Bias correction /adjustment methods

Home Page Forums C3S_422_Lot1_SMHI forum Bias correction /adjustment methods

This topic contains 16 replies, has 14 voices, and was last updated by  Jonas Olsson 1 week, 3 days ago.

Viewing 15 posts - 1 through 15 (of 17 total)
  • Author
    Posts
  • #4039

    FulcoLudwig
    Participant

    A. Explain briefly what data you are planning to use for your case study and what data analyses and or (bias) – corrections you have already done or are planning to do

    B. To advise you on the most appropriate bias correction/adjustment and data analyses methods please answer the following key questions
    1. Which indicators / climate variables are you most interested in (temperature, precipitation, soil moisture, streamflow etc.)?
    2. Are you interested in gridded data or point/station data?
    3. What local data do you have available for bias correction ( e.g. station data on precipitation and temperature or a particular gridded data set)? For which time period, daily/monthly data?
    4. For your case are you only interested in changes in average values, or also in extremes (e.g. extreme rainfall, floods, droughts etc.)?
    5. Do you need to correct precipitation data? If so, are the number of raindays very important or only the total amount of rainfall?

    • This topic was modified 3 weeks, 2 days ago by  Lorna Little.
    #4181

    Michele
    Participant

    UKZN Reply to above questions:

    A. Explain briefly what data you are planning to use for your case study and what data analyses and or (bias) – corrections you have already done or are planning to do

    • The grid cell in which the driver climate stations in the uMngeni catchment occurred was determined.
    • The monthly changes in Tmax, TMin and PPT for all the GCMs for this grid cell were obtained from the CDS for both emissions scenarios for the periods 2011 – 2040 and 2041 – 2070.
    • These changes were applied to the 30 year observed climate record (1971 – 2000)
    o Change in monthly max temperature was added to the observed temperature record (i.e. each day’s maximum temperature in January was increased by the January change factor for maximum temperature for a specific GCM and emission scenario)
    o Change in monthly min temperature was added to the observed temperature record (i.e. each day’s min temperature in January was increased by the January change factor for min temperature for a specific GCM and emission scenario)
    o The observed daily precipitation values were adjusted using the monthly percentage change in precipitation (each day’s precip in January was multiplied by the January percentage change factor for precip for a specific GCM and emission scenario)

    B. To advise you on the most appropriate bias correction/adjustment and data analyses methods please answer the following key questions

    1. Which indicators / climate variables are you most interested in (temperature, precipitation, soil moisture, streamflow etc.)?

    We require daily rainfall and temperature to run the models, thus we are most interested in precipitation and temperature.

    2. Are you interested in gridded data or point/station data?

    Point/station data

    3. What local data do you have available for bias correction ( e.g. station data on precipitation and temperature or a particular gridded data set)? For which time period, daily/monthly data?

    We have high quality, point based, observed daily rainfall and daily maximum and minimum temperature available for nine stations through the uMngeni catchment for the period 1950 – 2000. As well as 1’ x 1’ gridded data for temperature, precipitation and potential evaporation.

    4. For your case are you only interested in changes in average values, or also in extremes (e.g. extreme rainfall, floods, droughts etc.)?

    The extremes would be of interesting, but as we have chosen to apply the future scenarios as a change to historical, observed data we recognise that looking at changes in variability is not possible.

    5. Do you need to correct precipitation data? If so, are the number of raindays very important or only the total amount of rainfall?

    We have decided to use the change method, thus a limitation of our results is that there is no change in the variability and that we will only be able to assess means and thresholds.

