Cassava Response to Weather Variability in Eastern Africa

Abstract

Cassava is one of the most important crops in global food security. It is the second most important staple crop in Africa. Its significance is enhanced by the fact that it very well tolerates droughts, and therefore it may be a prospective response to climate change in hot and dry areas. The potentials of cassava are under-utilized in Eastern Africa, and there is a lack of research studies regarding climate impacts on cassava yields in this region. The present research focuses on cassava production in Eastern Africa, analyzing the relationship of cassava yields, harvested areas, temperature, and precipitation from 1961 to 2023. The statistical analysis applies panel regression for the 63 years of time series, for the 15 most important cassava producing countries of Eastern Africa. Findings show that while the impacts of rainfall are insignificant on yields, the effects of temperature are significantly positive, indicating yield and area increases with warming climate. An expansion of the cassava growing area and the expanding rural population contributed to decreasing yields, probably because of the expansion of smallholder subsistence farming, suffering from to limitations in other farming resources. Therefore, even if climate change may benefit cassava production, other factors create severe limitations on improving yields. However, the positive response of the crop to rising temperatures is a clear sign that it is a useful choice for food security under climate change and would deserve more support from agricultural policymakers in Eastern Africa.

1. Introduction

1.1. The Global Significance of Cassava in Fighting Food Insecurity

Cassava (Manihot esculenta Crantz) is increasingly recognized as a strategic crop in the global fight against food insecurity, particularly in sub-Saharan Africa, due to its agronomic, nutritional, and socio-economic contributions across different regions. However, due to inefficiencies related to production technology, market access, and insufficient processing facilities, it is still not utilized up to its potential [1].

According to FAO [1,2] cassava is a staple food for more than 800 million people globally, and there are varieties with yields up to 80 t per hectare in field trials, although the usual average yield is much lower. It is among the most important staple crops globally, ranking fourth after rice, wheat, and maize in terms of caloric contribution to human diets. It serves as a primary source of energy for over 800 million people, particularly in tropical and subtropical regions, where food insecurity remains a pressing challenge [1]. Its adaptability to marginal soils and tolerance to drought conditions make cassava a strategic crop for sustaining food production in areas vulnerable to climate variability. Its adaptability to marginal soils and tolerance to drought conditions make cassava a strategic crop for sustaining food production in areas vulnerable to climate variability [3]. Furthermore, cassava’s versatility extends beyond food; it is utilized for animal feed and industrial applications such as starch, ethanol, and bio-based products, thereby contributing to economic resilience in developing nations [3].

Cassava is the third most important carbohydrate source in the tropics, following rice and maize, and the second most important source of per capita calorie intake in Africa [4]. Currently, Africa holds 80% of the global cassava production area, and in 2021 global cassava production reached 315 million metric tonnes, with Africa contributing 65%, and sub-Saharan Africa 62%. The crop is well suited to smallholder farming, often farmed on plots smaller than 0.25 ha with traditional methods, and allows sustainable intensification through eco-friendly practices, as it tolerates marginal soils and minimal inputs [4]. It has been found to be resilient to climate variability, placing it as a sustainable response to natural disasters and climate change.

Cassava is primarily consumed in boiled or fried forms; processed forms are rare due to limitations of infrastructure [5]. Borku [6] emphasized cassava’s high nutritional value due to its high carbohydrate content and its protein-rich leaves. Bogale et al. [7] found that consumers respond positively to a mixture of cassava and teff flour, especially if cassava is not more than 30% of the blend. However, higher cyanide levels in certain varieties necessitate careful processing, and detoxification practices are common.

The potentials of cassava are studied for many countries of Africa, not only for its importance as a staple crop for millions of Africans, but because of its increasing significance as a cash crop, too [8]. Historically introduced to drought-prone regions to fill food gaps [9], cassava has not received adequate policy or extension support despite its ability to thrive in marginal soils and buffer against climate shocks. Food insecurity in sub-Saharan Africa is driven by a combination of demographic pressures, low agricultural productivity, environmental degradation, and recurrent droughts, especially as traditional grain-based approaches have failed to meet the nutritional needs of growing populations [10]. Cassava, as a drought-resistant and calorie-rich crop, offers a viable alternative for improving food security, particularly among smallholder farmers [11].

The world’s top exporter of cassava was Thailand in 2023, with 2812 million metric tonnes exported, which was 50.5% of global exports; followed by Vietnam (2152 million tonnes, 38.7%); the first African country, Cameroon, representing 2.82 million metric tonnes, or 0.0.51%; Nigeria with 2.63 million tonnes (0.047%); and Uganda with 1.64 million metric tonnes, or less than 0.0.3% [12].

Although cassava is widely grown in African countries, yield disparities persist throughout the continent, and Eastern African countries average 7.51 t/ha, below both the African average (8.55 t/ha) and the global average (10.62 t/ha) [4]. Adebayo [13] analyzed cassava production trends across 37 African countries, highlighting that 95.6% of production variability is explained by changes in the area harvested, while only 1.1% is due to yield variability.

According to the analysis by Adebayo [13], Nigeria, the Democratic Republic of Congo, and Ghana contribute 68.3% of Africa’s production increase, and Nigeria alone produced 59.4 million tonnes in 2019. This is nearly 20% of the world total, making Nigeria the world’s largest cassava producer. According to FAOSTAT [14] significant growth was experienced from 2000 to 2010 in Angola (the output of 2010 being 216% of that of 2000), Cameroon (301%), and Sierra Leone (537%), and Eastern African countries also showed impressive increase: Zambia (480%), Burundi (429%), Mauritius (394%), Kenya (212%), and Malawi (210%). However, average yields remain low at 7.7–8.9 t/ha, far below the potential yields of 40–80 t/ha under optimal conditions. West Africa dominates production, though low yields, especially in Nigeria (5.9–6.4 t/ha from 2020 to 2023), and low production in Eastern Africa are explained by problems of cassava mosaic and brown streak diseases [13]. As the example of the Democratic Republic of Congo shows, seedbed preparation and the right choice of local landraces could increase tuber yields by 32–68%, pointing to the importance of proper agronomy [15].

