Exploring Agronomic Management Strategies to Improve Millet, Sorghum, Peanuts and Rice in Senegal Using the DSSAT Models

Abstract

Achieving food security for a growing population under a changing climate is a key concern in Senegal, where agriculture employs 77% of the workforce with a majority of small farmers who rely on the production of crops for their subsistence and for income generation. Moreover, due to the underproductive soils and variable rainfall, Senegal depends on imports to fulfil 70% of its food requirements. In this research, we considered four crops that are crucial for Senegalese agriculture: millet, sorghum, peanuts and rice. We used crop simulation models to explore existing yield gaps and optimal agronomic practices. Improving the N fertilizer management in sorghum and millet resulted in 40–100% increases in grain yields. Improved N symbiotic fixation in peanuts resulted in yield increases of 20–100% with highest impact in wetter locations. Optimizing irrigation management and N fertilizer use resulted in 20–40% gains. The best N fertilizer strategy for sorghum and millet included applying low rates at sowing and in early development stages and adjusting a third application, considering the expected rainfall. Peanut yields of the variety 73-33 were higher than Fleur-11 in all locations, and irrigation showed no clear economic advantage. The best N fertilizer management for rainfed rice included applying 30 kg N/ha at sowing, 25 days after sowing (DAS) and 45 DAS. The best combination of sowing dates for a possible double rice crop depended on irrigation costs, with a first crop planted in January or March and a second crop planted in July. Our work confirmed results obtained in field research experiments and identified management practices for increasing productivity and reducing yield variability. Those crop management practices can be implemented in pilot experiments to further validate the results and to disseminate best management practices for farmers in Senegal.

1. Introduction

In the face of a rapidly expanding global population and the escalating challenges posed by climate change, achieving food security has become one of the foremost concerns of the 21st century. Senegal, located in the semi-arid region of West Africa, faces unique agricultural challenges that are intensified by climate variability and limited resources [1]. In 2020, agriculture contributed 15% to Senegal’s GDP and employed 77% of the workforce, with the majority of workers being smallholder farmers dependent on crop production for both subsistence and income generation [2]. However, the country’s agriculture is predominantly rainfed and highly vulnerable to climate fluctuations, including droughts, floods, rising temperatures, waterlogging and shorter growing seasons [3,4]. Despite the high proportion of the population engaged in farming, Senegal relies on imports to meet 70% of its food needs, largely due to the country’s underproductive soils and unpredictable rainfall patterns [5]. Numerous studies highlight the significant gap between current crop yields and the potential or attainable yields in Senegal [6,7]. The primary factors contributing to this yield gap include the inefficient or insufficient use of agricultural inputs, such as fertilizers, improved seed varieties and pesticides, as well as soil degradation caused by poor management practices and erosion, which are widespread in many agricultural regions.

Millet and sorghum are among the key staple crops in Senegal, playing a crucial role in both food security and the livelihoods of the population. These crops have long been an integral part of the country’s agricultural heritage [2]. In 2019, millet and sorghum were cultivated on a combined area of 1 million hectares, with yields of 600 kg/ha for millet and 800 kg/ha for sorghum [8], making them the most widely grown crops in Senegal. Known for their resilience to semi-arid conditions [9], as well as their nutritional benefits—rich in iron and zinc and free from gluten [10,11,12]—millet and sorghum are essential to both food security and sustainable agricultural development. Their consumption plays a pivotal role in combating malnutrition, particularly in regions where access to diverse and nutritious diets is limited.

Peanuts are a vital cash crop in Senegal, playing a significant role in the nation’s economy, food security and rural livelihoods. It provides a stable source of income for farmers, contributing to poverty reduction and broader socio-economic development [13,14,15]. The export of peanuts and their derivatives generates substantial revenue, bolstering foreign exchange earnings and supporting overall economic stability. Peanut cultivation is especially important for smallholder farmers, offering a means of income diversification and serving as a key driver of rural development [16]. In addition to its economic impact, peanuts enhance food security by offering a nutrient-rich food source that complements other dietary staples. Containing over 25% protein, along with essential nutrients such as calcium, folate and immune-boosting phytosterols [17], peanuts are a valuable component of the Senegalese diet [18]. Furthermore, peanut byproducts such as oil and meal play an important role in producing cooking oil and animal feed, thereby contributing to both food security initiatives and the sustainability of the livestock sector.

Finally, rice is a dietary staple in Senegal, providing 22–40% of the population’s caloric intake and 23–39% of its protein needs [19]. Rich in carbohydrates, vitamins and minerals, rice contributes to a balanced and diverse diet [19,20]. Its cultivation is also a cornerstone of Senegal’s economy, supporting the livelihoods of many smallholder farmers and serving as an important source of income diversification [21,22]. Despite its importance, domestic rice production—ranging between 1.3 and 1.4 million tons from 2020 to 2022—meets only about 28% of national demand [23]. In response, the government has implemented policies aimed at achieving rice self-sufficiency. Early efforts concentrated on harnessing the 240,000 hectares of potential farmland in the Senegal River Valley through irrigation development. While this strategy remains relevant, the release of improved rainfed rice varieties—such as NERICA by Africa Rice and the Sahel and ISRIZ varieties by the Institut Sénégalais de Recherches Agricoles (ISRA)—has shifted the focus. Currently, around 50% of national rice production originates from lowland and plateau areas in Casamance and the Groundnut Basin [24,25]. Significant efforts continue to promote the adoption of modern agricultural practices, improved seed varieties, and sustainable farming techniques to boost productivity and resilience [26,27].