    #4182

    Lyudmila Lebedeva
    Participant

    A. Explain briefly what data you are planning to use for your case study and what data analyses and or (bias) – corrections you have already done or are planning to do
    Local data collection includes meteorological measurements, ground thawing and freezing depths, ground temperature, and ground water and ice content, as well as river ice break-up and freeze-up dates, and river ice thickness,. Daily values of air temperature, precipitation and air humidity from CDS for historical and future (2011 – 2040) periods will be used as forcing data for the Hydrograph and HYPE models for production of the following CCIs: river ice characteristics and ground thawing depth. We will compare outputs of climate model ensemble with local meteorological data. We will also run the impact models with uncorrected forcing data, compare results with simulations using the reference data (local station data)and assess sensitivity of the CIIs to bias of the climate input. We will introduce bias correction depending on results of the two approaches.
    B. To advise you on the most appropriate bias correction/adjustment and data analyses methods please answer the following key questions
    1. Which indicators / climate variables are you most interested in (temperature, precipitation, soil moisture, streamflow etc.)
    Air temperature, precipitation and air moisture as forcing data for our impact models for river ice conditions and ground thawing.
    2. Are you interested in gridded data or point/station data?
    Both, gridded data to be distributed on watershed models, and point data for simulation of representative landscape types.
    3. What local data do you have available for bias correction (e.g. station data on precipitation and temperature or a particular gridded data set)? For which time period, daily/monthly data?
    Daily station data on air temperature, precipitation and air moisture for 1966-2012
    4. For your case are you only interested in changes in average values, or also in extremes (e.g. extreme rainfall, floods, droughts etc.)?
    The most crucial input for CIIs production in our case is probably the daily air temperature: both with regard to averages (daily, monthly and annual), and day to day, seasonal and between year variability, as well as the long term trends. We still need to assess the sensitivity of our impact models to precipitation biases, but most likely it should not be so important.
    5. Do you need to correct precipitation data? If so, are the number of raindays very important or only the total amount of rainfall?
    The most crucial input for CIIs production in our case is air temperature. However, we need to assess the importance of precipitation biases as well. If important, the total amount is probably more important than the temporal distribution since it is the heat content of the precipitation that should be of interest.

    #4184

    RemC
    Participant

    NCWQR

    A. Explain briefly what data you are planning to use for your case study and what data analyses and or (bias) – corrections you have already done or are planning to do.
    We need bias corrected and downscaled precipitation and temperature data. These data are the forcing input to our impact (SWAT and HYPE) models. We have downloaded data from EFSG but still has to process for use to our impact models. With the available bias-corrected data from CDS we still have to downscale for our use.

    B. To advise you on the most appropriate bias correction/adjustment and data analyses methods please answer the following key questions

    1. Which indicators / climate variables are you most interested in (temperature, precipitation, soil moisture, streamflow etc.)?
    – Both temperature and precipitation as these are the forcing input of our impact models. Soil moisture ca be used to validate the model. Streamflow will be calculated from the impact model.

    2. Are you interested in gridded data or point/station data?
    We need gridded data. Our current precipitation and temperature data is at 4×4 km grids.

    3. What local data do you have available for bias correction ( e.g. station data on precipitation and temperature or a particular gridded data set)? For which time period, daily/monthly data?
    Our current daily (1981-2017) precipitation and temperature data is at 4×4 km grids.

    4. For your case are you only interested in changes in average values, or also in extremes (e.g. extreme rainfall, floods, droughts etc.)?
    We are mostly interested in the extremes. This will affect highly our impact models.

    5. Do you need to correct precipitation data? If so, are the number of rain days very important or only the total amount of rainfall?
    We need to correct precipitation data and are also interested in the number of rain days

    #4185

    Mario Rohrer
    Participant

    A. Explain briefly what data you are planning to use for your case study and what data analyses and or (bias) – corrections you have already done or are planning to do

    We need bias corrected and downscaled precipitation, Tx and Tn temperature data, as well as evapotranspiration data. We have downloaded CMIP5-GCM data of South America, as well as NCEP R1 reanalysis data. Furthermore, based on these data, we have calculated the effective precipitation index for the historical baseline (1986-2005) and scenarios for several projections and periods. But this data are too coarse for our clients. Our final products will be based on the SMHI’s 50 km statistically downscaled variables calculated for HYPE for South America.