1.2. The Importance of Cassava in Eastern Africa

Cassava is a tropical root crop that needs at least 8 months of warm temperature to establish and grow. Cassava does not tolerate freezing conditions. It tolerates a wide range of soil pH from 4.0 to 8.0, is most productive in full sun, and is a hardy crop that produces tubers under marginal conditions, e.g., in drought or depleted soils [16]. Its tuberous root contains 30–40% dry matter and 25–30% starch. Nutritionally, cassava contains potassium, iron, calcium, vitamin A, folic acid, sodium, vitamin C, vitamin B-6, and protein. Cassava roots and leaves are good sources of carbohydrates, protein, vitamins, and minerals [17].

Although cassava has high potentials with respect to climate resilience, to generation of small-farm income, and to the fight against hunger and food insecurity in Eastern Africa, it is still largely under-utilized and not well researched in the region.

The beneficial contribution of the crop to household food security was established by a study of Tanzania [18], in spite of constraints presented by the prevalence of pests, poor market access, lack of processing infrastructure, and insufficient knowledge of proper agroeconomic practice.

Ethiopia also suffers from the under-utilization of cassava production [19], although the crop could well tolerate the climate variability and the poor soils. Processing is still rudimentary and poses health risks, notably cyanide poisoning, and without a multi-sectoral collaboration to improve cassava varieties, processing technologies, and value chains, progress cannot be reasonably expected [19,20,21]. To enhance productivity, improvements in the education level, cropping system, plant topography, and pest incidence [22] would be needed, as well as improved use of mobile phone use, land allocation, and better access to credit and extension services [8].

Cassava is a wide-spread crop in Uganda, with undoubted economic importance as a food, feed, and industrial crop [23], but, although disease-tolerant and high yielding improved varieties are available, farmers still grow local varieties of the crop. Countries such as Uganda, Tanzania, and Kenya have incorporated cassava into farming systems as a safeguard against crop failures resulting from erratic rainfall and soil degradation [24]. Its capacity to thrive in low-fertility soils and endure prolonged dry spells makes it indispensable for smallholder farmers facing climate uncertainty. Beyond its role in subsistence farming, cassava is increasingly driving agro-industrial development in the region, with expanding opportunities in starch processing, animal feed production, and bioenergy [25].

Similarly to other countries, cassava in Malawi plays a crucial role as a staple crop, feeding about 30 to 40% of the population, second only to maize [26,27], and its importance is underlined by the fact that it is considerably more drought tolerant than maize.

1.3. Cassava as a Response to Climate Change

Cassava possesses exceptional resilience to drought, high temperatures, and low-fertility soils, making it a strategic component of climate change adaptation and food security [1,3]. Its resilience to climate variability has been extensively researched, evaluating its yield and quality under various weather conditions.

The considerable resilience of cassava to climate variability is due to its tolerance to high temperatures, increased solar radiation, and water stress. This is achieved by the crop’s response to these factors of modifying its stomatal conductance and photosynthesis rates, while it can respond positively to higher CO₂ levels, with increasing yields and better tolerance to salinity and water deficits [28]. High temperatures with drought occurrence can lead not only to yield reduction but to increased toxicity in cassava, but high temperatures with adequate water availability can reduce toxicity. Therefore, precise irrigation management should be implemented in drought-prone areas [29].

According to Pipitpukdee et al. [30] climate change affects land use, yield, and overall production of cassava in Thailand, with a U-shaped relationship between temperature and yield. A moderate temperature increase was found to increase cassava yields up to a certain temperature limit, above which further warming leads to declining yields. Rainfall was also found to affect cassava production, especially the harvested area of cassava, as were extreme weather events, like La Niña. The mentioned study indicated a predicted 15–21% yield reduction, compared to baseline levels, due to climate change [30].

In Africa cassava yields were found to show mixed responses to various climate change scenarios, and potential improvements or declines were found in different regions with different regional climate change projections [31]. To fully utilize the resilience of the crop to weather variability, strategies of adaptation are advisable, including the introduction of irrigation facilities, the adoption of drought-tolerant varieties, and better soil management practices [32,33]. Under various climate change scenarios, cassava production was forecasted to change from –3.7% to +17.5% compared to its baseline levels [34]. The optimal growth conditions were found to be the temperature range of 25–29 °C with 1000–1500 mm annual rainfall, while elevated CO₂ resulted in enhanced photosynthesis and yield [28]. However, climate variability may increase the occurrence of extreme weather events, salinity risks, and shift pest and disease patterns, which may be damaging [34]. Regional studies indicate varied responses; South Africa may expect an increase, leading to industrial potential for starch production [35], while Brazil and China may face vulnerabilities in their different agroclimatic zones, harming their bioethanol production [36,37]. Nutritional limitations (low protein, high cyanide content) and pest pressures remain critical constraints [38]. Breeding for stress resistance and quality enhancement, alongside policy support, is essential to sustain cassava’s dual role in food security and industrial applications.

1.4. Models of Cassava Response to Climate Change

To evaluate cassava production in response to climatic variables, Pipitpukdee et al. [30] use the following models:

Area_it = α₀ + α₁Climate_it + α₂Price_it + α₃Irrigation_it + α₄Popdensity_it + α₅T_it + α₆T²_it + v_i + e_it

(1)

Yield_it = β₀ + β₁Climate_it + β₂Price_it + β₃Irrigation_it + β₄T_it + β₅T²_it + u_i + ϵ_it

(2)

In these relationships ‘Climate’ refers to mean and maximum temperature in the growing season and to total and maximum daily rainfall. ‘Irrigation’ is the percentage of irrigated area, and ‘Popdensity’ is the population density for the relevant location i and relevant year t. ‘Price’ is composed of cassava farm price and labor wage, while T is a variable representing the year, capturing technological advancement. The model coefficients were estimated for 77 provinces in Thailand, for the years from 1989 to 2016, using a panel regression analysis.

Orimoloye and Adigun [39] used monthly temperature and rainfall data from 2005 to 2009 in Nigeria to estimate the impact of these variables on cassava and maize yields, using a linear trend analysis. Their findings show that maximum temperature and rainfall significantly influenced cassava yields, the mean temperature impact was negative, the rainfall effect was positive, and minimum temperature did not have any significant impact.

Boansi [40] estimated the impact of several influencing factors on cassava yield, namely the area of cassava harvested; the total rural population as a proxy for labour availability; the real producer price ratios of cassava to maize, yam, and common beans; the nominal exchange rate; the total volume of rainfall by seasons; and the mean temperature by seasons. Using a linear relationship for the logarithms of these variables, the logarithm of cassava yield was estimated.

