d422lot1.smhi.5.1.1b: detailed workflows of each case
TRANSCRIPT
C3S_422_Lot1_SMHI – D5.1.1B |
Copernicus Climate Change Service
D422Lot1.SMHI.5.1.1B:
Detailed workflows of each case-study
on how to use the CDS for CII
production and climate adaptation
Full Technical Report:
Using seasonal forecast information to
strengthen resilience and improve food
security in West Africa
Bernard Minoungou, Dr Abdou Ali, Hamatan Mohamed
AGRHYMET Regional Centre
REF.: C3S_422_Lot1_SMHI D5.1.1B
C3S_422_Lot1_SMHI – D5.1.1B |
Copernicus Climate Change Service
Summary The case study focuses on improving the quality of seasonal forecast information to strengthen the resilience and improve the food security in West Africa,
especially on the Niger River Basin. We used seasonal climate forecasts and the HYPE hydrological model to predict some characteristics of the rainy season in
West Africa. The ECMWF seasonal forecast ensemble (system 5) from 1993 to 2015 (hindcast) and 2018 (forecast), available in the Climate Data Store (CDS) catalogue were used. The climatic variables considered are daily precipitation,
mean and extreme temperatures (minimum and maximum) at the seasonal scale. The main objective was to assess the ability of the HYPE hydrological
model to predict runoff over the historical period and to produce hydrological seasonal forecasts for 2018. The main season’s characteristics produced are: (i) cumulative rainfall map for the rainy season (May to November), (ii) map
showing the rainfall situation of the season (above, near or below normal considering 1993-2015 as reference period), (iii) graph of the mean seasonal
precipitation over the basin compared to the reference period (1993-2015), (iv) map showing the hydrological situation of the season (above, near or below normal considering 1993-2015 as reference period), (v) graph of the mean
seasonal streamflow over the Niger Basin compared to the reference period (1993-2015), (vi) graphs comparing monthly streamflow of 2018 to those of the
reference period (1993-2015). We organized a workshop to share and discuss produced indicators with clients.
These clients represented the national hydrology services of four countries in the Niger River Basin (Burkina Faso, Mali, Niger and Nigeria). This study shows that bias correction of climate data, which are the main inputs
to hydrological rainfall-runoff models, is a very important step and requires appropriate methods. Initialization of hydrological models was an important
phase in the study.
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B |
Contents
1- Case study description ........................................................................................................ 4
1.1 Issue to be addressed ......................................................................................................... 4
1.2 Decision support to client .................................................................................................. 4
1.3 Temporal and spatial scale ............................................................................................... 4
1.4 Knowledge brokering ........................................................................................................... 4
2- Potential adaptation measures ........................................................................................ 4
2.1 Lessons learnt ........................................................................................................................ 4
2.2 Importance and relevance of adaptation .................................................................... 5
2.3 Pros and cons or cost-benefit analysis of climate adaptation ............................. 5
2.4 Policy aspects ......................................................................................................................... 5
3- Contact ...................................................................................................................................... 6
3.1 Purveyors ................................................................................................................................. 6
3.2 Clients/users ........................................................................................................................... 6
4- Data production and results ............................................................................................. 6
4.1 Step 1: Data collection from CDS, quality assessment and bias adjustment8
4.2 Step 2: Hydrological forecasting................................................................................... 14
4.3 Step 3: Evaluate predictive capacity ........................................................................... 20
4.4 Step 4: Revised ensemble ............................................................................................... 20
4.5 Step 5: Stakeholder communication ........................................................................... 21
5- Conclusion of full technical report ................................................................................ 22
References ......................................................................................................................................... 24
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 4
1- Case study description
1.1 Issue to be addressed
The rainy season occupies a central place in socio-economic activities in the
Sahelian regions, as more than 80% of the population lives on agriculture and livestock. However, extreme hydroclimatic events such as droughts and floods affect these activities. Efforts made in recent years in the production of
hydroclimatic information to enhance the resilience of populations have become insufficient, given the variability and climate change.
1.2 Decision support to client
The clients, that are the national hydrology departments, received the
information produced, mainly the daily, monthly and seasonal cumulative rainfall and the daily, monthly and seasonal streamflow within River Niger basin. These information are usually used to support the decision of dam managers (to better
plan dam operations), farmers (to better plan their activities during the rainy season), and society (on the potential risks of floods).