Agriculture in Senegal is increasingly challenged by climate variability, limited resources, and a growing population. In this context, crop simulation models have become valuable tools for informed agricultural decision-making. These models offer farmers and policymakers insights into crop growth, yield projections, and resource management under diverse environmental and economic scenarios [4,28,29,30,31]. By simulating multiple scenarios, farmers can anticipate potential yield fluctuations and implement adaptive crop management strategies to reduce losses or capitalize on favorable conditions. This proactive approach strengthens the resilience of agricultural systems, supporting both food security and economic stability. Moreover, agricultural policies significantly shape the sector, and crop simulation models can aid policymakers by providing comprehensive assessments of the potential impacts of their decisions on crop production, resource use and environmental sustainability.

Crop simulation models can play a vital role in supporting both farmer decision-making and policy development by offering valuable insights into crop production strategies. However, a commonly cited concern among farmers and policy advisers is that while they recognize the potential of these models as decision-support tools, they often find them difficult to use due to complex interfaces. To address this challenge, Han et al. [31] developed SIMAGRI, a user-friendly interface designed with two key goals: to make crop models more accessible to farmers and policy advisers and to ensure simulations are based on appropriate inputs—such as soil characteristics, cultivar types, weather data and economic parameters. SIMAGRI enables users to run DSSAT simulations for a variety of “what-if” scenarios, incorporating different climate conditions (historical or forecasted) and crop management practices. The interface supports simulations using various combinations of cultivars, soil types, sowing dates, fertilizer applications, irrigation regimes and more. A graphical user interface (GUI) allows users to set up scenarios, execute simulations and view results in intuitive probabilistic formats, such as box plots or cumulative probability curves, facilitating direct comparisons between management strategies. Additionally, SIMAGRI integrates crop modeling with economic data—including prices and production costs—and presents outcomes in straightforward financial terms, such as gross margin, making the results more actionable for decision-makers.

In this study, we conduct a series of crop simulation runs for millet, sorghum, peanut and rice to explore combinations of management options across different regions of Senegal. The aim is to identify strategies that can inform on-farm decision-making and support the development of policies aimed at increasing productivity and reducing yield variability. This work is grounded in the Theory of Change framework outlined in Appendix A Figure A1, which guides our approach to linking simulation-based insights with actionable outcomes for both farmers and policymakers.

Specifically, we use the DSSAT crop models [32] to answer the following research questions:

1.1. General Questions for the Four Crops

• What are the gaps between current crop yield and attainable yields with crop management practices that maximize yields (planting dates, fertilizer, irrigation)?
• How does the yield interannual variability change with improved practices?
• What is the role of seasonal climate forecasts in adjusting crop management practices?
• What is the profitability of investing in improved varieties, fertilizers and irrigation?

1.2. Millet and Sorghum

• What is the impact of adjusting planting dates based on definitions of the onset of rainy season?
• What is the optimal strategy for N fertilizer management?

1.3. Peanuts

• What is the expected yield with varieties of different season lengths (long vs. short)?
• What is the feasibility of peanut production in different regions?
• What is the feasibility of irrigation for peanut production?

1.4. Rice

• What is the suitability of rainfed rice in selected regions of Senegal?
• What is the strategy for optimal N fertilizer management for rainfed rice?
• What are the crop irrigation needs in different regions, using different production strategies (to inform investments in irrigation infrastructure)?
• What is the feasibility of irrigated rice in single and double cropping systems in the Senegal River Valley?

2. Materials and Methods

2.1. Crop Simulation Models

We used the DSSAT version 4.7 crop models [32,33,34] that have been tested throughout the world and found to be robust tools for quantifying the impact of crop management practices on crop production [9,35]. Model calibration and performance evaluation were carried out using crop phenology, management and grain yields data obtained from research work in Senegal on millet, sorghum, peanut and rice. (The articles describing the calibration and testing of the DSSAT models that were used in this paper are presented in Appendix A 

Table A1. Description, application and source of collected data used in the calibration of the DSSAT models.

Data	Description	Period	Application	Source
Sorghum field experiment data	Determine the effect of fertilizer application on sorghum grain yield and formulation of tailored fertilization strategies according to sorghum varieties	2015–2016	Calibration/Evaluation	[38,39]
Peanuts management and yield data	Effects of fertilization rate and water availability on peanut growth and yield	2014–2015	Calibration/Evaluation	[29]
Soil data	Analysis of soil physical and chemical properties at the experimental sites in 2014	2014–2015	Calibration/Evaluation	[29]
Peanuts management and yield data	Evaluate the impact of climate change on peanut yield in Senegal	2014–2015	Calibration/Evaluation	[40]
Rice management and yield data	Measure the agronomic traits of four upland rainfed rice NERICA (NERICA 1 and NERICA 4 (95–100 days); NERICA 8 and NERICA 11 (75–85 days) in a sudano-sahelian zone	2013–2014	Calibration/Evaluation	[6]

, and the genetic coefficients used for each DSSAT crop model are shown in Appendix A 

Table A2. Genetic coefficients used in the simulation runs of the four DSSAT crop models based on the work included in Table A2.