Srivastava et al. [41] estimated the influencing factors of yields for Nigeria using a yield gap model. Their model included temperature, rainfall, and total radiation, and their findings indicate that precipitation and total radiation may be the most influential factors on cassava yield variability.

Mathematical models have been applied for estimating the yields and production of other crops in response to changing weather patterns. Miao et al. [42] used a similar model to that of Pipitpukdee et al. [30] to estimate the impact of climatic variables on the yields and areas of soybean and corn, though they did not find the impacts of prices significant on soybean yields. Amin et al. [43] used a linear trend model with minimum and maximum temperatures, total rainfall, and sunshine and humidity as independent variables to estimate their impact on the yields and cultivation areas of three rice crops and one wheat crop in Bangladesh. Their findings show very mixed impacts: sunshine, minimum temperature, and humidity were not significant on the dependent variables in most of the cases, but there was a considerable impact of the time trend, representing technological change and socioeconomic conditions.

As was shown in the above literature review, although the relationship between cassava production and climatic variables is a popular research topic, the climate impacts on cassava production in Eastern Africa are not well researched. Existing results deal with several African countries, covering a time span of not more than 30 years and neglecting the Eastern African region. However, according to the literature the response of cassava production to climate trends highly varies across regions and time windows, and therefore findings from other regions of the world cannot be easily adapted to Eastern Africa. Our analysis covers 63 years and 15 countries—all major cassava producers—in Eastern Africa, addressing this research gap.

2. Materials and Methods

2.1. Data

Cassava yield, harvested area, and total production data were collected from the FAOSTAT database for 15 countries (see Figure 1), Burundi, Comoros, Kenya, Madagascar, Malawi, Mauritius, Mozambique, Réunion, Rwanda, Seychelles, Somalia, Uganda, Tanzania, Zambia, and Zimbabwe, for the years 1961 to 2023 [14]. Climate variables, namely total annual precipitation and mean annual and seasonal temperature, were collected for the same countries and the same time period from the World Bank Climate Knowledge Portal [44]. Data for agricultural land use, population, and fertilizer application were derived from the FAOSTAT database [45,46,47]. The analysis used the following variables (see

Table 1. List of variables used in the analysis.

Variable Name	Meaning of Variable	Unit	ln-Transformed Value	Source of Data
Area	Total harvested area of cassava	hectares	lnArea	[14]
Yield	Cassava yield	kg per hectare	lnYield	[14]
Prod	Total annual output of cassava	tonnes	LnProd	[14]
Temp	Mean annual temperature	°C	LnTemp	[44]
Precip	Total annual precipitation	mm	LnPrecip	[44]
TS1, TS2, TS3, TS4	Seasonal mean temperatures for January–March, April–June, July–September, October–December	°C	LnTS1, LnTS2, LnTS3, LnTS4	[44]
ALand	Agricultural land as % total land area	%	LnALand	[45]
Nfert, Kfert, Pfert	Annual fertilizer use kg/ha, N, K and P fertilizers, resp.	kg/ha	LnNfert, LnKfert, LnPfert	[46]
CArea	Total land area of country	1000 ha	LnCArea	[46]
TPop	Total population	1000 persons	LnTPop	[47]
RPopPct	Proportion of rural population	%	LnRPopPct	[47]
Year	captures the annual technological and socioeconomic changes	1961–2023	lnYear	-

):

Time series were downloaded from the data sources denoted in Table 1. Altogether 945 cases (15 for 63 years each) were derived, for 14 variables, and then their ln-transformed values were taken. As not all data were available for each country and each year, data cleaning was performed by deleting cases with missing data, leaving 912 cases altogether, after deleting 33 country–year pairs.

2.2. Methods of the Statistical Analysis

Following the simple descriptive analysis of the variables, correlation analysis was performed between cassava production and weather variables. Then, as the core part of the analysis, a panel regression model was set up, based on the approaches of previous studies. The predictor variables were chosen following the previously reviewed literature. Therefore, yield was assumed to be influenced by precipitation and temperature [28,29,31,40,41], which is the primary focus of the analysis. Besides annual mean values of precipitation and temperature, seasonal mean temperature values were also included [40], as cassava shows different needs and sensitivities during the various stages of the crop growth cycle [2,41]. Other influencing factors were also considered, such as the harvested area [13,40], technology level, and nutrient availability [2,3]. As there are no detailed annual countrywide data about technology level and nutrient availability for the cassava crop, control variables were introduced instead. Overall countrywide fertilizer application of N, P, and K nutrients were used to capture the general nutrient supply for crops, and the size of the total agricultural land area expressed as percentage of the total land area was used to represent the importance of agriculture in the national economy. High levels of development are generally associated with decreasing shares of agricultural land, but this trend may change in countries with highly intensified agricultural sectors. Developing nations also have high proportion of their territory under agriculture, this being a major sector of their economies [49,50]. Although the development status of a country would be better captured by the share of agricultural income within GDP, limitations on data availability did not facilitate this choice. Therefore, the share of agricultural land and the level of fertilizer application can both reflect agricultural development and overall development differences. The total land area, the total population, and the percentage of rural population were added as control variables.

The theoretical foundation for the model is derived from the Cobb–Douglas production function [51,52], which expresses total output (Q_t) at time t, as a multiplicative function of inputs: capital (K), labour (L), and productivity (A), as Q_t = B × A_t × K_t^α × L_t^1−α. A generalized form of this equation is Q_t = B × A_t × R1_t^α1 × R2_t^α2 ×…× Rn_t^αn, where R1, …, Rn represent n different resources or factors of production, A represents technological progress, and B a constant factor. If Q_t is considered to be the agricultural output per hectare of a certain crop, then R1, R2, …, Rn can include physical, human, and natural resources [51]. Taking the natural logarithm of this equation, we arrive at the additive equation form: lnQ_t = B + lnA_t + α1 × lnR1_t+ α2 × lnR2_t+…+ αn × lnRn_t. If the output is agricultural yield, then the production resources can reasonably be taken as the influencing factors listed above: temperature, precipitation, area harvested, fertilization, and a factor describing the level of development. Based on these considerations the following model versions will be estimated:

lnYield = Intercept + a1 × lnWeather +a2 × lnArea + a3 × lnYear + a5 × lnFert +a5 × lnALand + a6 × lnControls

lnArea = Intercept + b1 × lnWeather +b2 × lnYear + b3 × lnFert + b4 × lnALand + b5 × lnControls