    B. To advise you on the most appropriate bias correction/adjustment and data analyses methods please answer the following key questions

    1. Which indicators / climate variables are you most interested in (temperature, precipitation, soil moisture, streamflow etc.)?

    We are interested in all variables you mentioned, in addition in evapotranspiration and runoff.

    2. Are you interested in gridded data or point/station data?

    We are primarily interested in gridded data. For runoff we are also interested in station/catchment data.

    3. What local data do you have available for bias correction ( e.g. station data on precipitation and temperature or a particular gridded data set)? For which time period, daily/monthly data?

    We have a data portal with station data of Peru. The data is in daily or subdaily resolution and is basically from 1964-2017. Moreover, we have TRMM-MPA-data from 1998 to 2018 (without S-Patagonia) and GPM-imerg data starting in 2002 in subdaily and 10 km*10 km resolution.

    4. For your case are you only interested in changes in average values, or also in extremes (e.g. extreme rainfall, floods, droughts etc.)?

    We need also extremes.

    5. Do you need to correct precipitation data? If so, are the number of raindays very important or only the total amount of rainfall?

    The numbers of raindays are also very important.

    #4190

    JoniDH
    Participant

    B. To advise you on the most appropriate bias correction/adjustment and data analyses methods please answer the following key questions
    1. Which indicators / climate variables are you most interested in (temperature, precipitation, soil moisture, streamflow etc.)?
    Initially mean annual temperature and precipitation, then mean monthly, monthly and daily temperature and precipitation.

    2. Are you interested in gridded data or point/station data?
    Gridded data for the whole of Costa Rica (high resolution, ~1km or higher).

    3. What local data do you have available for bias correction ( e.g. station data on precipitation and temperature or a particular gridded data set)? For which time period, daily/monthly data?
    Station data with daily and monthly temperature and precipitation records of various record length covering 1950 to 2016.

    4. For your case are you only interested in changes in average values, or also in extremes (e.g. extreme rainfall, floods, droughts etc.)?
    To assess biodiversity changes according to project objectives we initially only require average values but will be looking at extremes as well.

    5. Do you need to correct precipitation data? If so, are the number of raindays very important or only the total amount of rainfall?
    It is likely that we need to correct the data and the number of raindays and length of days without rain (climatic seasonality) is as important as rainfall amounts.

    #4191

    PeterBerg
    Participant

    A. Explain briefly what data you are planning to use for your case study and what data analyses and or (bias) – corrections you have already done or are planning to do

    We will run the HYPE model for the Niger basin, and will force it with seasonal forecasts. Two different bias correction methods will be used: (i) the DBS method as provided by the global climate service, and (ii) a copula based method by KIT that corrects precipitation and temperature to have the same joint distribution as the reference data (HydroGFD).

    B. To advise you on the most appropriate bias correction/adjustment and data analyses methods please answer the following key questions
    1. Which indicators / climate variables are you most interested in (temperature, precipitation, soil moisture, streamflow etc.)?

    Precipitation and temperature

    2. Are you interested in gridded data or point/station data?

    Gridded data

    3. What local data do you have available for bias correction ( e.g. station data on precipitation and temperature or a particular gridded data set)? For which time period, daily/monthly data?

    We will use the HydroGFD data provided by the global climate service at daily timescale.

    4. For your case are you only interested in changes in average values, or also in extremes (e.g. extreme rainfall, floods, droughts etc.)?

    We will look at e.g. the onset of the rainy season, and the total rainfall during the season.

    5. Do you need to correct precipitation data? If so, are the number of raindays very important or only the total amount of rainfall?

    The number of rainy days are important, and both methods will correct for this.