SORGHUM			MILLET
Coefficient		FaddaW	Coefficient	CIVT
P1		200.0	P1	100.0
P2		280.0	P2O	12.0
P2O		12.6	P2R	142.0
P2R		655.0	P5	390.0
PANTH		617.5	G1	1.0
P3		145.0	G4	0.6
P4		81.5	PHINT	43.0
P5		400.0	GT	1.0
PHINT		49.0	G5	11.0
G1		3.0
G2		6.0
PEANUTS	Cultivar		RICE
Coefficient	FLEUR 11	VAR 73-33	Coefficient	NERICA 1&4
CSDL	11.84	11.84	P1	380
PPSEN	0.0	0.0	P2R	100
EM-FL	16.0	20.0	P5	300
FL-SH	7.0	7.6	P2O	13
FL-SD	12.0	13.0	G1	75.0
SD-PM	36.0	40.0	G2	0.03
FL-LF	66.0	66.0	G3	1.0
LFMAX	1.7	1.55	G4	83.0
SLAVR	250	250	PHINT	37.0
SIZLF	25.0	15.0
XFRT	0.95	0.9
WTPSD	0.36	0.36
WTPSD	0.36	0.36
SFDUR	29.0	25.0
SDPDV	1.65	1.65
PODUR	8.0	10.0
THRSH	95.0	95.0
SDPRO	0.7	0.7
SDLIP	0.51	0.51

.) In all simulations, we started the model runs several months before the rainy season normally starts, and we used soil water content at a permanent wilting point as the initial condition for the runs. Unless specified, the sowing date was determined following Han et al.’s [36] criteria (sow when the soil total plant available water content—TAW—reaches 30% of its maximum value, TAW = 30). For each crop, we used cultivars that had been previously calibrated and tested in Senegal (Appendix A Table A1 and Table A2).

2.2. Sites

In this study, we selected ISRA’s four agricultural research experimental fields in Senegal, Bambey (14.71° N, 16.48° W), Nioro du Rip (13.76° N, 15.78° W), Sinthiou Maème (13.82° N, 13.9° W) and Kolda (12.88° N, 14.25° W), and four additional sites relevant for rice production (Appendix A Figure A3). Long-term weather data (i.e., daily maximum and minimum temperature, solar radiation and precipitation) from 1981 to 2018 were obtained from the ANACIM for defining meteorological onsets and for running the DSSAT millet, sorghum, peanut and rice models. The climate characteristics of the study sites are presented in Appendix A Figure A3.

In each study site, we used soils that are representative of those suitable for crop production and that represent the wide heterogeneity characteristic of Senegal: they are predominantly sandy in Bambey with a low water-holding capacity, while in Kolda, they are predominantly clayey with higher water holding capacity. Soils in Nioro and Sinthiou present an intermediate clayey-sandy texture. Due to agricultural practices and climatic conditions that cause wind and water erosion, land degradation has occurred in the different agroecological zones of Senegal. In addition, the removal of harvest residues typically for animal feed has accentuated the deterioration of soils due to net losses of organic carbon. Soils in the floodplains of the Senegal river valley are alluvial with high clay content and are locally described as Faux Hollaldé and Hollaldé.

2.3. Prices and Costs

Production costs and prices received by farmers used for the economic analyses in this article are presented in Appendix A (

Table A4. Results of calibrated DSSAT rice model comparing observed^* and simulated anthesis date, maturity date and grain yields.

		Nerica 1 *		Nerica 8 *
		Observed	Simulated	Observed	Simulated
Anthesis	Median	66.2	67.0	64.8	63.0
Date (DAP)	Q25	63.7	66.0	61.0	63.0
	Q75	70.0	68.3	66.6	65.0
Maturity	Median	92.0	91.0	86.0	88.0
date (DAP)	Q25	86.0	90.0	85.0	87.0
	Q75	95.0	94.0	94.0	90.0
Yield (kg/ha)	Median	1282	2180	1532	1896
	Q25	744	1629	840	1413
	Q75	2131	2626	2656	2210

* 306 and 116 field observations were used for Nerica 1 and Nerica 8, respectively.

). The information was obtained from different available sources and checked by ISRA. The base production cost for all crops include labor, soil preparation, pesticides, herbicides and P and K fertilizers.

3. Results and Discussion

3.1. Yield Gaps of Sorghum and Millet

The model runs under ideal conditions—i.e., affected only by solar radiation and temperature, with no water or nutrient limitations—resulted in mean grain yields of 12–15 Mg/ha for sorghum and 9–10 Mg/ha for millet. Given the short rainy season characteristic of much of Senegal, “potential yields” under such ideal conditions were considered too far off from realistically attainable yields in the field. Consequently, we defined “potential yield” as the attainable grain yield under observed rainfall conditions but without nutrient deficiencies. (In this paper, we define “potential,” “optimal,” “N-limited” yields in different ways for different crops.

Table A3. Grain prices and production costs for sorghum, millet, rice and peanuts in 2023 (CFA).

Sorghum
Base production cost		156,500 CFA/ha
Nitrogen fertilizer		1130 CFA/kg N
Sorghum price		160 CFA/kg grain
Sorghum byproduct		10 CFA/kg byproduct
Millet
Base production cost		154,000 CFA/ha
Nitrogen fertilizer		1130 CFA/kg N
Millet price		160 CFA/kg grain
Millet byproduct		5 CFA/kg byproduct
Peanuts
Base production cost		395,700 CFA/ha
Irrigation		630 CFA/mm
Peanut price		280 CFA/kg grain
Peanut byproduct		33 CFA/kg byproduct
Rice
Base production cost		391,000 CFA/ha
Irrigation		630 CFA/mm
Nitrogen fertilizer		1130 CFA/kg N
Rice price		180 CFA/kg Grain
Rice byproduct		0 CFA/kg byproduct