This approach is widely applied in previous studies. In this model the lnYear variable represents the effects of general socioeconomic and technological changes not captured by other variables. We tested several model versions, with slight modifications of the above equation forms. The temperature was tested with its de-meaned version (lnTempDM), and the square of the de-meaned temperature variable (lnTempDMSq) was included to capture a possible U-shape effect of temperature, as in [30], to incorporate the fact that there may be an upper temperature threshold for cassava growth. In addition, instead of the annual mean temperature (lnTemp), the seasonal temperature means (lnTS1, lnTS2, lnTS3, lnTS4) were also applied. The ln specification matches a multiplicative (elasticity) form, as in the Cobb–Douglas production function, which is generally accepted. This formulation is extensively used in agricultural production analyses accommodating a multiplicative association amongst inputs (area, climate variables, labor/capital, etc.) and output (yield or production) [51]. Applying natural logarithms also helps to (i) diminish skewness and kurtosis in highly skewed variables (i.e., land area, production, yield), making the distribution of transformed variables closer to normal; (ii) alleviate heteroskedasticity common in cross-sectional/time series panel data; and (iii) allow coefficient interpretation in terms of elasticities—e.g., a 1% change in precipitation corresponds to a percentage change in yield. Numerous recent studies adopting this methodology clearly state these reasons [52].

The model parameters were estimated using panel regression by the Linear Mixed-Effect Model estimation [53,54], using the MATLAB software package (release 25.2.0.3055257—R2025b, Update 2) [55], together with the Panel Data Toolbox for MATLAB [56]. For the reason that the dataset was the combination of cross sectional and time-series variation (panel structure), the use of panel regression allows for regulatory unobserved heterogeneity across cross-sectional units (e.g., region-specific fixed features such as persistent soil quality, institutional context, or management style) and for chronological trends distressing all units (e.g., general technological progress, policy changes). This enables the reduction in omitted-variable bias. Several studies in the literature on climate impacts on agriculture yield follow such an approach [57].

The panel data model can be formulated according to the following formula:

yit = α + Xitβ + µi + vit, i = 1,…, n, t = 1, …, T.

(3)

where i denotes individuals (e.g., countries) and t the time (years), α represents the overall grand mean for all countries and years, and µ_i represents the ith time-invariant individual effect, while v_it is the error component, assumed to be of zero expected value. The matrix X contains the predictor vectors, and the vector β contains the coefficients describing the effects of the predictor vectors on the dependent variable yit. The additive model is based on the ln-transformed format of the Cobb–Douglas multiplicative production function [56].

A fixed-effects model assumes that the µi intercepts are fixed values, not dependent on time, and they can correlate with the X predictor matrix. A random-effects model, however, assumes that the µi values are random values and are strictly uncorrelated with X and the error term. In the fixed-effects model, it is common to take the overall intercept as the mean of the individual effects; therefore, the formula may contain this overall intercept, and the individual effects can represent deviations from this overall mean. However, in the random-effects model, the individual deviations around the overall mean are not fixed values but random values following some specific distribution.

Testing whether the error term correlates with the regressor variables is the way to decide whether the fixed-effects model is appropriate or the random-effects model should be chosen. The classic test applied for this purpose is the Hausmann test. However, when in spite of the ln-transformation of the variables, the data still suffer from heteroskedasticity, a more robust version of the fixed-effects and random-effects model can be applied using the Panel Data Toolbox, and the robust Mundlak test can be applied to choose between a fixed-effects model and a random-effects model of panel regression [56]. The Mundlak test has a null hypothesis that the estimated coefficients are jointly zero, and there is no correlation between the time-invariant error term (unobservable factors) and the regressors. It can be applied reliably to models in which variables do not follow homoskedastic and normal distributions. The significance of the test value, i.e., acsmall p-value, suggests that the null-hypothesis is rejected, and the fixed-effects assumptions are satisfied. The validity of the β parameters (i.e., their being significantly different from zero) is tested by the Wald test—where the significance of the test value (i.e., a small p-value) means that the parameter vector is significantly different from zero [56].

3. Results

3.1. Descriptive Statistics of the Dataset

Table 2. Descriptive statistics and normality test for variables (for variable names see Table 1).

					Shapiro–Wilk
	Mean	Std. Dev	Min	Max	W	p
Yield kg/ha	8304.75	4728.280	1177.8	35,538.00	0.89160	3.14 × 10−25
Area harvested, ha	199,463.82	298,409.575	3.0	1.629 × 10+6	0.70017	1.14 × 10−37
Production, tonnes	1.329 × 10+6	1.865 × 10+6	75.0	8.372 × 10+6	0.73782	7.63 × 10−36
Precip, mm	1232.03	482.300	139.51	2849.55	0.99220	0.00007
Temp, °C	23.10	2.124	18.64	28.04	0.98368	8.85 × 10−9
Aland, %	48.47	19.595	3.30	83.90	0.96832	1.78 × 10−13
Kfert, kg/ha	11.84	30.626	0.00	162.09	0.42624	1.72 × 10−47
Nfert, kg/ha	21.55	39.139	0.00	173.91	0.59509	3.14 × 10−42
Pfert, kg/ha	5.38	14.173	0.00	126.15	0.39636	2.77 × 10−48
TPop, 1000 persons	10,392.848	12,098.432	44.10	66,617.6	0.7955	7.60 × 10−33
RPopPct, %	68.933	22.110	0.2590	102.55	0.8992	1.98 × 10−24
CArea, 1000 ha	34,113.167	33,412.150	46.00	94,730.0	0.8379	6.22 × 10−30
Ln Yield	8.92	0.512	7.071	10.478	0.99677	0.05613
LnArea	9.95	3.365	1.099	14.303	0.86360	1.24 × 10−27
LnProd	11.95	3.210	4.317	15.940	0.87256	7.66 × 10−27
LnPrecip	7.01	0.503	4.938	7.955	0.89069	2.53 × 10−25
LnTemp	3.14	0.092	2.925	3.334	0.98456	1.95 × 10−8
LnTPop	8.241	1.7964	3.7865	11.1067	0.9230	1.46 × 10−21
LnRPopPct	4.101	0.7476	−1.3509	4.6304	0.5174	7.70 × 10−45
LnCArea	8.849	2.5939	3.8286	11.4588	0.8221	4.40 × 10−31

shows the main characteristics of the dataset used for the study. As minimum and maximum values indicate, the harvested area and the total production varies greatly between countries and years. The yields also show great variation from 1.18 t/ha to 35.5 t/ha, with an average of 8.3 t/ha during the analyzed 63 years. Mean annual precipitation and temperature were 1232 mm and 23.10 °C, respectively, with moderate variability. The distribution of the data deviates from the normal distribution both for the original data and the ln-transformed data, with the only exception of lnYield, as small p-values of the Shapiro–Wilk test show. However, the further statistical analyses do not require normal distribution of the variables, so this does not hinder the analysis.

