Evolution of Agroclimatic Indicators in Senegal Using CMIP6 Simulations

Abstract

Climate has a strong influence on agriculture, which is considered the most dependent human activity on climate variations. The future performance of the Senegalese agricultural sector will depend on its ability to adapt to the negative impacts of climate change. This study demonstrated that the impact of three climate change scenarios (ssp126, ssp245 and ssp585) on the evolution of 14 agro-climatic indicators is already evident in Senegal in the near and distant future. Indeed, the results obtained show a generalized decrease over the whole country in seasonal rainfall totals of about −10% in the near future (2020–2049; PSE horizon) up to −40% in the distant future (2070–2099) for the ssp585 scenario. This decrease in precipitation will be associated with two phenomena, namely a shortening of the rainy season due to increasingly late starts and an increase in dry spells, particularly the DSl and DSxl. The other trend observed is an increase in the frequency and intensity of extreme rainfall events (R99 and R20), which illustrates an increasingly chaotic distribution of rain in the future. Finally, this characterization of agroclimatic indicators made it possible to evaluate and classify the sensitivity of four global models corrected by the CFD-t method in order to run agronomic simulations and to explore adaptation strategies for farmer management in the future.

1. Introduction

Climate change is now considered one of the major challenges facing the world’s populations [1]. Most sectors of human activity on which development efforts are focused are climate-sensitive. The agricultural sector is one of the most vulnerable sectors to climate change [2, 3, 4, 5, 6]. Indeed, the Sahelian zone is already subject to high spatiotemporal variability in rainfall. The Sahel experienced continuous drought from the late 1960s to the mid-1990s and early 2000s [7, 8, 9, 10]. During these decades of drought, the isohyets lost between 20 and 50% of their annual cumulative rainfall compared to the 1950s and 1960s. Precipitation has also decreased between 20 and 40%, as well as average temperatures have increased by 1.3°C [11]. This has been described as the most negative trend in precipitation in the contemporary world [12]. The impacts on the populations of sub-Saharan Africa have been very significant with per capita food production falling sharply from 98% in the 1960s to about 86% in the mid-1980s [13, 14]. On average, each inhabitant had lost 12% of the food grown in the 1980s, when drought was at its peak [15]. Climate change will exacerbate these impacts on agriculture. There is growing evidence that changes in climate parameters, namely precipitation, temperature and the intensity and frequency of extreme events, will affect the agricultural sector in several ways [16]. Increased temperatures and the possibility of more extreme thermal events will impact crop productivity. Heading is one of the most sensitive phenological stages to temperature extremes for all cereal species, and during this stage of development, temperature extremes will significantly affect crop production [16, 17]. While increased pockets of drought during the rainy season will limit water availability, thereby promoting soil and groundwater salinization ultimately reducing crop area [18, 19]. Studies have also shown a proliferation of parasites harmful to plants and an increasing incidence of pest attacks which could be detrimental to agricultural production [20]. Given this context of high exposure to the impacts of climate change, the implementation of adaptation measures is an imperative issue for the agricultural sector now and in the future. Thus, several agro-climatic indicators have been created so that agricultural actors can better understand the effects of climate change on their production systems and prepare the necessary adaptations. However, most of the studies that evaluate these agro-climatic indicators focus on global or regional scales. The objective of this paper is to quantify the impacts of global warming on 14 selected agroclimatic indicators in Senegal. Among these indicators, we have the start of the rainy season, different categories of dry sequences and the frequency and intensity of extreme rainfall events. The study was carried out through a spatiotemporal assessment of changes in these agro-climatic indicators in the near (2020–2049) and distant (2070–2099) future in relation to the historical period (1985–2014) for the Shared Socioeconomic Pathways scenarios (ssp126, ssp245 and ssp585).

2. Data and methods

2.1 Data

In this study, four CMIP6 general circulation models (Table 1; [25]) and three greenhouse gas emission scenarios known as Shared Socioeconomic Pathways (SSPs; [26, 27]) (SSP1-2.6, SSP2-4.5, and SSP5-8.5,) were used. The CMIP6 GCM runs were developed in support of the Intergovernmental Panel on Climate Change Sixth Assessment Report (IPCC AR6). Each of the climate projections includes daily precipitation and temperature variables for the periods 1950–2014 (“hindcast”) and 2015–2100 (“forwardcast”). A bias reduction and correction procedure, described in the following paragraph, was applied to the historical and future climate projections to provide a set of climate change projections at high resolution (0.25°) relative to their native resolution that can be used to assess the impacts of climate change on processes that are sensitive to smaller-scale climate gradients and the effects of local topography on climate conditions.