    #4198

    Jonas Olsson
    Participant

    Tokyo:

    A. Explain briefly what data you are planning to use for your case study and what data analyses and or (bias) – corrections you have already done or are planning to do

    We plan to use precipitation with analyses focusing on extremes. We will mainly look at relative future changes, without considering bias. But we also want to make some assessment of the impact of bias correction on the future change of extremes. We have access to bias-corrected GCM precipitation (against APHRODITE) through the DIAS climate service.

    B. To advise you on the most appropriate bias correction/adjustment and data analyses methods please answer the following key questions
    1. Which indicators / climate variables are you most interested in (temperature, precipitation, soil moisture, streamflow etc.)?

    We are primarily interested in sub-daily precipitation, but as we understand we will only have access to daily (GCM) precipitation in our domain. But it would be very good to have spatio-temporally downscaled as well as bias-corrected precipitation, if that can be provided.

    2. Are you interested in gridded data or point/station data?

    Both.

    3. What local data do you have available for bias correction (e.g. station data on precipitation and temperature or a particular gridded data set)? For which time period, daily/monthly data?

    We have very high-resolution precipitation observations from around 150 (?) stations in the Tokyo area for several decades (?).

    4. For your case are you only interested in changes in average values, or also in extremes (e.g. extreme rainfall, floods, droughts etc.)?

    Also extremes.

    5. Do you need to correct precipitation data? If so, are the number of raindays very important or only the total amount of rainfall?

    Only the extremes.

    #4215

    phil.graham@smhi.se
    Participant

    VIETNAM
    A. Explain briefly what data you are planning to use for your case study and what data analyses and or (bias) – corrections you have already done or are planning to do

    Nawapi has set up their own basin scale HYPE application for the Srepok River. They will use this to assess long-term climate change impacts from the future climate projections and short-term seasonal forecasts from the seasonal forecasting outputs. They have never worked with bias corrections Before.

    B. To advise you on the most appropriate bias correction/adjustment and data analyses methods please answer the following key questions
    1. Which indicators / climate variables are you most interested in (temperature, precipitation, soil moisture, streamflow etc.)?
    Daily temperature and precipitation will be used as input in the Srepok HYPE setup. Streamflow and soil moisture will be outputs.

    2. Are you interested in gridded data or point/station data?
    Station data.

    3. What local data do you have available for bias correction ( e.g. station data on precipitation and temperature or a particular gridded data set)? For which time period, daily/monthly data?
    Daily data from 13 observations stations have been used for calibration and validation for the period 1979-2015.

    4. For your case are you only interested in changes in average values, or also in extremes (e.g. extreme rainfall, floods, droughts etc.)?
    Averages and Extremes.

    5. Do you need to correct precipitation data? If so, are the number of raindays very important or only the total amount of rainfall?
    Yes and Yes!

    #4216

    Felix_van_Veldhoven
    Participant

    Climate Adaptation Services

    A. Explain briefly what data you are planning to use for your case study and what data analyses and or (bias) – corrections you have already done or are planning to do

    Developing suitability maps for relevant crops (barley/hops) by connecting to CRUCIAL data and develop water availability indicators using stream flow, runoff and specifically the temporal aspects for selecting the critical periods in the year.

    B. To advise you on the most appropriate bias correction/adjustment and data analyses methods please answer the following key questions

    1. Which indicators / climate variables are you most interested in (temperature, precipitation, soil moisture, streamflow etc.)?

    Streamflow, precipitation (extremes), temperature (extremes), soil moisture, crop specific indicators for barley and hobs (CRUCIAL)

    2. Are you interested in gridded data or point/station data?

    Mainly gridded data

    3. What local data do you have available for bias correction ( e.g. station data on precipitation and temperature or a particular gridded data set)? For which time period, daily/monthly data?

    None

    4. For your case are you only interested in changes in average values, or also in extremes (e.g. extreme rainfall, floods, droughts etc.)?

    Also in extremes

    5. Do you need to correct precipitation data? If so, are the number of raindays very important or only the total amount of rainfall?