in the Appendix A summarizes these definitions for each crop.) In practical terms, this would require applying fertilizers whenever the crop shows signs of deficiency. While such intensive nutrient management is not feasible in the field, it provides a useful benchmark for the climate-based potential sorghum yield for evaluating the conceivable benefits of improved fertilizer practices. Among nutrients, phosphorus (P) and potassium (K) management tends to be more straightforward than nitrogen (N). Soil tests for P and K can be calibrated to specific soil types, and these nutrients typically have some residual effect, meaning they are not entirely lost if not taken up by the current crop. In contrast, N has little to no residual effect—any N not absorbed by the crop is often lost through leaching, runoff or denitrification. For this reason, our analysis focused on N management, assuming that P and K were non-limiting. We compared “potential yields” under no N limitation to simulated yields obtained using relatively high N fertilizer rates applied at three key crop growth stages. This approach allowed us to assess both the potential for yield improvement and the effectiveness of different N management strategies.

3.1.1. Sorghum Yield Gap

Optimal N fertilizer management in this study was defined as applying 20 kg N/ha at sowing, 20 kg N/ha 36 days after sowing and 60 kg N/ha at panicle initiation (51 days after sowing). Grain yields obtained with this N fertilizer strategy were defined as “optimal yields.” In Bambey—a location characterized by severe water limitations—the difference between “potential” and “optimal” yields was minimal, suggesting that nitrogen availability was not the primary constraint to achieving higher yields (Figure 1a). In contrast, results from the wetter locations—Kolda, Nioro and Sinthiou—indicate that improved N fertilizer management could significantly enhance grain yields. In these regions, “potential” sorghum yields were 40–80% higher than those achieved under the “optimal” nitrogen regime (Figure 1a), highlighting a substantial yield gap attributable to nitrogen limitations. Possible strategies for optimizing nitrogen management in these environments are discussed in a later section.

3.1.2. Millet Yield Gap

Optimal fertilizer management for millet was defined as applying 20 kg N/ha at sowing, 20 kg N/ha 20 days after sowing (near panicle initiation) and 30 kg N/ha 45 days after sowing (approximately 10 days before anthesis). Similar to the findings in sorghum, millet grain yield response to eliminating the N deficiencies in the “potential” yields were 80–100% higher than the “optimal” fertilizer strategies. Unlike sorghum, in the case of millet, even in the water limited location (Bambey), removing N deficiencies also resulted in important grain yield gains (Figure 1b). Millet is well adapted to low soil water content, and therefore, N response is evident even in environments with high water deficiencies such as Bambey.

3.2. Peanut Yield Gap

Following a similar criterion as for sorghum and millet, “potential” peanut yield was defined as the yield attainable with observed rainfall and temperature and with full symbiotic nitrogen (N) fixation capacity—i.e., when all of the nitrogen required for crop growth and development is supplied through symbiotic fixation. This ideal scenario was compared to the more frequent N-limited situation in which symbiotic fixation falls short and does not provide the total N demand over the growing season (in our case, we simulated the provision of 60% of the total N demand through symbiotic fixation). The results, presented in Figure 2, highlight the critical role of effective symbiotic N fixation: peanut yields were 20–100% higher under full symbiotic fixation conditions. As expected, the greatest yield response was observed in Kolda, the location with the highest rainfall, where favorable moisture conditions likely enhanced the efficiency of N fixation and overall crop performance.

3.3. Rice Yield Gap

To estimate yield gaps in rice, we analyzed three scenarios: (1) potential yield—with no water or N limitations; (2) no N limitation—with optimal irrigation and unlimited nitrogen; and (3) optimal yield—with optimal irrigation and N fertilizer applied at 30 kg N/ha at three growth stages: sowing, 25 days after sowing (DAS) and 45 DAS. “Optimal irrigation” was defined as applying 10 mm of water whenever soil moisture dropped below 50% of the total available water (TAW).

The results underscore the critical importance of optimizing both irrigation and nitrogen management. In most locations, improved water use alone led to an approximate 20% increase in grain yield, while optimized nitrogen application contributed an additional 20% yield gain (Figure 3). The only exception was Kolda, the rainiest site, where enhancing water management beyond the tested strategy had little to no effect, likely due to sufficient natural rainfall.

3.4. Optimal N Fertilizer Recommendations for Sorghum and Millet

3.4.1. Sorghum

We evaluated the yield response of sorghum and millet using optimal planting dates. Our definition of the “optimal planting date” was guided by the findings of Han et al. [36], who identified the onset of the rainy season as the date when soil water content reaches 30% of the total plant-available water (TAW) in the soil. Based on this criterion, we used the DSSAT model’s automated planting function to initiate sowing when the soil moisture reached this threshold. To validate or refine this approach, we also tested an alternative definition of optimal planting—using a higher soil moisture threshold of 95% TAW (i.e., almost field capacity), which delayed planting by 10 to 20 days. The results confirmed that the 30% TAW threshold remains the most effective criterion for defining optimal planting dates, aligning with Han et al.’s [36] definition of the onset of rains (Appendix A Table A4). Subsequently, the response to nitrogen fertilizer was evaluated based on this confirmed definition of the optimal planting date.

We studied the sorghum response to N fertilizer applied at three growth stages: at sowing, 36 days after sowing (DAS) and at panicle initiation (51 DAS). The design of the simulation experiments allowed us to study the sorghum response to N applied at each one of those three growth stages and to identify optimal combinations of application rates and timing.