No	Models	Institutions	Horizontal resolution Lon × Lat	Reference
1	CNRM-CM6-1	Centre National de Recherches Météorologiques (CNRM); Centre Européen de Recherches et de Formation Avancée en Calcul Scientifique, France	1.41° × 1.41°	Voldoire et al. [21]
2	CNRM-ESM2-1	Centre National de Recherches Météorologiques, Toulouse, France	1.41° × 1.41°	Séférian et al. [22]
3	IPSL-CM6A-LR	Institut Pierre Simon Laplace, Paris, France	2.5° × 1.26°	Boucher et al. [23]
4	MRI-ESM2-0	Meteorological Research Institute (MRI), Japan	1.13° × 1.13°	Yukimoto et al. [24]

Table 1.

List of CMIP6 models used in this study, modeling centers, horizontal resolution and references.

2.2 Bias correction

A bias correction procedure with the CDF-t method was applied on each grid point with the GCM data via the historical Global Meteorological Forcing Dataset (GMFD; [28]). For each daily parameter, the algorithm generates the cumulative distribution function (CDF) for the GMFD data and for the retrospective GCM simulations. The corresponding source values (day of year +/− 15 days) are clustered and sorted over the period from 1960 to 2014 at various probability thresholds to produce a quantile map between models and observations. Based on this map, the model values in the different quantiles (e.g., p = 90%) can be translated into corresponding GMFD values for the same quantile. Assuming that the CDF of the simulations is stable during the retrospective and prospective periods, the algorithm simply searches for the probability quantile associated with the predicted climate variations from the estimated CDF of the historical data and then accepts these as the adjusted climate projections. The climate projections adjusted in this way have the same CDF as the GMFD data; therefore, possible biases in the statistical structure (variance, in particular) of the original GCM outputs are eliminated by this procedure. At the end of the bias correction step, the previously extracted parameter climate trends (precipitation and temperature) are added to the fitted model climate fields.

2.3 Agroclimatic indicators

Table 2 provides details on the 14 agroclimatic indicators calculated in this study. These include seasonal rainfall characteristics (cumulative seasonal rainfall, number of rainy days, rainfall intensity, extreme rainfall, dry sequences, start, end and length of the rainy season). A dry sequence is obtained by counting the number of consecutive days without rainfall that lie between two rainy days (a day is rainy when its cumulative rainfall is greater than or equal to 1 mm). These are intra-seasonal droughts, also called rainfall breaks. We have defined four types: DSs, DSm, DSl and DSxl (see Table 2). The dry sequences that determine the potential risks of reseeding are the longest rainfall breaks observed over the 30 days after sowing [29]. Here, DSl and DSxl present themselves as effective indicators for detecting false starts to the season [18]. Next, in order to assess the ability of global models to adequately represent the agronomic start of the rainy season, we used Sivakumar’s [30] definition of local rainy season start. Based on an agro-nomic criterion, he considers the start of the rainy season in the Sahelian and Sudanian regions as the date, starting on May 1, when there is a rainfall of at least 20 mm on three consecutive days, with no dry spells of more than seven days in the following 30 days. The end of the agronomic season is defined by considering the date after September 1 when a long dry period of 20 days occurs. The duration of the agronomic season is defined for a given season as the difference between the start and end date of the rainy season according to the agronomic criterion. Finally, indicators illustrating the intensity (R95 and R99) and frequency (R10 and R20) of rainfall events are used in addition to the cumulative season (PRCTOT).

Indicator	Details	Definitions	Units
PRCPTOT	Total annual daily precipitation	Total annual rainy days >1 mm	mm
SDII	Daily intensity index	Ratio of annual total to number of rainy days	mm/day
R95 (HIP)	Rather rainy days	Total daily rainfall >95th percentile	mm
R99 (HIP)	Extreme daily rainfall	Total daily rainfall >99th percentile	mm
R10 (FIP)	Number of rainy days	Number of days with rain >10 mm	days
R20 (FIP)	Number of very rainy days	Number of days with rain >20 mm	days
WD	Rainy days	Number of days with rain >1 mm	days
DSs	Dry spells of short duration	Dry spell between 1 and 3 days	days
DSm	Dry spells of medium duration	Dry spell between 4 and 7 days	days
DSl	Dry spells of long duration	Dry spell between 8 and 15 days	days
DSxl	Dry spells of extreme duration	Dry spell >15	days
Onset	Start date of the agronomic season	Sivakumar Definition	days
Cessation	End date of the agronomic season	Sivakumar Definition	days
Length	Length of the agronomic season	Sivakumar Definition	days

Table 2.

Definitions of the agroclimatic indicators used.