    No

    #4218

    Guillermo Grau
    Participant

    OXFAM

    A. What data you are planning to use for your case study and what data analyses and or (bias) corrections you have already done or are planning to do?

    We are using GCM at daily (or more) resolution, both the historical and at least one projection. From CMIP5 (EC-EARTH): min, avg and max temperature, precipitation. Secondary inputs: droughts, dry spells. From Cordex: potential evapotranspiration, total soil moisture content
    From other sources: fixed soil attributes (field capacity, saturation point and wilting point), crop requirements. We intend to do bias correction with quantile mapping and we might be interested in some downscaling to slightly increase the resolution.

    B. To advise you on the most appropriate bias correction/adjustment and data analyses methods please answer the following key questions

    1. Which indicators / climate variables are you most interested in (temperature, precipitation, soil moisture, streamflow etc.)? Temperature, soil moisture, evapotranspiration. Given that it has been stated that soil moisture is not one of the high-quality outputs of the model, what could be an alternative approach based on other variables?

    2. Are you interested in gridded data or point/station data? Gridded data.

    3. What local data do you have available for bias correction (e.g. station data on precipitation and temperature or a particular gridded data set)? For which time period, daily/monthly data? ERA5 dataset at 30km spatial res, global coverage sub-daily temporal res, 2010-present. Or regional: West Africa (CORDEX AFR, 0.44° resolution) and Central America (CORDEX CAM, 0.44° resolution) for the bias correction and downscaling.

    4. For your case are you only interested in changes in average values, or also in extremes (e.g. extreme rainfall, floods, droughts etc.)? Changes in average values at daily resolution, and extreme events for temperature.

    5. Do you need to correct precipitation data? If so, are the number of raindays very important or only the total amount of rainfall? Only total amount of rainfall.

    #4219

    Guillermo Grau
    Participant

    UN Habitat

    A. What data you are planning to use for your case study and what data analyses and or (bias) corrections you have already done or are planning to do?

    We are using GCM at daily (or more) resolution, both the historical and at least one projection. From CMIP5 (EC-EARTH): min, avg and max temperature, precipitation. Also interested in (some of them already available): heatwaves, tropical nights, ice days, zero crossings of air temperature days, heating degree days, cooling degree days, insolation, short-duration extreme precipitation, snow cover, snow water equivalent, water runoff or discharge, growing season, droughts, floods.

    We are downscaling and bias-correcting these products (at least temperature and precipitation) with the MACA algorithm (Multivariate Regression by Combining selected Analogues) for 4 cities around the world: Maputo, Dakar, Asunción, Port Vila.

    B. To advise you on the most appropriate bias correction/adjustment and data analyses methods please answer the following key questions

    1. Which indicators / climate variables are you most interested in (temperature, precipitation, soil moisture, streamflow etc.)? Primarily temperature and precipitation. Secondary: others listed above.

    2. Are you interested in gridded data or point/station data? Gridded data.

    3. What local data do you have available for bias correction (e.g. station data on precipitation and temperature or a particular gridded data set)? For which time period, daily/monthly data? We have explored ERA5, but its resolution was not sufficient; we are working to source data from local authorities.

    4. For your case are you only interested in changes in average values, or also in extremes (e.g. extreme rainfall, floods, droughts etc.)? Extremes are also relevant (floods, droughts).

    5. Do you need to correct precipitation data? If so, are the number of raindays very important or only the total amount of rainfall? Number of raindays, as well as the number of consecutive raindays.

    #4220

    Guillermo Grau
    Participant

    PwC

    A. What data you are planning to use for your case study and what data analyses and or (bias) corrections you have already done or are planning to do?

    We are using GCM at daily (or more) resolution, both the historical and at least one projection. From CMIP5 (EC-EARTH): min, avg and max temperature, precipitation, number of lightnings, wind speed. From Cordex: daily max 1-hour precipitation rate (extreme). From other sources: soil characteristics (field capacity, saturation point, wilting point). We are discussing about the possibility of downscaling with MACA as in the UN Habitat use case.