The trends of cassava production are presented in Figure 2. The largest producers in the 2020s are Tanzania, Mozambique, and Malawi (with 6 to 8 million tonnes annually), followed by Zambia, Burundi, Madagascar, and Uganda (2 to 4.5 t), then Rwanda and Kenya (1 to 1.3 t), while the rest of the countries produce less than 0.25 tonnes each. Most of these countries show a steady growing trend from 1960 to 2023, with the remarkable exception of Uganda. The largest production areas, however, are possessed by Uganda, Tanzania, and Mozambique (800 to 1200 thousand hectares each), followed by Burundi, Zambia, Madagascar, and Malawi (250–360 thousand hectares), while the rest of the countries have a much less area, under 100 hectares. It is worth noting that in Réunion, with a production area of 500 hectares in 1961, the land allocated to cassava showed a steady decline, and by 1998 the area under cassava fell under 100 hectares, and from 2007 no cassava-producing area is recorded in the crop statistics. This is rather surprising, considering that between 1997 and 2006 the yields were between 11 and 20 t/ha, which is not the lowest in the region.

The fluctuations in production are well explained by the variations in yields. The steadiest yield increase was achieved in Malawi, where the average annual yield reached nearly 25 t/ha by 2023. While Zambia achieved outstanding yields between 2015 and 2021 of 35 t/ha, these high values could not be maintained, and by 2022 and 2023 yields fell back to 11.7 t/ha. Several countries of the region produced yields in the range of 12–15 t/ha in 2022–2023 (Kenya, Mauritius, Rwanda, Seychelles), while another group (Somalia, Mozambique, Madagascar, Tanzania, Comoros, and Zimbabwe) achieved low yields between 5 and 10 t/ha. It is surprising to see that Uganda, with a medium yield of 15 t/ha between 1999 and 2008, turned to a decreasing yield path, and by 2023 its yield is less than 2 t/ha, the lowest in the region.

Although it is not the aim of the present study to give a thorough climate analysis for the selected Eastern African countries, the following figures and tables present a general overview of the temporal trends of rainfall and temperature from 1961 to 2023 (Figure 3 and

Table 3. Rainfall and temperature statistics 1961–2023, forecast for 2030, and parameters of a fitted linear regression trend, for Eastern Africa.

Rainfall	Min	Max	Mean	Forecast-2030	R2	Slope	Intercept
Burundi	1763.1	2531.9	2076.0	1996.89	0.0361	1.651	−1212.2
Comoros	920.3	1731.6	1263.1	1198.96	0.0007	−0.207	1674.7
Kenya	515.7	1109.5	745.0	654.81	0.0395	1.444	−2132.0
Madagascar	1115.5	1693.1	1412.5	1103.88	0.0123	0.752	−85.6
Mozambique	643.0	1079.2	869.8	768.71	0.0119	−0.599	2062.5
Mauritius	705.5	1605.1	1075.7	822.71	0.0028	0.563	−45.4
Malawi	978.4	1595.8	1342.0	1162.48	0.0011	−0.254	1847.9
Reunion	1049.0	2849.6	1684.6	1532.1	0.0317	3.340	−4967.8
Rwanda	1411.2	1991.6	1682.3	1789.4	0.0910	2.108 *	−2517.8
Somalia	139.5	458.7	284.8	250.06	0.0135	0.399	−510.1
Seychelles	1022.4	1603.7	1318.7	1911.93	0.00004	−0.043	1403.4
Tanzania	919.7	1625.7	1148.9	1065.28	0.0169	0.863	−570.5
Uganda	1392.4	2356.2	1799.7	1815.13	0.0637	2.660 *	−3499.8
Zambia	843.4	1262.5	1093.3	927.39	0.0260	−0.812	2710.4
Zimbabwe	392.0	1089.3	684.1	587.07	0.0172	−0.990	2656.4
Temperature	Min	Max	Mean	Forecast-2030	R2	Slope	Intercept
Burundi	19.7	21.7	20.5	21.99	0.6353	0.0203 **	−19.922 **
Comoros	24.6	26.1	25.2	26.04	0.6174	0.020 **	−6.3287 **
Kenya	24.4	26.5	25.3	24.7	0.7130	0.024 *	−18.372 **
Madagascar	22.2	23.6	22.8	26.59	0.7003	0.024 **	−12.945 **
Mozambique	23.3	25.2	24.1	24.04	0.5991	0.026 **	−15.107 **
Mauritius	23.1	24.6	23.7	25.37	0.5536	0.020 **	−6.636 #
Malawi	21.4	23.4	22.2	24.38	0.5766	0.030 **	−20.485 **
Reunion	20.2	21.6	20.8	23.54	0.5416	0.019 **	−6.783 **
Rwanda	18.6	20.6	19.4	21.53	0.6638	0.023 **	−19.901 **
Somalia	26.1	28.0	26.9	20.76	0.6623	0.027 **	−15.807 **
Seychelles	25.4	27.0	26.1	28.13	0.6262	0.023 **	−9.423 **
Tanzania	22.1	23.9	22.8	26.93	0.6511	0.023 **	−16.340 **
Uganda	22.2	24.1	23.0	24.23	0.6121	0.019 **	−13.331 **
Zambia	21.3	23.7	22.2	24.44	0.6452	0.0320 **	−29.261 **
Zimbabwe	20.4	23.3	21.5	23.81	0.4809	0.0320 **	−29.524 **

Source of 2030 forecasts: World Bank Climate Change Knowledge Portal [44]. #: significant at 10%, *: significant at 5%, **: significant at 1%.

).

Annual rainfall values vary greatly among the countries (see Figure 3 and Table 3). The driest country, consistently, is Somalia, with annual sums varying between 140 mm and 459 mm, with an average of 285 mm, followed by Zimbabwe, Mozambique, and Kenya (averaging between 681 and 870 mm), while the other extremes are Burundi and Uganda (2076 and 1800 mm average values, respectively).

