of bioclimatological predictors; thus, other climatological datasets should be considered, and, for data consistency, the BCVs should also be derived from the respective datasets. Here, we investigate how well the BCVs are represented by different datasets for the extended Mediterranean area within the period 1970–2020, how different calculation schemes affect the representation of BCVs, and how deviations among the datasets differ regionally. We consider different calculation schemes for quarters/months, the annual mean temperature (BCV‐1), and the maximum temperature of the warmest month (BCV‐5). Additionally, we analyzed the effect of different temporal resolutions for BCV‐1 and BCV‐5. Differences resulting from different calculation schemes are presented for ERA5‐Land. Selected BCVs are analyzed to show differences between WorldClim, ERA5‐Land, E‐OBS, and CRU. Our results show that (a) differences between the two calculation schemes for BCV‐1 diminish as the temporal resolution decreases, while the differences for BCV‐5 increase; (b) with respect to the definition of the respective month/quarter, intra‐annual shifts induced by the calculation schemes can have substantially different effects on the BCVs; (c) all datasets represent the different BCVs similarly, but with partly large differences in some subregions; and (d) the largest differences occur when specific month/quarters are defined by precipitation. In summary, (a) since the definition of BCVs matches different calculation schemes, transparent communication of the BCVs calculation schemes is required; (b) the calculation, integration, or elimination of BCVs has to be examined carefully for each dataset, region, period, or species; and (c) the evaluated datasets provide, except in some areas, a consistent representation of BCVs within the extended Mediterranean region.
- Open Access
- 8 May 2023
- Published by Wiley
Blessing and curse of BioClim variables: A comparison of different calculation schemes and datasets for species distribution modeling within the extended Mediterranean area
Bioclimatic variables (BCVs) are the most widely used predictors within the field of ecological niche modeling. However, recent studies indicate that BCVs alone are not sufficient to describe these limits and more (bio)climatological variables should be considered. Unfortunately, the most popular database WorldClim offers only a limited selection of predictors; thus, other gridded station-based observations or reanalysis (GSOR) datasets should be taken into account. In the present study, we investigate how well the BCVs are represented by different GSOR datasets for the extended Mediterranean area within the period 1970-2020, how deviations among the datasets differ regionally and how different calculation schemes affect the representation of BCVs. We consider different calculation schemes for quarters/months, the annual mean temperature and the maximum temperature of the warmest month and show the effects on the respective BCV. Differences resulting from different calculation schemes are presented for ERA5-Land. Selected BCVs are analyzed to show differences between WorldClim, ERA5-Land, E-OBS and CRU. Our results show that (a) deviations between the two calculation schemes for annual mean temperature (maximum temperature of the warmest month) diminish (increase) when the temporal resolution is decreased; (b) with respect to the definition of the respective month/quarter, temporal shifts can have substantially different effects on the BCVs, depending on region; (c) overall, all datasets represent the different BCVs similarly, but with partly large differences in some subregions; (d) the largest differences occur when specific month/quarters are defined by the seasonal cycle of precipitation. In summary, (a) since the definition of BCVs matches different calculation schemes, a transparent communication of the BCVs calculation schemes is inevitable; (b) the calculation, integration or elimination of BCVs has to be examined carefully for each dataset, region, period or species; (c) the GSOR-Data provide, except some subregions, a consistent representation of BCVs within the extended Mediterranean region.