Sorghum responded positively to nitrogen (N) fertilizer applications at sowing and at 36 days after sowing (DAS), but the strongest yield response was observed when fertilizer was applied near panicle initiation (51 DAS). Given the frequency of intense rainfall events during the growing season—which can lead to significant N losses through leaching and runoff—splitting fertilizer applications is a reasonable strategy to reduce these losses. The first two applications (at sowing and 36 DAS) support early crop establishment and vegetative growth, a phase during which the crop’s N demand is relatively low but critical for proper development. The third application, just before panicle initiation, coincides with the crop’s peak growth period and highest N demand, explaining the particularly strong yield response at this stage. These results highlight the potential for a growth stage–specific N recommendation system, such as the one proposed by [37] for barley, which combines plant and soil indicators. Analysis of 35 years of data across four locations suggests that the most effective fertilizer strategy involves applying 20 kg N/ha at sowing and 20 kg N/ha around 30 DAS and then adjusting the third application—just before panicle initiation—based on location-specific factors and expected seasonal rainfall (

Table 1. N Fertilizer response of sorghum (kg grain/ha) in four locations of Senegal (Bambey, Kolda, Nioro and Sinthiou). N fertilizer (kg N/ha) applied at sowing, 36 days after sowing (DAS) and 51 DAS.

	N Fertilizer (kg N/ha) Applied at Sowing-36 DAS-51 DAS
	0-0-0	20-0-0	20-20-0	20-20-30	20-20-60	20-40-0	20-40-30	20-40-60
Bambey
Median	124	674	928	1123	1241	1034	1207	1386
Q25	107	634	839	1039	1128	943	1131	1217
Q75	136	705	1028	1339	1458	1132	1489	1605
Kolda
Median	2738	3415	4134	5032	6050	4875	5713	6816
Q25	2379	2984	3607	4526	5465	4302	5285	6037
Q75	2989	3707	4421	5309	6510	5119	6187	7471
Nioro
Median	2463	2937	3712	4674	5371	4352	5185	5939
Q25	2261	2830	3395	4223	5091	3984	4794	5468
Q75	2723	3325	4012	4865	5711	4685	5498	6432
Synthiou
Median	2677	3223	3829	4749	5572	4541	5279	6288
Q25	2437	2899	3498	4319	5182	4069	4951	5563
Q75	3062	3631	4338	5200	5974	4984	5779	6663

). We also found a significant correlation between the onset of the rainy season and total seasonal rainfall: for every 10-day delay in onset, total rainfall decreases by approximately 65–80 mm. This relationship, along with seasonal climate forecasts—such as those provided by the IRI (International Research Institute for Climate and Society, Columbia University)—can inform adaptive nitrogen management. Table 3a summarizes sorghum fertilizer recommendations tailored to expected rainfall conditions.

We then conducted an economic analysis of different nitrogen (N) fertilization strategies using the 2024 market prices for sorghum (as received by farmers), the base cost of sorghum production (excluding N fertilizer) and the prevailing price of N fertilizer. Gross margins were calculated for each fertilization strategy to assess economic viability. As shown in Figure 4, omitting N application at panicle initiation (51 DAS) frequently resulted in negative economic outcomes. In contrast, the application of N at panicle initiation consistently yielded the highest economic returns, underscoring the importance of this growth stage for optimizing profitability. Furthermore, increasing N rates beyond 20 kg/ha at sowing or 36 DAS—whether through single applications (e.g., 20-40-0 vs. 20-20-0) or split applications (e.g., 20-40-60 vs. 20-20-60)—did not lead to improved economic performance. These findings suggest that the most efficient fertilization strategy balances agronomic response with cost-effectiveness, with particular emphasis on the timely application of N at panicle initiation.

3.4.2. Millet

We also analyzed millet’s response to nitrogen (N) fertilization using optimal sowing dates tailored to each location. As with sorghum, we compared the optimal planting threshold defined by Han et al. [36]—where sowing occurs when plant-available water in the soil (TAW) reaches 30%—with a more conservative approach that delays sowing until TAW reaches 95%, following additional rainfall events. The results confirmed that the 30% TAW threshold resulted in superior performance: higher and more stable grain yields, as well as planting dates that were 10–20 days earlier (

Table 2. N fertilizer response of millet (kg grain/ha) in three locations of Senegal (Bambey, Nioro and Sinthiou). N fertilizer (kg N/ha) applied at sowing, 20 days after sowing (DAS) and 41 DAS.

	N Fertilizer (kg N/ha) Applied at Sowing, 20 DAS and 41 DAS
	0-0-0	20-0-0	20-20-0	20-20-30	20-20-60	20-40-0	20-40-30	20-40-60
Bambey
Median	91	358	546	1196	1581	599	1256	1694
Q25	76	241	360	866	1132	372	931	1172
Q75	115	423	657	1625	1980	776	1680	2229
Nioro
Median	217	534	614	1631	2377	670	1655	2411
Q25	185	449	517	1328	1687	578	1365	1735
Q75	231	587	729	1777	2530	808	1854	2597
Synthiou
Median	260	587	683	1691	2287	728	1727	2338
Q25	234	539	637	1535	1881	679	1563	1942
Q75	281	615	780	1821	2605	853	1881	2695

).

Millet’s response to N fertilizer rates and application timing closely mirrored that of sorghum. Moderate N applications—20 kg N/ha at sowing and 20 DAS (near panicle initiation)—were essential for proper crop establishment and early growth. However, the most pronounced yield response was observed with an application at 45 DAS, just before anthesis, indicating that this growth stage is critical for optimizing N uptake and grain production (Figure 5).