1.3 Temporal and spatial scale
This work is focused on seasonal hydrological forecasts. These forecasts covered
the Niger River basin, which has an area of 2 170 000 km2. The Niger Basin was subdivided into 803 sub-basins and the information was produced at the scale of these sub-basins.
1.4 Knowledge brokering
The AGRHYMET Regional Centre, in collaboration its partners, organizes the
PRESAO (Seasonal Forecast over West Africa) forum for the elaboration of seasonal hydrological forecasts in the beginning of May every year. This forum brings together all the national hydrology and meteorological services, basin
organizations and the departments in charge of disaster and risk management. The information purveyors therefore have a knowledge of the needs of the clients
in terms of information to be provided to them. The interactions with customers are done each year through a forum. As part of the C3S_422_Lot1_SMHI contract, we held a physical meeting with customers to share with them the
products from the case study.
2- Potential adaptation measures
2.1 Lessons learnt
A similar study previously conducted forcing the HYPE hydrological model with
the ECMWF System 4 seasonal forecast indicates that the quality of seasonal hydrological forecasts may be improved. Accessibility to data and methodology
was made possible thanks to the collaboration between the AGRHYMET Regional Center and its partner institutions including SMHI. Interactions between institutions producing climate data and large-scale tools and local institutions
allow their practical use to solve recurring societal problems in our regions such
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 5
as adaptation to extreme events (floods and droughts), improvement of
agricultural production. It is also important to point out that some planned activities are not always carried out due to lack of technical resources. This is the
case, for example, of the Distribution Based Scaling (DBS) bias correction method that was originally planned for this study.
2.2 Importance and relevance of adaptation
In West Africa, climate variability and change were assessed on the basis of indigenous knowledge. For example, the behavior of some birds is used to
determine the start of the rainy season. The efforts made in recent years and the current outlook for seasonal forecasting are helping to better qualify and quantify
the physical processes responsible for the different characteristics of the rainy season in West Africa. For instance, in agriculture, an evaluation showed that agrometeorological assistance through the provision of climate services
contributed to an increase in agricultural production in Mali of about 30% (Faustin Gnoumou, personal communication). Until now, agro-hydro-
meteorological assistance to end-users through seasonal forecasting has been based on purely statistical methods. With the new approach (using deterministic predictions) that is expected to be developed, resilience to climate variability and
change is likely to be enhanced. We will be able to provide local characteristics such as the periods of occurrence of extreme hydrological events, their amplitude
and the availability of water resources and associated margins of error.
2.3 Pros and cons or cost-benefit analysis of climate adaptation
Studies have shown that end users can benefit from climate change adaptation in terms of increased income and reduced risk through the use of seasonal forecasting despite the uncertainties associated with current forecasts. Generally,
investments that prevent climate-related disasters are much lower cost than the economic and human losses caused by such disasters. One of the major
shortcomings of previous hydrological forecasts (statistical approach) is that they did not meet users' expectations for assessing water resources for decision support. As part of this study, it was possible to produce quantitative
hydrological forecasts on the availability of water resources at the Niger Basin scale.
2.4 Policy aspects
In the past, adaptation strategies have focused on managing crises rather than preventing them. With the increase in extreme weather events and the high cost
of crisis management, public opinion has become aware of the importance of climate change adaptation measures in anticipating such events. The activity of
seasonal hydrological forecasts in West Africa is already included in policies to strengthen resilience to climate change and variability. However, the previous
forecasts no longer sufficiently met the expectations of increasingly demanding users and decision-makers. Thus improved forecasts better met existing policies.