- 8 May 2023
- Published by Wiley
Open AccessBlessing and curse of BioClim variables: A comparison of different calculation schemes and datasets for species distribution modeling within the extended Mediterranean area
Bioclimatic variables (BCVs) are the most widely used predictors within the field of ecological niche modeling. However, recent studies indicate that BCVs alone are not sufficient to describe these limits and more (bio)climatological variables should be considered. Unfortunately, the most popular database WorldClim offers only a limited selection of predictors; thus, other gridded station-based observations or reanalysis (GSOR) datasets should be taken into account. In the present study, we investigate how well the BCVs are represented by different GSOR datasets for the extended Mediterranean area within the period 1970-2020, how deviations among the datasets differ regionally and how different calculation schemes affect the representation of BCVs. We consider different calculation schemes for quarters/months, the annual mean temperature and the maximum temperature of the warmest month and show the effects on the respective BCV. Differences resulting from different calculation schemes are presented for ERA5-Land. Selected BCVs are analyzed to show differences between WorldClim, ERA5-Land, E-OBS and CRU. Our results show that (a) deviations between the two calculation schemes for annual mean temperature (maximum temperature of the warmest month) diminish (increase) when the temporal resolution is decreased; (b) with respect to the definition of the respective month/quarter, temporal shifts can have substantially different effects on the BCVs, depending on region; (c) overall, all datasets represent the different BCVs similarly, but with partly large differences in some subregions; (d) the largest differences occur when specific month/quarters are defined by the seasonal cycle of precipitation. In summary, (a) since the definition of BCVs matches different calculation schemes, a transparent communication of the BCVs calculation schemes is inevitable; (b) the calculation, integration or elimination of BCVs has to be examined carefully for each dataset, region, period or species; (c) the GSOR-Data provide, except some subregions, a consistent representation of BCVs within the extended Mediterranean region.
- Open Access
- 16 April 2021
- Published by Springer Nature in Theoretical and Applied Climatology
- Vol. 145 (1-2) , 31-46
Assessing local daily temperatures by means of novel analog approaches: a case study based on the city of Augsburg, Germany
Within the scope of urban climate modeling, weather analogs are used to downscale large-scale reanalysis-based information to station time series. Two novel approaches of weather analogs are introduced which allow a day-by-day comparison with observations within the validation period and which are easily adaptable to future periods for projections. Both methods affect the first level of analogy which is usually based on selection of circulation patterns. First, the time series were bias corrected and detrended before subsamples were determined for each specific day of interest. Subsequently, the normal vector of the standardized regression planes (NVEC) or the center of gravity (COG) of the normalized absolute circulation patterns was used to determine a point within an artificial coordinate system for each day. The day(s) which exhibit(s) the least absolute distance(s) between the artificial points of the day of interest and the days of the subsample is/are used as analog or subsample for the second level of analogy, respectively. Here, the second level of analogy is a second selection process based on the comparison of gridded temperature data between the analog subsample and the day of interest. After the analog selection process, the trends of the observation were added to the analog time series. With respect to air temperature and the exceedance of the 90th temperature quantile, the present study compares the performance of both analog methods with an already existing analog method and a multiple linear regression. Results show that both novel analog approaches can keep up with existing methods. One shortcoming of the methods presented here is that they are limited to local or small regional applications. In contrast, less pre-processing and the small domain size of the circulation patterns lead to low computational costs.
- 16 April 2021
- Published by Springer Nature in Theoretical and Applied Climatology
- Vol. 145 (1-2) , 31-46
Open AccessAssessing local daily temperatures by means of novel analog approaches: a case study based on the city of Augsburg, Germany
Within the scope of urban climate modeling, weather analogs are used to downscale large-scale reanalysis-based information to station time series. Two novel approaches of weather analogs are introduced which allow a day-by-day comparison with observations within the validation period and which are easily adaptable to future periods for projections. Both methods affect the first level of analogy which is usually based on selection of circulation patterns. First, the time series were bias corrected and detrended before subsamples were determined for each specific day of interest. Subsequently, the normal vector of the standardized regression planes (NVEC) or the center of gravity (COG) of the normalized absolute circulation patterns was used to determine a point within an artificial coordinate system for each day. The day(s) which exhibit(s) the least absolute distance(s) between the artificial points of the day of interest and the days of the subsample is/are used as analog or subsample for the second level of analogy, respectively. Here, the second level of analogy is a second selection process based on the comparison of gridded temperature data between the analog subsample and the day of interest. After the analog selection process, the trends of the observation were added to the analog time series. With respect to air temperature and the exceedance of the 90th temperature quantile, the present study compares the performance of both analog methods with an already existing analog method and a multiple linear regression. Results show that both novel analog approaches can keep up with existing methods. One shortcoming of the methods presented here is that they are limited to local or small regional applications. In contrast, less pre-processing and the small domain size of the circulation patterns lead to low computational costs.
