Page 143 - AJWEP-v22i3

Projected trends in extreme heat in Senegal from 2020 to 2080

Dehban et al., Mmame and Ngongondo, and Zareian approach has been employed by several authors,
42
32
et al., have utilized these methods to evaluate the including Goudiaby et al. and Bodian et al. In this
43
49
48
models’ ability in reproducing reference data. KGE is study, the 95 percentile was selected as the threshold
th
a performance metric that indicates the model’s ability for detecting heat waves.
to reproduce an observed or reference time series. It is
calculated based on the correlation coefficient (r), biases 2.3.3. Spatialization of extreme heat
(β), and variability (γ) using the following formula: The inverse distance weighted (IDW) interpolation
method was used to spatialize the results at the scale
2
1
KGE 1 ( r 1) 2 β γ 1 2 (I) of Senegal. This is a robust method that allows the
estimation of values across a surface based on known
With. β sim , sim data points. It is based on the principle that each
obs obs measuring point influences its surrounding area, with the
Where: influence decreasing as distance increases. This method
was used to map the 95 percentile across Senegal.
th
σ = standard deviation
μ = mean 2.3.4. Heat wave anomalies
sim = simulation data The Lamb index (1982) was applied to identify interannual
obs = observation data anomalies of extreme heat in Senegal for the periods 2021
Percentage Bias is used to evaluate model – 2050 and 2051 – 2080. This index helps assess the
performance by comparing simulated data to reference interannual variability of extreme temperatures and detect
or observed data. It accounts for systematic bias in fluctuations in hydro-climatic regimes and climatic
50
simulated data and indicates the extent to which the phenomena. 28,51 A positive index indicates an increase in
model overestimates or underestimates the observed temperatures, while a negative index reflects a decrease. 30
data. The closer the pBias value is to zero, the more x x �
accurately the model reproduces the reference data. It is IL�� i� m (III)
calculated using the following formula:

n Qsim i Qobs i where:
[]
pBiais 100 (II) IL = Lamb index
n
i1 Qobs i () x = 95 annual percentiles for a station during a year
th
i1
i
x = annual average of the 95 percentile of the mean
th
m
where: temperature at the station during the study period.
Q = values at time step The trend of extreme heat was obtained using
n = number of time step the Mann-Kendall test, 52,53 a non-parametric test
that identifies trends in a given time series. The null
2.3.2. Data resampling and extraction hypothesis for this test is the absence of a trend. When
Canadian Earth System Model Version 5 data have been the null hypothesis is rejected, the alternative hypothesis,
uploaded to the WCRP platform (https://esgf-node. which indicates the existence of a trend, is accepted,
llnl.gov/search/cmip6/) in netCDF format. They were depending on the level of significance. The Mann-
54
then resampled to reduce the size of the grids. This Kendall test provides three key pieces of information:
55
operation reduced the data to a spatial resolution of (i) The Kendall tau or Kendall’s rank coefficient, which
0.0625° × 0.0625° per grid. The data were then extracted measures the tendency of the slope. The positive or
based on the coordinates of the 40 grid points selected negative rate reflects an upward or downward trend,
for this study (Figure 1), and the 95 annual percentiles respectively. The formula is as follows:
th
of the SSP1-2.6, SSP2-4.5, and SSP6-8.5 scenarios Var (s) = n(n - 1)(2n + 5)/18 (IV)
were calculated. In previous studies, Da Silva et al. where
44
and Teegavarapu et al. demonstrated the effectiveness
45
of this approach in capturing the spatial variability of S jn1 in signxi ( xj) (V)
climatic variables in the study area using simulated j1 ij1
data. Among the available interpolation methods, we where
utilized bilinear interpolation, 46,47 implemented through sign (xi – xj) = {1 if (xi - xj) > 0; 0 if (xi - xj) = 0; −1
the Climate Data Operator software. This resampling if (xi - xj) < 0 (VI)

Volume 22 Issue 3 (2025) 137 doi: 10.36922/AJWEP025150107