3. Results

3.1 Changes in agroclimatic indicators in the near (2020–2049) and far future (2070–2099)

To study the impacts of climate change on the evolution of these indicators on each grid point of Senegal, the changes between the historical period (1985–2014) and the climate simulations SSP126, SSP245 and SSP585 for the near future (2020–2049) and the far future (2070–2099) are represented in Figure 1. The middle of the whisker diagram represents the spatial average over Senegal, and it is important to note that here the multi-model average is considered. An amplification of the changes in the evolution of the agroclimatic indicators is observed when moving from the near future to the distant future and from the sustainability scenario SSP126 to the fossil fuel development scenario, SSP585. In the near future, seasonal precipitation totals will be broadly the same or even slightly lower than what is observed in the historical model simulations. This slight decrease is more a consequence of a high occurrence of dry sequences and a shortening of the rainy season illustrated by a late start to the season. This evolution will become more pronounced in the distant future with a decrease in average over the country of −10% for SSP126, −20% for SSP245 and over −35% for SSP585.

Figure 1.

Projected changes in the evolution of the 14 agroclimatic indicators. Changes are defined as the difference between the historical period (1985–2014) and the climate simulations SSP126 (top panel), SSP245 (bottom panel left) and SSP585 (bottom panel right) for the near future (2020–2049; boxplot in black) and the distant future (2070–2099; boxplot in red).

By analyzing the distribution of precipitation over the season, it is clear that the trend observed over the period 2020–2049 is confirmed and amplified in the distant future. Indeed, a strong increase in dry sequences (DSs and DSxl) will be observed, which will increase from +15% for SSP 126 to +50% for SSP585. This situation will favor the risks of reseeding and post-flowering water stress. Although at the beginning of the rainy season (June to July), the occurrence of rainfall breaks of 8 to 14 days is a fairly frequent event, and the observation of a DSl or DSxl after the semi flowering is not something that farmers want. Indeed, even if young millet plants show a certain capacity to adapt to such pockets of drought, the same cannot be said for maize, sorghum or certain legumes (groundnuts, cowpeas, etc.). The water requirement of crops during the heading-maturity phase (or reproductive phase) is one of the main factors conditioning their final yield. In general, this period corresponds to the critical reproductive phase for non-photoperiodic crops with a cycle length of about 90 days. A late start of the agronomic season, from 10 days for SSP126 to about 20 days for SSSP585, will also be observed, whereas the end of the season seems to be less sensitive to the effects of climate change. Paradoxically, extreme rainfall events, notably R99 (mm) and R20 (days), will increase by about 10–15% and 20–25%, respectively, by 2050 before decreasing slightly in the distant future. These results are perfectly consistent with studies [31, 32, 33] demonstrating the installation of a new rainfall regime over the Sahel characterized by false starts, a shortening of the rainy season, an increased frequency of intense rainfall and an occurrence of dry sequences, which is directly associated with global warming.

3.2 Spatial distribution of the change in agroclimatic indicators

In order to better analyze the spatial distribution of these indicators, the average change of these indicators on each grid point is represented in Figure 2. This spatial distribution of changes on these agroclimatic indicators allows to measure the spatial heterogeneity of the impact of global warming.

Figure 2.

Spatial distribution of changes in the 14 agroclimatic indicators in addition to the mean surface temperature (Tas). The changes are defined as the difference between the historical period (1985–2014) and the SSP126 climate simulations for the near future (2020–2049).

Indeed, despite a general decline in cumulative seasonal rainfall on all pixels of the country, some areas of the country will be more impacted than others. The center of the country and even the Central East will suffer the largest decreases in terms of cumulative seasonal exceeding −20% followed by the south with rates between −15 and − 10%. The north and the coastal area of the country will record the smallest decreases (less than −10%). Precipitation intensity indicators will increase in the coastal areas, the southeast and the north except for R99 which will increase only in the south of the country. The 8-to-15-day dry spells (DSl) will increase more in the south of the groundnut basin, while DSxl will increase almost throughout the country. Finally, the delays in the start of the season observed in Figure 1 affect the groundnut basin (Fatick, Kaolack and Kaffrine), the center-east and the southeast of the country more, so the seasons are shorter in these areas of the country. This characterization of agroclimatic indicators allowed us to set up this ranking matrix of four global climate models (GCMs), namely CNRM-CM6-1, CNRM-ESM2-1, IPSL-CM6A-LR and MRI-ESM2-0 (see Figure 3).

Figure 3.

Ranking of the four models used according to the 14 agroclimatic indicators. The sign (+) of the arrow indicates an increase in the near future of the indicator and the sign (−) a decrease. For example, M2, which is the model where warming is stronger, tends to capture later season starts.