- 9 April 2020
- Published by Inter-Research Science Center in Climate Research
- Vol. 80 (1) , 19-42
Seasonal assessments of future precipitation extremes in the Mediterranean area considering nonstationarities in predictor-predictand relationships
- 9 April 2020
- Published by Inter-Research Science Center in Climate Research
- Vol. 80 (1) , 19-42
Seasonal assessments of future precipitation extremes in the Mediterranean area considering nonstationarities in predictor-predictand relationships
- Open Access
- 9 March 2020
- Published by Copernicus GmbH
Assessment of future heat events for the city of Augsburg by means of a normal vector based analog approach
Under enhanced anthropogenic greenhouse gas forcing heat waves are only one example of climatic risks mankind has to deal with. Especially in urban areas where most of the people will live until the end of the 21st century heat waves are a serious risk factor since the urban heat island will reinforce such events. For the city of Augsburg, new analog methods are utilized for assessing the development and impacts of heat waves taking into account the varying urban structure. For model calibration the temperature data from the Augsburg-Mühlhausen weather station operated by the German Weather Service (DWD) and atmospheric circulation variables of the ERA5 reanalysis data set were used to analyze the recent temperature development. For this purpose, the least deviation of the normal vector was used to determine a subsample of analogs corresponding to the day of interest. The normal vector was derived from the regression plane of the prevailing circulation on the respective day. Subsequently, the temperature patterns were used to define the analog day from the subsample. For future periods, the same method was applied to model data for two representative concentration pathways (RCP4.5, RCP8.5) of different general circulation models (GCM: ACCESS1-0, CNRM-CM5, MPI-ESM-LR). Thus, we derive future time series of analogs corresponding to events prevailing in the observational period. To account for projected trends of the GCMs, the trends of all time-series were first removed and, after the analog selection process, added again according to the trends of the GCMs. Temperature extremes are defined as days with temperatures exceeding the 90th quantile (Q90) and heat days are defined as days where at least two temperature indices (TMIN, TMEAN, TMAX) exceed Q90. When at least three consecutive days are defined as heat day a heat wave is proclaimed. Analysis have shown that under consideration of RCP8.5 (RCP4.5) and all model runs the number of heat days in the end of the 21st century will be nine (five) times higher than within the reference period 1970-2000. Furthermore, the mean duration of heatwaves will extend by factor four (two), whereby heat waves of more than 30 (15) consecutive days are possible.
- 9 March 2020
- Published by Copernicus GmbH
Open AccessAssessment of future heat events for the city of Augsburg by means of a normal vector based analog approach
Under enhanced anthropogenic greenhouse gas forcing heat waves are only one example of climatic risks mankind has to deal with. Especially in urban areas where most of the people will live until the end of the 21st century heat waves are a serious risk factor since the urban heat island will reinforce such events. For the city of Augsburg, new analog methods are utilized for assessing the development and impacts of heat waves taking into account the varying urban structure. For model calibration the temperature data from the Augsburg-Mühlhausen weather station operated by the German Weather Service (DWD) and atmospheric circulation variables of the ERA5 reanalysis data set were used to analyze the recent temperature development. For this purpose, the least deviation of the normal vector was used to determine a subsample of analogs corresponding to the day of interest. The normal vector was derived from the regression plane of the prevailing circulation on the respective day. Subsequently, the temperature patterns were used to define the analog day from the subsample. For future periods, the same method was applied to model data for two representative concentration pathways (RCP4.5, RCP8.5) of different general circulation models (GCM: ACCESS1-0, CNRM-CM5, MPI-ESM-LR). Thus, we derive future time series of analogs corresponding to events prevailing in the observational period. To account for projected trends of the GCMs, the trends of all time-series were first removed and, after the analog selection process, added again according to the trends of the GCMs. Temperature extremes are defined as days with temperatures exceeding the 90th quantile (Q90) and heat days are defined as days where at least two temperature indices (TMIN, TMEAN, TMAX) exceed Q90. When at least three consecutive days are defined as heat day a heat wave is proclaimed. Analysis have shown that under consideration of RCP8.5 (RCP4.5) and all model runs the number of heat days in the end of the 21st century will be nine (five) times higher than within the reference period 1970-2000. Furthermore, the mean duration of heatwaves will extend by factor four (two), whereby heat waves of more than 30 (15) consecutive days are possible.