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 6
3- Contact
3.1 Purveyors
Abdou Ali, Minoungou Bernard and Hamatan Mohamed AGRHYMET Regional Centre, Niamey, Niger
3.2 Clients/users
Nakohoun Lokou Pascal, Direction des Etudes et de l’Information sur l’Eau,
Ouagadougou, Burkina Faso Sidibé Ibrahima, Direction Nationale de l'Hydraulique du Mali, Bamako, Mali
Housseini Mohamed, Direction Nationale de l'Hydraulique, Niamey, Niger Olorunoje Addi Shuaib, Nigeria Hydrological Services Agency (NIHSA), Abuja,
Nigeria
4- Data production and results The aim of this study is to produce hydrological seasonal forecasts using the
Hydrological Prediction for Environment (HYPE) model, when forced with the System 5 seasonal forecast from ECMWF. The main steps of the study included:
Data collection: The data collected include ECMWF seasonal forecast ensemble (system 5) available in the CDS catalogue, WFDEI-reanalysis data, river discharge station data;
Quality assessment and bias adjustment by considering WFDEI-Reanalysis data as reference data. The quantile-quantile method was used for bias
correction; Forecasting river discharge with HYPE for Niger River: streamflow were
forecasted by forcing the Niger-HYPE model with ECMWF seasonal climate
forecasts at the scale of the rainy season in the Sahel (May-November). To properly characterize the initial conditions of the hydrological model, the
model was run over the historical period with the WFDEI re-analysis data. The hydrological forecast were done by considering first the ECMWF system 5 raw seasonal forecast and the ECMWF system 5 bias adjusted
seasonal forecast; The assessment of the predictive capacity;
The ensembles are revisited by putting less weight on low-performing ensemble members;
Interactions with clients at the regional level.
Area of study, observational data and hydrological model
The basic requirement to produce operationally hydrological seasonal forecast is
that all data sources are operationally available, which means that the data is
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 7
continuously updated near real-time. The input data selected could be grouped
as hydrological measurements, climate and reanalysis data as well as seasonal
climate forecasts.
We used a semi-distributed, process-oriented, rainfall-runoff model called HYPE
(Lindström et al., 2010). The HYPE model was setup for the Niger River Basin
using available datasets on topography (Lehner et al., 2006), land use (JRC,
2003), soil (Batjes, 2012; FAO et al., 2009), lakes (Lehner and Döll, 2004), and
reservoirs (Lehner et al., 2011). The model uses the WFDEI reanalysis dataset
as climate forcing for daily precipitation, mean temperature, minimum,
temperature and maximum temperature (Weedon et al., 2012). The potential
evapotranspiration was estimated from daily minimum and maximum
temperature using the equation by (Hargreaves and Samani, 1982). The 2.1
million km2 river basin (figure 1) was divided into 803 linked sub-basins based on
the topography and derived drainage direction. The sub-basin size was chosen to
fit the coarseness of the available forcing data (0.5°x0.5°). Each sub-basin was
further divided into areas of homogeneous soil and land cover (in total 57 unique
hydrologic response units). The 48 largest reservoirs and lakes in the basin were
included in the model. A tailor-made extension to the HYPE model was developed
in order to simulate the Inland Niger Delta wetland. The principle was to allow
flooding of incoming discharge onto a dynamic floodplain area (growing and
shrinking in size with water stage) on which precipitation and evaporation occurs.
Part of the water on the floodplain can also return to the river system as the
flood is receding.
Figure 1: Selected key input data for the Niger-HYPE 2.0 model in the Niger River Basin.
Note that although the basin stretches into Sahara, the hydrologically active part of the basin is limited by the desert in the North
The model was calibrated for the period 1994-01-01 to 2009-12-31 against 56
discharge stations spread across the hydrologically active part of the basin,
(ABN, 2008; GRDC, 2012); and against potential evapotranspiration from the
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 8
MODIS satellite (Mu et al., 2007). A conjunctive multi-objective calibration
approach was employed in order to obtain a balanced model for both river flow
and evaporation. The parameters in each sub-basin were not calibrated to each
gauging station. Instead they were linked to soil type and land cover type, and
applied throughout the basin. As such the model was calibrated against all
stations simultaneously (to assist in making predictions in ungauged basins). The
calibration objective was hence to obtain a balanced model performance at all
stations simultaneously with this single parameter set, rather than to obtain
optimal performance at any given station. The period 1979-01-01 to 1993-12-31
was used for independent validation.
4.1 Step 1: Data collection from CDS, quality assessment and bias
adjustment
Description
Data collection: Seasonal climate forecasts available from the C3S and the CDS
catalogue were used to force the Niger-HYPE model. In particular, ECMWF
System 5 seasonal forecasts (S5) was used. A seasonal forecast is produced
each month. The forecasts have an initial date of the 1st of each month, and run
for 7 months. Forecast data and products are released at 12Z UTC on a specific
day of the month. For SEAS5, this is the 5th. While SEAS5 is expected to be
operational from 1 Nov 2017, forecasts for 1 Jan 2017 through to 1 Oct 2017
have also been completed and archived for reference.