However, analyzing rainfall trends, there is no significant trend, neither positive or negative, as none of the temporal trends nor the regression slopes differ significantly from zero—with Uganda and Rwanda as the exceptions, where slight positive trend coefficients of 2.1 and 2.6 mm per year were found to be significant. However, these values are extremely small, indicating less than 0.1% annual increase in the average precipitation. The forecasted values for 2030 were derived from the projections of the World Bank Climate Change Knowledge Portal [44].

Table 4. Correlation coefficients between cassava production and weather variables.

	Yield	Area	Prod	Precip	Temp	TS1	TS2	TS3
Yield	1.0000
Area	−0.2592	1.0000
Prod	0.0782	0.8375	1.0000
Precip	−0.0393	0.0446	0.0667	1.0000
Temp	0.1141	0.0108	−0.0245	−0.5882	1.0000
TS1	0.0411	0.0796	0.0173	−0.5560	0.8715	1.0000
TS2	0.1337	−0.0631	−0.1024	−0.5283	0.9732	0.8714	1.0000
TS3	0.1687	−0.0632	−0.0821	−0.3609	0.8684	0.5564	0.8562	1.0000
TS4	0.0401	0.1222	0.1170	−0.6976	0.8550	0.7546	0.7368	0.6106

illustrates the correlation between cassava yield, production, harvested area, temperature, and precipitation data. The strong positive correlation between harvested area and total production is not surprising, but it is surprising to see, that the correlation between total production and yield per hectare is rather weak. The negative correlation between yield and area harvested, although weak, may be somewhat surprising, too. The strong negative correlation between mean annual temperature and total annual precipitation is in line with expectations, indicating a warm and dry weather pattern for Eastern Africa. This is also reflected by the negative correlation between precipitation and seasonal temperatures, though it is less marked for the July–September season. Annual mean temperature has a strong positive correlation with mean temperatures of each season.

3.2. Results of the Panel Regression Models

Table 5. Parameter estimates of fixed-effects models for LnYield (15 countries, 912 valid cases).

	Model-1		Model-2		Model-3		Model-4		Model-5
R2	0.1830		0.1852		0.2055		0.2155		0.2157
Mundlak test	55.7029	***	143.1093	***	1511.8	***	94.3787	***	143.5588	***
Independent	Coefficient		Coefficient		Coefficient		Coefficient		Coefficient
lnYear	6.3389		6.6203
LnArea	−0.2077	#	−0.2141	#	−0.1963	#	−0.1747		−0.1987	#
LnPrecip	−0.0521		−0.0469		0.0006		−0.0457		−0.0405
LnTemp	4.1153	#
LnTempDM			3.8116	#
LnTempDMSq			−5.5294		−7.0245
LnTS1					1.2616		0.9949		1.0721
LnTS2					1.0965	#	0.9923	#	1.0054	#
LnTS3					0.5851		0.6449		0.6483
LnTS4					2.1139	*	1.7289	#	1.7400	#
LnNfert					0.0103
LnPfert					−0.0838	#	−0.0985	*	−0.0950	*
LnKfert					0.0400		0.0486		0.0517
LnALand					0.0960				0.1419
LnCArea	8.3480		8.5470		7.9985		9.6514
LnTPop	0.0627		0.0680		0.2199	*	0.2778	**	0.2658	**
LnRPopPct							−0.3900	**	−0.3669	**
Constant									−4.8878

***: p < 0.001, **: p < 0.01, *: p < 0.05, #: p < 0.1, models robust to heteroskedasticity, adjusted for 15 clusters, unbalanced panel.

and

Table 6. Parameter estimates of fixed-effects models for LnArea (15 countries, 912 valid cases).

	Model-1		Model-2		Model-3		Model-4		Model-5
R2	0.2338		0.2574		0.4137		0.3807		0.3747
Mundlak test	113.025	***	166.841	***	61,464.000	***	162.505	***	41.148	***
Independent	Coefficient		Coefficient		Coefficient		Coefficient		Coefficient
lnYear	10.4748		11.5040				7.5280
LnPrecip	0.2674	#	0.2844	#	0.1782		0.1350		0.1623	#
LnTemp	6.0486	#
LnTempDM			4.4043
LnTempDMSq			−26.5522		5.7639
LnTS1					0.2946
LnTS2					1.9006	#	1.5502	*	1.9444	#
LnTS3					1.5062	*	1.5451		2.0552	#
LnTS4					1.1035
LnNfert					0.0069
LnPfert					0.1533	*	0.1522	*	0.1545	*
LnKfert					−0.0044
LnALand					1.2943	*	1.1804	**	1.1825	**
LnCArea	−37.0960	**	−34.9981	*	−55.6428	*
LnTPop	0.0501		0.0741		0.1132				0.1440
LnRPopPct							0.2438		0.1979
Constant									−9.6710	*

***: p < 0.001, **: p < 0.01, *: p < 0.05, #: p < 0.1, models robust to heteroskedasticity, adjusted for 15 clusters, unbalanced panel.

present the results of the fixed-effects panel regression for the yield and the harvested area as dependent variables, respectively. Five slightly different model versions were tested, and each model used the ln-transformed values of the variables. The first model was the simplest, with the ln-values of precipitation, temperature, and harvested area as main predictors for the yield model, adding the ln-transformed value of year, total population, and country area as control factors capturing the impacts of the changing external socioeconomic environment. The second model version used the de-meaned version of the temperature variable together with its squared value, with the purpose of capturing any nonlinear effect of temperature. The third model version omitted the lnYear variable and introduced the ln-values of fertilization and agricultural land share, and instead of the annual mean temperature it included the ln-transformed values of seasonal mean temperatures. The fourth and the fifth models added the ln-transformed value of the percentage of rural population as an indicator of labour availability for agriculture and slightly modified the control variables while keeping the weather variables that were found significant in the previous models.

The panel regressions were set up both for the fixed-effects model version and the random-effects model version, and then the Mundlak test was applied to decide which of the two is the proper one. As the Mundlak test values were significant in each model, the relevance of the random-effects model was rejected, and the fixed-effects model was chosen as the proper one. The R² values are rather modest, indicating a moderate fit and predictive power, but our aim was not to set up a full predictive model but to identify how the changing weather patterns influence cassava production.

As the coefficients and the associated p-values show, the impact of precipitation is not significant for LnYield. A rather consistent positive significance of LnPrecip was found in three models for LnArea but not for the other two model versions. The coefficients indicate that an increase of 1% in the annual precipitation leads to a 0.16 to 0.28% increase in harvested area. In the first two models the impact of LnYear was directly included but was not significant either for LnYield or for LnArea.