In addition, we found a significant correlation between millet yields and total rainfall from 41 DAS to crop maturity (Appendix A, Figure A3. This finding, along with the observed relationship between rainfall onset and growing season rainfall, informed the development of fertilizer recommendations. Table 3 presents a summary of optimal N fertilizer strategies for millet derived from 35 years of simulations.

Table 3. Summary of recommendation of N fertilizer application for (a) sorghum and (b) millet, based on onset date and rainfall forecast after panicle initiation (DAS = days after sowing).

(a) ONSET	Sorghum Forecast * Post 51 DAS	N Recommendation ** (0 + 36 DAS + 51 DAS)
Early	N or A	20 + 20 + 60
Early	B	20 + 20 + 30
Late	N or A	20 + 20 + 30
Late	B	20 + 20 + 0
(b) ONSET	Millet Forecast * Post 41 DAS	N Recommendation ** (0 + 20 DAS + 41 DAS)
Early	N or A	20 + 20 + 30
Early	B	20 + 20 + 0
Late	N or A	20 + 20 + 30
Late	B	20 + 20 + 0

* N = normal, B = below normal, A = above normal precipitation; ** N recommendations in kg N/ha.

Table 3. Summary of recommendation of N fertilizer application for (a) sorghum and (b) millet, based on onset date and rainfall forecast after panicle initiation (DAS = days after sowing).

(a) ONSET	Sorghum Forecast * Post 51 DAS	N Recommendation ** (0 + 36 DAS + 51 DAS)
Early	N or A	20 + 20 + 60
Early	B	20 + 20 + 30
Late	N or A	20 + 20 + 30
Late	B	20 + 20 + 0
(b) ONSET	Millet Forecast * Post 41 DAS	N Recommendation ** (0 + 20 DAS + 41 DAS)
Early	N or A	20 + 20 + 30
Early	B	20 + 20 + 0
Late	N or A	20 + 20 + 30
Late	B	20 + 20 + 0

* N = normal, B = below normal, A = above normal precipitation; ** N recommendations in kg N/ha.

3.5. Peanuts: Planting Dates, Varieties, and Feasibility of Irrigation

3.5.1. Planting Dates

We evaluated optimal sowing dates for peanuts by comparing two varieties, Fleur-11 and 73-33, the latter having a growing season approximately 15–20 days longer. Although yield differences between the two sowing date definitions were not substantial, the approach using a soil water threshold of 30% TAW (total available water) generally outperformed the 95% TAW threshold. Consequently, all subsequent analyses were conducted using sowing dates based on the 30% TAW threshold.

We further analyzed yield performance and gross margins for both varieties under optimal sowing dates. Across all four locations, 73-33 consistently produced higher yields than Fleur-11. The smallest yield advantage for 73-33 was observed in Bambey, the site characterized by a late onset of rains and the lowest total seasonal rainfall. This suggests that a shorter-season variety like Fleur-11 may still be better suited for more water-limited environments (

Table 4. Yields (kg/ha) and gross margins (CFA/ha) of rainfed peanuts for two cultivars (73-33 and Fleur 11) in four locations of Senegal.

Cultivar	73-33				Fleur 11
Site	Bambey	Kolda	Nioro	Sinthiou	Bambey	Kolda	Nioro	Sinthiou
Yield (kg/ha)
Median	1099	2500	2012	1870	2342	4114	2612	2819
Q25	744	2266	806	1573	1786	3637	1490	2244
Q75	1626	2598	2254	2239	3090	4399	3272	3322
Gross Margin (CFA/ha)
Median	23,590	512,628	324,879	308,530	−49,876	302,828	−25,829	−25,829
Q25	−104,670	455,641	−75,657	189,819	−175,313	252,345	−205,295	−205,295
Q75	203,010	564,101	411,182	418,506	35,616	366,724	129,320	129,320

).

A basic economic analysis was also performed using 2023 market prices for peanuts and input costs. Results indicated that Fleur-11 frequently resulted in negative gross margins in three of the four locations. While 73-33 outperformed Fleur-11 across all sites in terms of both yield and gross income, its performance in Nioro was marked by high variability, with a notable risk of negative returns as shown in Table 4.

3.5.2. Irrigation

We assessed the feasibility of irrigating peanut crops across the four study locations using the following irrigation strategy: apply water (up to 10 mm per event) whenever soil moisture dropped below 50% of the total available water (TAW) and continue until field capacity is restored. While grain yield consistently responded positively to irrigation across all sites, the economic analysis revealed that irrigation does not present a clear advantage in any of the locations considered (Figure 6). In the driest sites—Bambey and Nioro—the relatively modest yield increases did not offset the high costs associated with water access and irrigation infrastructure, making such investments difficult to justify. However, irrigation in these areas did significantly reduce the probability of negative gross margins, indicating a potential role in risk mitigation rather than yield maximization. Conversely, in the wetter sites—Kolda and Sinthiou—the yield gains from irrigation were minimal, and variability in economic returns remained largely unchanged with or without irrigation.