Quality control: To force the HYPE model, climate data are estimated at the
scale of the 803 sub-watersheds in the Niger basin. The data are available on
regular grids and it was first necessary to analyze their consistency once
estimated on the subwatersheds. We compared the seasonal climate forecast
data with the reference data used to calibrate the HYPE model.
Bias adjustment: Bias correction uses historical WFDEI re-analysis data to
readjust the simulations from the seasonal forecasts (ECMWF system 5). This
correction uses a statistical transformation whose goal is to find a function h
which considers the data of the model as being equal to the distribution of the
data observed according to the following formula:
𝑃𝑜 = ℎ(𝑃𝑚) (1)
𝑃𝑜 = 𝑂𝑏𝑠𝑒𝑟𝑣𝑒𝑑 𝑑𝑎𝑡𝑎
𝑃𝑚 = 𝑆𝑖𝑚𝑢𝑙𝑎𝑡𝑒𝑑 𝑑𝑎𝑡𝑎
In practice, one must first find a common period for both data sources. The
WFDEI historical data, used to calibrate the HYPE model, being available over the
period 1979-2013 and those of the seasonal forecasts over the historical period
1993-2015, the period 1993-2013 was considered for the correction of bias.
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 9
The statistical transformation is an application of the probabilistic integral
transformation (Angus 1994) and if the distribution of the variable of interest is
known, the transformation is defined as:
𝑃𝑜 = 𝐹0−1(𝐹𝑚(𝑃𝑚)) (2)
𝑤ℎ𝑒𝑟𝑒 𝐹𝑚 𝑖𝑠 𝑡ℎ𝑒 𝑐𝑢𝑚𝑢𝑙𝑎𝑡𝑖𝑣𝑒 𝑑𝑖𝑠𝑡𝑟𝑖𝑏𝑢𝑡𝑖𝑜𝑛 𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛 (𝐶𝐷𝐹) 𝑜𝑓 𝑃𝑚
𝐹0−1 𝑖𝑠 𝑡ℎ𝑒 𝑖𝑛𝑣𝑒𝑟𝑠𝑒 𝑜𝑓 𝑡ℎ𝑒 𝐶𝐷𝐹 𝑐𝑜𝑟𝑟𝑒𝑠𝑝𝑜𝑛𝑑𝑖𝑛𝑔 𝑡𝑜 𝑃𝑜.
The practical goal is to find the best approximation of h and different approaches
have been proposed in the literature. The method chosen to solve equation (2)
was to use the CDFs of observed and simulated values. Following the procedures
of Boé et al (2007), CDFs are estimated using empirical percentile tables. Values
between percentiles are obtained by linear interpolation. The "qmap" package
developed under the R language contains this method of bias correction. This is
what we used in this study.
Results Data collection: ECMWF SEAS5 hindcast and forecast were collected. Historical
data range from 1993 to 2015 while the 2018 forecast was considered. Each
year, data cover West Africa over the period from May to November. The 25
ensembles member were used.
Quality control: The figures below represent the parameters estimated at the
level of the sub-basins of the Niger River. Considering the outputs of the model
initialized on April of each year, the season covers the period from May to
November of each year. Over this period, the seasonal average rainfall over the
period 1993-2013 varies from 5 mm in the North to over 3280 mm in the South,
considering the average of the 25 ensemble member used (see figures 2 and 3).
WFDEI mean inter-annual rainfall varies from 28 mm in the North to 2383 mm in
the South. Figure 4 below shows the absolute difference between the average
inter-annual rainfall over the period 1993-2013 and that of system 5. This
corresponds to a relative error of -80 to 80% over the entire Niger basin between
re-analysis data and system 5 data (Figure 4). The mean temperature varies
from 20.62 ° C to 33.62 ° C, over the period 1993-2013 for the re-analysis data
and between 17.92 ° C and 31.70 ° C for the average of 25 member sets at the
same time. The relative error between these two data sources ranges from -4 to
0.5 ° C (Figure 7). The inter-annual comparison of the two data sources is
represented by the figures 8 and 9. Figures 10 and 11 indicate the PBIAS for
comparison between the two datasets.