The impact of temperature shows a more complex pattern. In the models for LnYield, its impact was significantly positive, both for the simple annual mean and for the de-meaned annual mean, and also for the seasonal means, while the squared value of the de-meaned annual mean showed no significant impact on LnYield. As the coefficients show, a 1 percent increase in mean annual temperature leads to an approximately 4% (3.8 to 4.1) increase in yield, or a 1% warming in the second and fourth seasons (April to June or October to December) leading to a 1% or 1.7 to 2.1% yield rise, respectively. For the LnArea models a 1% increase in the annual mean temperature leads to a remarkable 6% rise in the cassava area, or a 1% warming in the second and the third seasons initiates a 1.5–2.05% increase in the harvested area. The squared value of the mean temperature was not found significant, indicating that the effect of temperature in the current temperature range is linear in Eastern Africa. The positive impacts of temperature on the harvested area are reasonable, as higher yields make cassava production more profitable; therefore, it is reasonable to allocate larger areas to the crop [58]. For the yield model the harvested area was also included as an independent variable, and its impact is significantly negative on yield: a 1% increase in the production area decreases yield by 0.19–0.21%. This may be due to the standard microeconomic concept that the areas brought later under cultivation are usually less productive, leading to a decrease in average yield [58].

Models 3 and 4 show that while N and K application did not have any impact, the P-fertilization negatively influenced the yield (1% increase in fertilization leading to a 0.08–0.099% of yield decrease) and positively the production area (1% increase in fertilization leading to a 0.15% increase in area). As the fertilization data do not refer to the cassava crop but to an overall application, higher P-fertilization probably indicates more intensive agricultural management approaches, which is reflected in the slight increase in cassava areas. The decreasing yield may mean that the P-fertilization is applied to crops better responding to it than cassava, such as maize, cereals, and leafy vegetables [2,59,60]. Cassava in Eastern Africa shows less response to phosphorus than to nitrogen or potassium due to its efficient symbiotic relationship with soil fungi [2]. The share of agricultural land did not show any significant effect on cassava yields, but a positive effect was found on the cassava production area indicating a 1.2% response to 1% growth in agricultural land, which suggests that the cassava area increased its share within the total agricultural area in the region.

Finally, it is worth noting that the cassava area showed a negative relationship to the total country area: a country having a 1% larger area measured in thousand hectares tends to have 35–55% smaller cassava area measured in hectares, suggesting that larger countries tend to specialize in other, possibly more profitable, crops. The increase in total population, on the other hand, had a positive relationship to cassava yields, while the increase in the proportion of rural population had the opposite impact.

Choosing the fifth model version, as one with the best—or nearly the best—R-square value, the significant predictors and their coefficients can be used to model the yield and area trends from 1961 to 2023. These modelled series serve only as illustrations, as the modelled equations are not reliable enough, but they can indicate overall tendencies and ideas about the future outlook. The following model equations were derived from Model 5 in Table 5 and Model 5 in Table 6. The ln values were transformed to level values, and Figure 4 shows the example of a few selected countries.

Yield = − 4.888 − 0.199 × LnArea + 1.005 × LnTS2 + 1.740 × LnTS4 − 0.095 × LnPfert + 0.266 × LnTPop − 0.367 × LnRPopPct

LnArea = −9.671 + 0.162 × LnPrecip + 1.944 × LnTS2 + 2.055 × LnTS3 + 0.154 × LnPfert + 1.182 × LnALand + 0.198 × LnRPopPct

As the six examples show, there is a definite increasing tendency of cassava areas, while yields show a decreasing pattern; although, for some countries the decrease seem to slow down, or stop (Kenya, Malawi, Tanzania). If the climatic, agronomic, and socioeconomic environment progresses in the same direction in the future, then the same yield and area patterns can be expected for the near future.

4. Discussion

4.1. Cassava Yield Responses to Changes in Weather

The analysis of cassava production in 15 countries of Eastern Africa revealed a specific relationship between cassava yields and rainfall and temperature. The basic findings showed that, based on historical data of the time period of 1961 to 2023, precipitation has no significant effect on cassava yields in Eastern Africa. A similar result of no significant relationship was found between rainfall and cassava yield in Lagos between 1995 and 2018 [32], and in Zimbabwe between 2007 and 2011 [61]. As Mupakati [61] explains, cassava is not affected by lack of water, except at around planting, when it requires sufficient water to initiate its development. Amin et al. [43] also found that rainfall had no effect on wheat and rice yields in Bangladesh.

Significant positive effects of rainfall were found on cassava yields in Nigeria between 1995 and 2019 [41], and between 2005 and 2009 [39], in Togo between 1978 and 2009 during the main production season [40], and in 77 provinces of Thailand between 1989 and 2016 [30]. Jarvis et al. [34] found that rainfall was positively related to total production in Uganda, Tanzania, and Nigeria, but had a negative impact in the Democratic Republic of Congo, Mozambique, Angola, Ghana, Madagascar, and Ivory Coast. Blanc [31] also established a negative impact of rainfall on cassava yields for sub-Saharan Africa between 1961 and 2002. These findings show the rather mixed response of cassava yields to rainfall, depending on the geographical location, and the associated average rainfall levels, even within Africa. The Eastern African region has not been thoroughly analyzed in this respect so far. Our analysis showed that during the past 63 years precipitation patterns fluctuated a lot, but no significant increasing or decreasing trend could be established, and the fluctuations did not influence yields significantly.

Our findings show that temperature significantly and positively affected cassava yields. This suggests that, in spite of the rising temperature, cassava yields may respond positively to climate change in the region. This is in agreement with Boansi [40] regarding cassava production in Togo in the short run, with Jarvis et al. [34] for cassava production several African countries, and with Adebayo [13] in African regions. Contrary to our findings, no significant relationship was found between cassava yields and temperature in Nigeria in 1995–2019 [41], nor in sub-Saharan Africa between 1961 and 2002 [31]. A short-term analysis of cassava production in Nigeria between 2005 and 2009 [39] revealed a negative influence of maximum temperature on cassava yields, and a similar negative relationship was found between cassava yields and minimum temperatures in Lagos, during 1995–2018 [32], as well as about yields of wheat and rice in Bangladesh [43]. We can again conclude that the impacts of temperature are controversial, depending on the geographical location and probably on the actual temperature ranges within which the relationships are analyzed. Pipitpukdee et al. [30], in their analysis of cassava, also established a positive impact of temperature on yields, at least up to a certain temperature threshold, and identified a U-shaped growth curve of cassava yields in relation to temperature for Thailand. Our results indicate that in the current temperature range Eastern Africa is still in the cool stage of the cassava growth curve, and yields still show a linear positive response to warming.