3.6. Rice

The DSSAT rice simulation model was previously calibrated by [38] for two groups of cultivars: Nerica 1 and Nerica 4, characterized by a growing period of 95–100 days, and Nerica 8 and Nerica 11, with a shorter cycle of 75–85 days. For validation purposes, observed field data from 2012 to 2014 were collected and compared with model outputs for anthesis and maturity dates, as well as grain yield ranges (Appendix A, Table A4. As anticipated, the DSSAT model demonstrated satisfactory performance in simulating both phenological development and yield. However, the observed variability in flowering and maturity dates, as well as yields, was consistently greater than that produced by the model simulations. Given the comparable yield levels between the two cultivar groups, the shorter-duration varieties (represented by Nerica 8 and Nerica 11) were selected for subsequent simulations due to their potential advantage under water-limited conditions

3.6.1. N Fertilizer Strategy for Rainfed Rice

The response of rainfed rice to nitrogen (N) fertilization was evaluated using simulations based on optimal sowing dates, defined as the point at which plant-available water reaches 30% of the soil’s total available water capacity (TAW = 30%). The study was conducted across two locations: Kolda and Tambacounda (Figure 7). Nitrogen was applied at three crop growth stages: at planting, 25 days after sowing (DAS; approximately 10 days prior to panicle initiation) and 45 DAS (approximately 20 days before anthesis). Results indicated a consistent positive yield response to N applications at sowing and at 25 DAS, up to a maximum of 30 kg N/ha per application. Additional N applied at 45 DAS also contributed to yield increases; however, in most cases, rates exceeding 30 kg N/ha did not lead to further yield improvements. Based on these findings, a general recommendation for N fertilization in rainfed rice under similar agroecological conditions would consist of applying a total of 90 kg N/ha, distributed evenly as 30 kg N/ha at sowing, 25 DAS and 45 DAS.

3.6.2. Irrigated Rice

We evaluated the optimal sowing date for rice across six locations in Senegal, each characterized by distinct climatic conditions. The onset of the rainy season was defined using the same threshold employed by [36], specifically when soil water content reaches 30% of total available water (TAW = 30%). Optimal sowing dates ranged from June 15 in Kolda to between July 15 and July 30 in St. Louis and Bambey (

Table 5. Optimal planting dates (day of year), irrigation needs (mm), grain yields (kg/ha) and gross margins (CFA/ha) of irrigated rice in six locations of Senegal.

	Bambey	Kolda	St Louis	Kaolak	Thies	Tambactou
Optimal Planting
Date (DOY)
Median	200	165	214	182	200	159
Q25	182	159	196	170	189	155
Q75	213	172	223	194	213	165
Irrigation
need (mm)
Median	165	21	205	75	144	57
Q25	101	18	175	39	101	22
Q75	212	40	224	114	165	81
Yield (kg/ha)
Median	7326	7000	7619	7516	7570	7479
Q25	6682	6448	6842	7033	6898	7040
Q75	7537	7421	7908	7870	7945	7799
Gross Margin
(CFA/ha)
Median	179,305	227,602	206,493	265,063	223,976	257,656
Q25	116,928	184,085	99,341	228,539	171,500	211,849
Q75	244,269	269,655	245,101	309,359	299,669	312,356

). Simulated grain yields under these optimal planting dates were relatively uniform across locations, ranging from 4000 to 5000 kg/ha.

Subsequently, we estimated the irrigation requirements necessary to attain these yields. Results revealed considerable variation in irrigation needs, with drier regions such as Bambey and St. Louis requiring between 150 and 250 mm/year, while wetter locations such as Kolda and Tambacounda required less than 100 mm/year. Intermediate values were observed in Kaolak and Thiès, where irrigation needs ranged from 75 to 150 mm/year (Table 5). Although economic analyses confirmed that rice yields responded positively to irrigation in all sites, achieving irrigation demands greater than 200 mm/year may be economically unfeasible without substantial infrastructure investment or ready access to water sources. In regions such as the Senegal River Valley (e.g., St. Louis), irrigation is more viable due to the proximity to reliable water supplies. In contrast, locations such as Bambey, which are remote from water sources, present significant logistical and economic challenges for irrigation.

As part of Senegal’s ongoing national efforts to increase rice self-sufficiency and reduce dependency on imports, one promising strategy is to intensify production in irrigated areas through double cropping. We explored the feasibility of growing two rice crops per year by simulating multiple combinations of planting dates. While practical factors such as labor availability and competition with other agricultural activities (e.g., horticulture or rice processing) are important, these were not explicitly included in the present analysis. Instead, the focus was on identifying planting date combinations that maximized gross margins. The simulated strategies included a first rice crop sown in either January, February or March and a second rice crop sown in June, July or August.

Our initial results indicated that rice yields and their variability remained relatively stable across all sowing dates under optimal irrigation (triggered at TAW = 50%). However, gross margins (expressed in CFA/ha) varied significantly depending on the combination of sowing dates, rice market prices and irrigation costs. We tested four scenarios combining two rice prices (150 and 200 CFA/kg) and two irrigation costs (650 and 1100 CFA/mm). Under low irrigation cost conditions, the most favorable economic outcome was obtained when the first crop was sown in January or February, followed by a second crop in July (

Table 6. Total gross margin for double rice crop in St. Louis using different scenarios of rice price (150 and 200 CFA/kg) and irrigation cost (650 and 1100 CFA/mm).