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 10
Figure 2: Seasonal average (1993-2013) precipitation
(mm) from WFDEI reanalysis data
Figure 3: Seasonal average (1993-2013) precipitation
(mm) from ECMWF system 5 data (mm)
Figure 4: Relative error
between WFDEI and ECMWF System 5 Precipitation
Figure 5: Seasonal average
(1993-2013) temperature (°C) from WFDEI reanalysis data
Figure 6: Seasonal average
(1993-2013) temperature (°C) from ECMWF system 5 data
Figure 7: Relative error
between WFDEI and ECMWF System 5 temperature
Figure 8: Temporal evolution of WFDEI precipitation compared to precipitation of ECMWF system 5 (before bias correction)
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 11
Figure 9: Temporal evolution of WFDEI temperature compared to temperature of ECMWF system 5 (before bias correction). Y axis is Temperature in degree Celsius.
Figure 10: PBias (in %) between WFDEI
precipitation and ECMWF System 5 Precipitation
Figure 11: PBias (in %) between WFDEI
temperature and ECMWF System 5 temperature
Bias correction: The figures (12-15) represent the quantile-quantile plots of the
data from the ECMWF system 5 (before and after bias correction) according to
those of the WFDEI data. The first ensemble of seasonal forecast data from
models initialized on May 1 were used to illustrate the results obtained. After bias
correction, the inter-annual comparison of the two data sources is represented
by the figures 16 and 17. Figures 18 and 19 indicate the PBIAS for comparison
between the two datasets. It is realized that the bias correction has improved the
quality of the temperature data. This is not the case for precipitation.
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 12
Figure 12: Q-Q Plot of WFDEI Precipitation vs ECMWF system 5 Precipitation corrected and
uncorrected (corrected in green and uncorrected in red)
Figure 13: Figure 12: Q-Q Plot of WFDEI
maximum temperature vs ECMWF system 5 maximum temperature (corrected in green and
uncorrected in red)
Figure 14: Figure 12: Q-Q Plot of WFDEI
minimum temperature vs ECMWF system 5 minimum temperature (corrected in green and
uncorrected in red)
Figure 15: Figure 12: Q-Q Plot of WFDEI mean
temperature vs ECMWF system 5 mean temperature (corrected in green and
uncorrected in red)
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 13
Figure 16: Temporal evolution of WFDEI precipitation compared to precipitation of ECMWF system 5 (after bias correction)
Figure 17: Temporal evolution of WFDEI temperature compared to temperature of ECMWF system 5 (before bias correction)
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 14
Figure 18: PBias (in %) between WFDEI
precipitation and ECMWF System 5 Precipitation (after bias correction)
Figure 19: PBias (in %) between WFDEI
temperature and ECMWF System 5 temperature (after bias correction)
4.2 Step 2: Hydrological forecasting
Description:
Once climate data, i.e., precipitation and daily seasonal temperatures, were
available, it was used to force the HYPE model with these data. To properly
characterize the initial conditions of the model, the model was run over the
historical period with the WFDEI re-analysis data.
Results
Figures 20 to 24 show the results for simulated streamflow at 5 stations for daily
values during the period from 1993-2013. As a preliminary evaluation, we
computed the percent bias (PBias) and the (KGE) for the complete observed-
simulated record (1993-2013). We consider as “observed” data here, the
discharges obtained by forcing HYPE with the WFDEI re-analysis data.
Figure 20: Comparison of the simulated
(orange) and simulated with WFDEI (blue) long-term (1993-2013) annual mean
streamflow at Koulikoro station
Figure 21: Comparison of the simulated
(orange) and simulated with WFDEI (blue) long-term (1993-2013) annual mean
streamflow at Lokoja station
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 15
Figure 22: Comparison of the simulated
(orange) and simulated with WFDEI (blue) long-term (1993-2013) annual mean
streamflow at Makurdi station
Figure 23: Comparison of the simulated
(orange) and simulated with WFDEI (blue) long-term (1993-2013) annual mean
streamflow at Niamey station
Figure 24: Comparison of the simulated (orange) and simulated with WFDEI (blue) long-term (1993-2013) annual mean streamflow at Sirba station
The figures below give KGE and PBIAS on the main hydrological stations
considered. For example for Koulikoro station some years are clearly different
whereas for Niamey the performance is not really related to any given year. To
improve the quality of the model, investigations must be conducted to better
understand the initial conditions considered by the model.