This diagram combining temperature changes, and precipitation characteristics relative to the 1985–2014 history allow us to determine, in terms of trend, whether the models are warm/dry/extreme/early-onset, warm/moist/extreme/early-onset, cool/moist/moderate/early-onset, cool/dry/moderate/early-onset for the near-future SSP126 (2020–2049). For example, temperatures are expected to increase in this period by +0.98°C with IPSL-CM6A-LR (M3), +1.07°C with CNRM-CM6-1 (M1), +1.09°C with MRI-ESM2-0 (M4) and + 1.1°C with CNRM-ESM2-1 (M2). With regard to precipitation, the changes are variable: With 3/4 models predicting a decrease in precipitation on average over Senegal ranging from −1 to −13%, the largest decrease is recorded by the warmest model, namely CNRM-ESM2-1 (M2). While the second warmest model MRI-ESM2-0 (M4) predicts an increase in precipitation explained mainly by a high occurrence of events such as R95, WD, R10 and longer seasons compared to the historical period. Finally, the IPSL-CM6A-LR, which is the model with less warming compared to the other three models, predicts a climate with an increase in extreme rainfall events such as R99, SDII, R20 and all categories of rainfall breaks (DSs, DSm, DSl and DSxl). This approach could allow for sensitivity testing with agronomic impact models. In addition, it will allow us to better understand the response of crops to different degrees of warming and extreme climate in the different agroclimatic indicators defined (Table 2).

4. Discussions and conclusions

This assessment of the evolution of 14 agroclimatic indicators shows that the effects of climate change that will affect all of Africa are already visible at the local level. Knowledge about future climate trends and their impacts has increased in recent years in Africa. According to the IPCC, monsoon rainfall is expected to increase over the central Sahel and decrease over the extreme western Sahel (Senegal area). The monsoon season is expected to have a delayed start. However, at a more localized scale, the question of the reliability of this information arises. The ability of global models to represent the Sahelian climate faces difficulties related to the limited availability of quality climate data and the many uncertainties intrinsic to the physics of the models. In order to better understand the impact of climate change on key indicators for the agricultural sector, four corrected global models are used to characterize their evolution in the near (2020–2049) and distant (2070–2099) future.

The results corroborate with the conclusions of the latest IPCC report that predict a decrease in rainfall in the western Sahel, agricultural droughts and extreme rainfall events. Indeed, we show that climate change will lead to an almost constant decrease for the different socioeconomic scenarios, of the seasonal cumulative rainfall in the near future below −10% for the ssp126 scenario and about −40% in the distant future for the ssp585 scenario. This decrease in rainfall is rather explained by an increase in rainfall breaks combined with a shortening of the rainy season as illustrated by the results. The frequency of the different categories of dry sequences will tend to increase to reach +20% for the DSl. This trend will persist in the distant future and could exceed +50% with the ssp585 scenario. The shortening of the rainy season will be mainly explained by a delay in the start of the rainy season which could be 1 month compared to the average observed over the 1985–2014 history. Furthermore, despite a decrease in precipitation, extreme rainfall events will increase in intensity (R99) and frequency (R20) in the near future for all scenarios before decreasing in the distant future. This increase in the frequency of heavy rainfall corroborates the results of Taylor et al. [34], Panthou et al. [35] and Chagnaud et al., [36] who attribute this intensification of rainfall to global warming, which particularly affects temperatures in the Sahara. Thus, a warmer Sahara intensifies convection in Sahelian storms through increased wind shear and changes in the Saharan air layer [37]. We also analyzed the strong spatial disparities behind these future trends. The results indicate that the central part of the country and the Central East will experience the largest decreases in terms of seasonal accumulation. Precipitation intensity indicators will increase most strongly in the coastal areas, the southeast and the north. The increase in dry sequences of the DSl and DSxl type and the delays in the start of the season in the southern part of the groundnut basin constitute a source of vulnerability for the agricultural sector, with significant consequences for food security, diseases, farm capital, loss of livestock, etc. [38]. In conclusion, our work could be the starting point for simulations of yields, biomass and several relevant indicators for the agricultural sector in order to quantify the impact of climate change on agriculture in Senegal. These studies could be carried out on various spatial scales (field, commune, department or region). This will allow for the exploration of adaptation strategies such as scaling up investments in irrigation, fertilization or pesticides and accelerating the adoption of climate-proof agricultural technologies such as drought-resistant crop varieties and collaboration between farmers and breeders for the mass use of organic fertilization.

Acknowledgments

We would like to thank the Projet d’Appui au Plan National d’Adaptation du Sénégal (PNA-FEM) for funding this research program to assess the impact of climate change on the Senegalese agricultural sector. The PNA-FEM is funded by UNDP and the Global Environment Facility to strengthen the capacity of sectoral ministries and local governments to better assess the impacts of climate change and to adapt decision-making to the risks posed by climate change.