	Total Double Crop Gross Margin (CFA/ha) in St. Louis
	Rice = 150 CFA/kg and Irrigation Cost = 650 CFA/mm
	Jan–Jun	Jan–Jul	Jan–Aug	Feb–Jun	Feb–Jul	Feb–Aug	Mar–Jun	Mar–Jul	Mar–Aug
Median	56,260	141,340	44,254	57,240	117,555	19,345	253,480	309,052	210,906
Q25	34,218	105,081	−1746	8518	63,335	−37,548	125,691	194,535	103,742
Q75	125,195	198,576	88,894	99,855	175,053	90,693	352,979	386,198	299,541
	Rice = 150 CFA/kg and Irrigation Cost = 1100 CFA/mm
Median	−223,955	−123,335	−224,448	−224,810	−166,200	−266,710	140,161	227,396	118,574
Q25	−264,778	−144,205	−261,814	−293,090	−199,475	−311,905	−1906	93,992	9561
Q75	−155,943	−55,810	−171,844	−191,925	−75,885	−177,100	252,186	303,469	198,418
	Rice = 200 CFA/kg and Irrigation Cost = 650 CFA/mm
Median	529,195	623,990	504,500	518,225	592,345	470,165	491,644	538,595	423,831
Q25	501,292	574,263	449,533	456,480	514,440	390,165	363,167	446,576	307,694
Q75	613,694	701,624	563,585	577,633	671,165	546,780	590,479	630,042	518,652
	Rice = 200 CFA/kg and Irrigation Cost = 1100 CFA/mm
Median	242,202	369,182	240,524	232,520	314,855	175,330	365,481	450,821	304,328
Q25	200,843	319,198	179,989	162,388	251,060	111,328	232,107	326,906	219,217
Q75	336,370	447,738	294,531	276,908	404,675	285,248	487,980	542,816	410,319

). Under high irrigation costs, a later first sowing date in March combined with a second sowing date in July yielded the highest margins. Across all scenarios, the combination of a January planting for the first crop and a July planting for the second crop showed the highest frequency of favorable gross margins.

Finally, we assessed the impact of irrigation management strategies—specifically, the threshold of soil moisture used to trigger irrigation—on gross margins. Using the best-performing double cropping strategy (January and July planting), we tested two thresholds: (a) irrigating when TAW falls to 50% and (b) delaying irrigation until TAW drops to 30%. Delaying irrigation to the 30% TAW threshold resulted in substantial reductions in gross margins (

Table 7. Gross margins of rice double crops planted in January (first crop) and July (second crop) using two criteria to trigger irrigation: TAW = 30% and TAW = 50%. P5, P25, P75 and P95 are the 5th, 25th, 75th and 95th percentiles, respectively.

1st Crop 2nd Crop	Jan 30 * Jul 30	Jan 50 Jul 30	Jan 30 Jul 50	Jan 50 Jul 50
Median	208,875	368,340	410,435	542,975
P5	−110,907	186,703	15,953	392,922
P25	76,183	300,123	332,988	494,355
P75	318,305	487,533	471,738	607,480
P95	422,989	609,047	582,882	707,378

* Jan 30, Jan 50 = first crop planted in January using total available water (TAW) of 30% and TAW 50%, respectively, as trigger for irrigation. Jul 30, Jul 50 = second crop planted in July, using TAW 30% and TAW 50%, respectively, as trigger for irrigation.

). Specifically, applying the 30% TAW threshold for both cropping cycles led to a 64% reduction in gross margins compared to the 50% TAW threshold. Delaying irrigation in only one of the two seasons resulted in gross margin losses ranging from 25% to 35%.

4. Conclusions

The best sowing date for the four crops included in the study was based on the definition of the onset of the rainy season described by Han et al. [36], i.e., sow when the soil water content reaches 30% of the maximum plant available water holding capacity (TAW = 30%).

The yield gaps in sorghum and millet of crops with observed weather and no N limitations vs. well fertilized crops was 40–80%, indicating the large potential to improve N fertilizer management based, e.g., on recommendation systems that consider expected rainfall conditions after panicle initiation. Yield gaps in rice indicated that optimizing water use resulted in a 20% increase in grain yield and optimizing N fertilizer use resulted in additional 20% yield gains, evidencing the potential for improving both water and N fertilizer management.

Sorghum, millet and rainfed rice responded to early applications of N fertilizers, i.e., at planting and before or around panicle initiation, which are critical for early growth and development. However, the strongest response in grain yield was found at later growth stages, i.e., immediately before the crops reach their fastest growth rates. The optimal N fertilizer rates at these later growth stages depend on the rainfall, and consequently, fertilizer recommendations can be adjusted considering the expected rainfall that is included in seasonal climate forecasts.

Peanut yields and gross margins obtained with variety 73-33 were 25–100% higher than with Fleur-11. Crop yields were affected by N symbiotic fixation: yields were 20–100% lower when symbiotic fixation provided only 60% of the crop N needs. Peanut yields responded consistently to irrigation in the four locations included in the study, but the economic analyses suggested that irrigation is not a clearly viable strategy in any of the locations, primarily due to the high costs associated with water access and infrastructure.

Optimal sowing dates of irrigated rice varied from mid-June in Kolda to mid-late July in St. Louis and Bambey. The obtained yields with optimal sowing dates were similar in all locations and varied between 4000 and 5000 kg/ha. The economic analyses indicated that rice responds well to irrigation in all locations, but the large needs of irrigation can only be realistically satisfied in regions with easy access to existing water sources such as the Senegal River Valley.

Economic assessments of double rice cropping systems in the Senegal River Valley indicated that the optimal strategy for planting dates involves sowing the first crop in March and the second crop in July. This combination yielded the most favorable gross margins, especially under scenarios of high irrigation costs.

The findings of our study corroborate results from previous field experiments and further identify agronomic management practices that can enhance productivity, in-crease farmers’ income and reduce yield variability. Our results highlight the value of using crop simulation models to evaluate a broad range of management strategies under diverse climatic and economic scenarios over extended time periods. This approach facilitates the identification of practices with the greatest potential to consistently improve cropping system performance. The most promising practices identified through simulation can subsequently be tested in pilot field experiments to validate model outputs and serve as demonstration sites for the dissemination of best management practices to farmers in Senegal.