Figure 25: KGE between simulated with WFDEI
and simulated discharge at Koulikoro station
Figure 26: PBIAS (in %) between simulated with
WFDEI and simulated discharge at Koulikoro station
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 16
Figure 27: KGE between simulated with WFDEI
and simulated discharge at Lokoja station
Figure 28: PBIAS (in %) between simulated with WFDEI and simulated discharge at Lokoja station
Figure 29: KGE between simulated with WFDEI
and simulated discharge at Makurdi station
Figure 30: PBIAS (in %) between simulated with
WFDEI and simulated discharge at Makurdi
station
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 17
Figure 31: KGE between simulated with WFDEI
and simulated discharge at Niamey station
Figure 32: PBIAS (in %) between simulated with
WFDEI and simulated discharge at Niamey station
Figure 33: KGE between simulated with WFDEI
and simulated discharge at Garbey Kourou a station
Figure 34: PBIAS (in %) between simulated with
WFDEI and simulated discharge at Garbey Kourou station
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 18
Forecast for 2018: After evaluating the quality of seasonal forecasts over the historical period, the
forcing of the model also produced the 2018 seasonal hydrological forecasts. Some derived products are represented by the figures 37, 39 and 41 and graphs 38, 40 and 42 below.
In figures 37 and 39, for “Compare to normal conditions”, green (yellow) [orange] colors indicate the probability of forecasts being above (near) [below]
normal conditions for the forecast period (May to November). The reference data is based on system5. The thresholds to define the normal conditions are the 66th and 33rd percentiles for the seasonal averages as these
are derived from the HYPE model simulation for the period 1993-2015.
The graphs of figure 40 showed one seasonal forecast at monthly time step from May to November.
Figure 35: Seasonal (May to November) total rainfall in the Niger River Basin from ECMWF climate seasonal forecasts in
2018
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 19
Figure 36: Graph of the mean seasonal precipitation over the river Niger basin compared to the reference period
Figure 37: Seasonal precipitation forecast of 2018
compared to reference
Figure 38: Graph of the mean seasonal streamflow over the river Niger basin compared to the reference period
Figure 39: seasonal streamflow forecast of 2018 compared
to reference
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 20
Figure 40: 2018 monthly streamflow (May to November) in some stations
4.3 Step 3: Evaluate predictive capacity
Description
It was a question of checking whether the forecasting of the characteristics of the
rainy season corroborated with the observations.
Results
The flows were above normal on all the hydrometric stations situated on the
main course of the Niger River or on the outlet of the large sub-basins. These observations match well with those of the seasonal hydrological forecasts of
2018. However, given the problems of model initialization and bias correction, forecasts should mainly be improved on small sub-watersheds with a fairly rapid
response.
4.4 Step 4: Revised ensemble
Description
It is to construct a revised ensemble, by putting less weight on low-performing
ensemble members. As a reminder, 25 ECMWF ensemble members were used to
produce seasonal hydrological forecasts.
Results
Figures 25 to 34 show the performance of each ensemble member to reproduce
the flow over the historical period considered (1993-2015). These performance
criteria show that all ensemble member have almost the same predictability
capabilities. Given this fact, one way of representing the products would be to
consider the characteristic quantiles of flows, obtained from the 25 ensemble
member (minimum, first quantile, median, third quantile and maximum). The
graphs of figure 41 below show the performance of the model obtained taking
into account the mentioned quantiles.
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 21
Figure 41: Evaluation of the model's performance by considering the quantiles of streamflow
4.5 Step 5: Stakeholder communication
Description
The communication took place through a workshop organized from 20 to 21 2018 December in Niamey (Niger). It saw the participation of the representatives of the national hydrology departments of four countries of the Niger River basin
(Burkina Faso, Mali, Niger and Nigeria). The main objectives of this workshop were to share and get feedback from clients on the new hydrological products of
seasonal forecasts developed as part of the C3S_422_Lot1_SMHI contract implementation.
Results
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 22
Some points discussed during this workshop include:
Presentation of the C3S_422_Lot1_SMHI contract; Presentation of the seasonal forecast approach using hydrological rainfall-
runoff modeling; Presentation of new seasonal hydrological forecast products followed by
exchanges;
Presentation of the interactive atlas.
Figure 42: Group photo of workshop participants
Figure 43: Presentation session of the interactive atlas
Clients said that using climate seasonal data for hydrological rainfall-runoff
model meets their needs more than using old statistical methods of seasonal forecasting. The main suggestions made would be to consider the season's
analyzes according to the flood return periods specific to each hydrometric station. They also propose to define warning thresholds from which potentially floodable areas can be identified from the results of forecasts.