- Open Access
- 12 November 2018
- Published by Wiley in International Journal of Climatology
- Vol. 39 (4) , 1912-1926
A circulation‐based approach considering non‐stationarities within the scope of statistical downscaling: An example of seasonal daily precipitation extremes in the Mediterranean area
Non‐stationarities within the predictor‐predictand‐relationships can substantially affect model skill of statistical downscaling approaches. For this reason, an approach is presented which takes varying predictor‐predictand‐relationships explicitly into account. Seasonal aggregated daily precipitation extremes (90th, 95th and 99th quantiles) were assessed by means of 31 year running contiguous calibration periods. Here, when the end of the time series was reached years from the beginning were attached. Thus, a regression model ensemble (RME) becomes available, with the number of established statistical models corresponding to the number of years within the time series. With respect to non‐stationarities in the projection periods, a novel approach is presented. A single model of the RME is determined for each sub‐period which seems to be the most suitable model (MSM) for assessing precipitation extremes within the respective period. In order to define the MSM weighted correlations of atmospheric circulation composites, represented by the geopotential heights, between reanalysis and model runs were calculated for all significant predictors. The reanalysis period which exhibits the highest mean correlation coefficient is then considered for the projections of the respective period of the model runs. Subsequently, a second selection process was performed in order to determine whether the regression model (RM) established within this period or the model with the highest skill within this period is used for assessing future precipitation extremes. Results have shown that projections can significantly differ depending on whether a non‐stationary or a stationary downscaling set‐up is used for the assessments.
- 12 November 2018
- Published by Wiley in International Journal of Climatology
- Vol. 39 (4) , 1912-1926
Open AccessA circulation‐based approach considering non‐stationarities within the scope of statistical downscaling: An example of seasonal daily precipitation extremes in the Mediterranean area
Non‐stationarities within the predictor‐predictand‐relationships can substantially affect model skill of statistical downscaling approaches. For this reason, an approach is presented which takes varying predictor‐predictand‐relationships explicitly into account. Seasonal aggregated daily precipitation extremes (90th, 95th and 99th quantiles) were assessed by means of 31 year running contiguous calibration periods. Here, when the end of the time series was reached years from the beginning were attached. Thus, a regression model ensemble (RME) becomes available, with the number of established statistical models corresponding to the number of years within the time series. With respect to non‐stationarities in the projection periods, a novel approach is presented. A single model of the RME is determined for each sub‐period which seems to be the most suitable model (MSM) for assessing precipitation extremes within the respective period. In order to define the MSM weighted correlations of atmospheric circulation composites, represented by the geopotential heights, between reanalysis and model runs were calculated for all significant predictors. The reanalysis period which exhibits the highest mean correlation coefficient is then considered for the projections of the respective period of the model runs. Subsequently, a second selection process was performed in order to determine whether the regression model (RM) established within this period or the model with the highest skill within this period is used for assessing future precipitation extremes. Results have shown that projections can significantly differ depending on whether a non‐stationary or a stationary downscaling set‐up is used for the assessments.