    B. To advise you on the most appropriate bias correction/adjustment and data analyses methods please answer the following key questions

    1. Which indicators / climate variables are you most interested in (temperature, precipitation, soil moisture, streamflow etc.)? Temperature, precipitation, lightnings, wind speed.

    2. Are you interested in gridded data or point/station data? Gridded data.

    3. What local data do you have available for bias correction (e.g. station data on precipitation and temperature or a particular gridded data set)? For which time period, daily/monthly data? We are sourcing data from regional authorities and institutes: annual/monthly historical grids.

    4. For your case are you only interested in changes in average values, or also in extremes (e.g. extreme rainfall, floods, droughts etc.)? Extremes are crucial for this case study.

    5. Do you need to correct precipitation data? If so, are the number of raindays very important or only the total amount of rainfall? Precipitation probability distribution function, number of raindays, as well as the number of consecutive raindays

    #4221

    Bernard
    Participant

    AGRHYMET Regional Centre

    A. Explain briefly what data you are planning to use for your case study and what data analyses and or (bias) – corrections you have already done or are planning to do :

    The services that we have planned to produce are higher accuracy in seasonal forecasts in West Africa. This case study will use a CDS seasonal forecast ensemble (starting with ECMWF system 5). The climatic variables to be considered are daily precipitation, mean and extreme temperatures (minimum and maximum) at the seasonal scale. For this purpose, the quantile-quantile method was used to bias adjust the data. These data will then be used as climatic inputs of the Niger-HYPE model. The model, once implemented, will generate flows at the scale of the sub-basins of the Niger River.

    B. To advise you on the most appropriate bias correction/adjustment and data analyses methods please answer the following key questions
    1. Which indicators / climate variables are you most interested in (temperature, precipitation, soil moisture, streamflow etc.)?
    The climatic variables to be considered are daily precipitation, mean and extreme temperatures (minimum and maximum) at the seasonal scale.

    2. Are you interested in gridded data or point/station data?
    The aim is to use seasonal forecast gridded data over west africa to estimate climate parameters at the sub-watershed scale.

    3. What local data do you have available for bias correction ( e.g. station data on precipitation and temperature or a particular gridded data set)? For which time period, daily/monthly data?
    The NIGER-HYPE model was calibrated with WFDEI re-analysis data. Therefore, these were considered as reference datasets for bias correction. This correction was carried out at the daily scale because the NIGER HYPE model is at the daily time step.

    4. For your case are you only interested in changes in average values, or also in extremes (e.g. extreme rainfall, floods, droughts etc.)?
    For our case study, we are interested in assess predictive capacity: dry spells, accumulated volumes (runoff), extreme wet and dry events (e.g. maximum daily, 3-day and weekly rainfall and discharge during the season), assess predictive capacity at local scale.
    5. Do you need to correct precipitation data? If so, are the number of raindays very important or only the total amount of rainfall?
    The precipitation data correction is needed. Regarding the parameters we would like to forecast, the number of raindays is important.

    #4222

    CIIFEN
    Participant

    A. We plan to use monthly precipitation and temperature (maximum and minimum) data. Data will be the input for a statistical model. Not sure if we need to bias correct data for seasonal scale and for a categorical (terciles) outputs of the statistical model.

    B1. Precipitation and temperature (maximum and minimum).
    B2. Gridded data.
    B3. Stations data (precipitation, maximum and minimum temperature) at daily time scale from 1980-2010.
    B4. Extremes of rainfall and temperature.
    B5. If we need to correct precipitation data (not sure yet), normally the total amount of precipitation is provided to users but number of rainy days could be interesting and a better skill of the statistical model than the amount of rainfall.

Viewing 15 posts - 1 through 15 (of 17 total)

You must be logged in to reply to this topic.