The size of the harvested area negatively impacted the yield in our model. A similar relationship was found by Boansi [40] reagarding cassava production in Togo, and by Blanc [31] in sub-Saharan Africa. Contrary to that, yields were found to be positively affected by harvested area in Lagos during 1995–2018 [32], and a positive impact was found on total output, though not on yield, in 37 African countries [13]. This mixed relationship may be due to better management practices or better varieties.

The widespread decline in cassava yields, demonstrated for many countries by our research in both the observed and the modeled yield time series, indicates systemic issues affecting productivity across the region. Such patterns are consistent with regional assessments conducted by the Food and Agriculture Organization, which highlight that cassava production in many African countries continues to face constraints related to soil nutrient depletion, limited access to high-quality planting materials, and increasing climatic stressors [1], and only a few countries demonstrate near-zero or marginally positive yield growth, indicating relatively stable production systems. As the International Fund for Agricultural Development notes, regions with better institutional support and targeted investments in root-and-tuber value chains are more likely to sustain yield stability despite bio-physical and economic pressures [62]. According to our findings, the positive effects of population growth and the negative influence of the rural population ratio suggest that the need to feed larger populations generates a pressure to improve the yield of this important staple crop.

4.2. Climate Change Impacts on Harvested Areas of Cassava

In our findings the impacts of temperature and rainfall were significantly positive on the harvested area of cassava. This suggests that with increasing temperature the harvested area is expected to increase, but this is mitigated by the impact of slightly decreasing rainfall. Similarly to our findings, Adebayo [13] also established positive effects of temperature on harvested areas. Jarvis et al. [34] found that rainfall was positively related to harvested area in Uganda, Tanzania, and Nigeria, but had a negative impact in the Democratic Republic of Congo, Mozambique, Angola, Ghana, Madagascar, and Ivory Coast. Olanda Souza et al. [36] found a positive response of harvested area to rainfall in Brazil, but a negative response to temperature. Pipitpukdee et al. [30] found no significant impact of rainfall on the production area of cassava in Thailand. Amin et al. [43], in analyzing climate impacts on wheat and rice production in Bangladesh, found that neither rainfall nor temperature had any effect on wheat harvested areas, while the impacts on rice areas depended on the variety.

Although our findings point at yield and area increases over warming temperatures, other factors can considerably modify this. Expanding growing areas contribute to decreasing yields, probably because the areas brought newly to cultivation are of weaker productive capacity, and other farming resources may also become limiting. The expanding rural population seems to negatively affect cassava yields due to the spreading of smallholder subsistence farming suffering from poor soil fertility, often farming degraded lands, lacking fertilizers and quality planting materials, and having limited access to other farming resources. Therefore, even if climate itself is beneficial, other factors create severe limitations on improving yields.

These results underline that there is a strong difference across cassava producing regions in the world, and due to geographical and climatic conditions, cassava production may react differently to climate trends. These responses may also vary within shorter time windows. This underlines the importance of the present analysis for Eastern Africa, for the very long time span of 63 years, because studies performed elsewhere cannot be generalized without considering regional differences.

5. Conclusions

5.1. Conclusions and Policy Implications

Climate change is an important threat to food security, especially in Africa, where the current warm and relatively dry climate is expected to turn even warmer and dryer, which can remarkably change the growing conditions for many staple crops. Warming temperatures and decreasing precipitation may change the spreading of pests and diseases, putting further stress on food production. Therefore, it is crucial to develop agrotechnology and crop varieties that can tolerate a hotter and dryer climate. Cassava, as a drought-resistant crop resilient to climatic changes and poor soils, may require less intensive technologies and may still produce well under conditions considered unfavourable to most of the currently grown staple crops in Africa. However, the response of cassava yields to temperature and rainfall significantly differ across geographical regions; therefore, research results established in particular regions of the world cannot be directly extrapolated to other regions. This necessitates area-specific analyses, taking into account the regional differences in temperature and rainfall trends, as well as typical varieties and agronomy applied in particular regions.

Our results show for Eastern Africa the indifference of the cassava crop yield to rainfall, and a positive yield response to warming temperatures. This is, however, not uniform throughout the world and is not general even for Africa. Therefore, country-wide analyses are essential to gain proper understanding of the potentials of cassava in fighting food insecurity under warming climate scenarios.

The predominance of decreasing yield trends also suggests that increases in cassava output in several countries are being driven more by expansion of cultivated area rather than improvements in productivity. Our findings reinforce that warming temperatures may lead to the expansion of cassava areas, against other crops that are more sensitive to climatic changes. This trend echoes warnings from the African Development Bank, which stresses that reliance on area expansion rather than yield enhancement is unsustainable, particularly in regions experiencing rapid population growth and land pressure [63]. Moreover, the observed yield stagnation supports conclusions from the International Institute of Tropical Agriculture, which emphasizes the need for improved cassava varieties, enhanced soil fertility management, and climate-smart interventions to reverse declining yield trends across the continent [64].

Overall, the yield patterns observed in this analysis underscore the need for policy shifts emphasizing productivity enhancement over land expansion, with special attention paid to small-scale subsistence farmers. This recommendation is consistent with the agricultural transformation frameworks advanced by FAO, IFAD, and AfDB, all of which highlight the importance of sustainable intensification and investment in innovation to address persistent yield constraints.

5.2. Limitations and Further Research

Limitations of the present study are mainly related to the choice of data. These include the use of annual and seasonal weather data instead of monthly ones, the omission of minimum and maximum weather values, the omission of information about the exact geographic location of the analyzed countries, and the use of country-level data instead of regional data on cassava production. These can be incorporated into a more in-depth analysis in a later phase of the research. Another area of deepening the analysis is to include more specific independent variables that represent technological and socioeconomic development, such as GDP per capita, prices of cassava and substitutes, investments in cassava production, and indicators of market access. However, these may be difficult to incorporate into the analysis due to data accessibility limitations. The present model structure uses the largest possible database of 63 years and 15 countries, focusing on basic climate-related effects on cassava in Eastern Africa, a relatively under-researched region so far.