5- Conclusion of full technical report
Seasonal hydrological forecasts are part of the annual activities of the AGRHYMET Regional Center for more than 20 years. They allow to have a global vision of the
rainy season, necessary for a better planning of the agro-hydro-pastoral activities. The use of Copernicus Climate Change Service data (seasonal climate forecasts) as input for the HYPE model has provided more information to end-
users to support decision-making. The methodology developed in this case study will be improved and will be used for annual hydrological seasonal
forecasting in the West African region. That allowed us to produce the following information: i) cumulative rainfall map for the rainy season (May to November), (ii) map showing the rainfall situation of the season (above, near or
below normal considering 1993-2015 as reference period), (iii) graph of the mean seasonal precipitation over the basin compared to the reference period
(1993-2015), (iv) map showing the hydrological situation of the season (above, near or below normal considering 1993-2015 as reference period), (v) graph of the mean seasonal streamflow over the Niger Basin compared to the reference
period (1993-2015), (vi) graphs comparing monthly streamflow of 2018 to those of the reference period (1993-2015).
The meeting with the clients showed that using climate seasonal data for hydrological rainfall-runoff model meets their needs more than the methods previously used (empirical methods of seasonal forecasting). The
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 23
predictability of 2018 hydrological seasonal products were assessed. The main
challenges we faced were the initialisation of the model, the bias correction (the reference data to be considered and the appropriate method). Investigations in
these areas should continue to improve the quality of results.
Copernicus Climate Change Service
C3S_422_Lot1_SMHI – D5.1.1B | 24
References
Batjes, N.H., 2012. ISRIC-WISE derived soil properties on a 5 by 5 arc-minutes
global grid (ver. 1.2) ( No. 2012/01). ISRIC - World Soil Information, Wageningen, The Netherlands.
FAO, IIASA, ISRIC, ISS-CAS, JRC, 2009. Harmonized World Soil Database (version 1.1). FAO and IIASA, Rome, Italy.
Gudmundsson, L., Bremnes, J. B., Haugen, J. E, Engen-Skaugen, T., 2012.
Technical Note: Downscaling RCM precipitation to the station scale using
statistical transformations – a comparison of methods. Hydrol. Earth Syst. Sci.,
16, 3383–3390, 2012 www.hydrol-earth-syst-sci.net/16/3383/2012/
doi:10.5194/hess-16-3383-2012
Hargreaves, G.H., Samani, Z.A., 1982. Estimating potential evapotraspiration. J Irrig Drain Engrg 108, 225–230.
JRC, 2003. Global Land Cover 2000 database [WWW Document]. URL http://bioval.jrc.ec.europa.eu/products/glc2000/glc2000.php (accessed 4.1.12).
Lehner, B., Döll, P., 2004. Development and validation of a global database of
lakes, reservoirs and wetlands. J. Hydrol. 296, 1–22.
doi:10.1016/j.jhydrol.2004.03.028
Lehner, B., Liermann, C.R., Revenga, C., Vörösmarty, C., Fekete, B., Crouzet, P.,
Döll, P., Endejan, M., Frenken, K., Magome, J., Nilsson, C., Robertson, J.C., Rödel, R., Sindorf, N., Wisser, D., 2011. High-resolution mapping of the world’s reservoirs and dams for sustainable river-flow management. Front. Ecol. Environ.
9, 494–502. doi:10.1890/100125
Lehner, B., Verdin, K., Jarvis, A., 2006. HydroSHEDS Technical Documentation.
World Wildlife Fund US, Washington, DC, U.S.A
Lindström, G., Pers, C., Rosberg, J., Strömqvist, J., Arheimer, B., 2010.
Development and testing of the HYPE (Hydrological Predictions for the
Environment) water quality model for different spatial scales. Hydrol. Res. 41,
295–319. doi:10.2166/nh.2010.007
Weedon, G.P., Gomes, S., Balsamo, G., Best, M.J., Bellouin, N., Viterbo, P., 2012. WATCH forcing data based on ERA-INTERIM [WWW Document]. URL ftp://rfdata:[email protected] (accessed 12.15.12).
Weedon, G. P., G. Balsamo, N. Bellouin, S. Gomes, M. J. Best, and P. Viterbo
(2014), The WFDEI meteorological forcing data set: WATCH Forcing Data
methodology applied to ERA‐Interim reanalysis data, Water Resour. Res., 50,
7505–7514, doi: 10.1002/2014WR015638