Land Suitability Assessment and Gap Analysis for Sustainable Taro (Colocasia esculenta (L.) Schott) Production in Rwanda Using Remote Sensing Data and a Fuzzy AHP Model

Highlights

What are the main findings?

• The analysis revealed that a significant portion of Rwanda land is well-suited for taro, with 22.8% classified as highly suitable and 55.7% as moderately suitable. A key finding was that within existing farmland, over 28% of the highly suitable land remains untapped for taro cultivation, indicating substantial room for expansion, particularly in the Eastern province.
• Methodologically, this study is the first to integrate multi-source remote sensing data within a Fuzzy-AHP-GIS framework for Rwanda, creating a transferable model for assessing underutilized crops to enhance food security.

What are the implications of the main findings?

• The findings offer an actionable blueprint for policymakers to direct the agricultural resources allocation and support services to high-potential districts via optimizing land use and investment.
• Closing the identified production gap through the adoption of improved practices on suitable land can transform taro from a marginal crop into a cornerstone of Rwanda climate-resilient agricultural strategy. This shift would enhance national food security, provide a global model for sustainable development for developing countries, and attract more research attention to orphan crops such as taro.

Abstract

Taro (Colocasia esculenta (L.) Schott) is a nutritionally important and climate-resilient crop with high potential for enhancing food security. Despite its significance, taro remains underutilized and excluded from major agricultural policies in Rwanda, resulting in low national yields. This gap hinders evidence-based planning and limits the crop contribution to resilience amidst population growth and climate change. By taking Rwanda as an example, a worldwide top 10 taro-producing country but still facing food insecurity issues, this study conducted a nationwide land suitability assessment to identify optimal areas for taro cultivation and quantify the production gap. The Fuzzy Analytic Hierarchy Process (AHP) model was integrated with GIS, where climatic, topographic, and a remotely sensed soil dataset were weighted and combined to generate a composite suitability index. Results revealed that 22.8% of Rwanda’s land is highly suitable (S1) and 55.7% is moderately suitable (S2) for taro cultivation. Within agricultural land, 30.2% is highly suitable, of which a significant portion (28.7%) remains largely underutilized, especially in the Eastern province. The national production gap was estimated at 32.4%, with over half of the districts exceeding 30%. The study highlights the importance of aligning taro cultivation with biophysical suitability and integrating spatial planning into national agricultural policies. The developed suitability map provides a critical decision-support tool for policymakers, agricultural planners, and extension services. By promoting sustainable taro production, improving farmer livelihoods and food security in Rwanda, it provides a global model for sustainable development for developing countries and advances research on orphan crops such as taro. The methodology offers a replicable framework for evaluating underutilized crops globally, contributing to sustainable agricultural diversification and food security.

1. Introduction

Taro (Colocasia esculenta (L.) Schott) is a nutritionally important crop and is recognized globally for helping eradicate food insecurity, especially in developing countries [1]. In Rwanda, taro (locally known as “amateke”) remains a traditional staple and grows across diverse ecological zones [2]. Rwanda is ranked among the world’s top 10 taro-producing countries [3]. However, taro is excluded from major crops policy, such as Rwanda’s Crop Intensification Program (CIP), which aims to boost the yield of staple crops, and therefore receives limited public support [4,5]. As a result, taro yield remains below 7 t/ha, while yields in Ethiopia and Kenya exceed 20 t/ha [6,7]. About 31% of the Rwanda population faces food insecurity [8]. This context reveals a critical opportunity in improving taro production via scientific land suitability mapping, which is essential for evidence-based policies, optimizing land use, and improving national food security. This study develops a land suitability framework to address this need. The resulting spatial analysis guides three priorities: maximizing yields in target zones, informing policy to better integrate taro in national programs, and promoting sustainable cultivation in optimal areas.

Taro thrives in agro-ecological zones with 25–35 °C temperatures and 1500–2000 mm annual rainfall [9,10]. It prefers fertile loamy or clayey soil with good drainage and moisture retention. Saline or poorly drained soils impede corm growth [9,11,12]. However, Rwanda lacks spatial planning tools and land suitability data. This limits informed decisions and targeted investment in taro cultivation [13]. The absence of such targeted data directly contributes to the crop’s neglect in agricultural policy and hinders efficient land allocation.

Land suitability analysis is vital for strategic planning under changing climate conditions by guiding optimal use of natural resources [14,15]. This analysis evaluates how suitable the land is for crop farming. It matches crop ecological requirements with physical land traits to evaluate cultivation potential. Selecting relevant criteria is critical, especially for taro, which needs warm temperatures, moist, well-drained soils, and favorable topography [16,17]. Due to data limitations and local variability, no universal standard exists for suitability assessments.

Studies in Ethiopia and Nigeria show that terrain distribution, soil texture, acidity, and organic matter are significant for taro and cocoyam suitability [18]. Clay-rich floodplains were found to be more favorable than acidic sandy uplands [19]. Mugiyo et al. emphasized rainfall as the dominant driver for underutilized crops like taro [20]. This highlights its importance for site-specific crops and resilient agricultural systems. Feizizadeh and Blaschke grouped key land suitability factors such as elevation, slope, pH, climate, and groundwater [21] into four main categories: topography, climate, soil, and water availability.

Multi-Criteria Decision Analysis (MCDA), by using Analytical Hierarchy Process (AHP), has been widely applied in crop suitability studies. It helps weigh environmental and soil factors [22]. For example, Sawant et al. used GIS-AHP to map rice and coconut suitability, classifying land into different suitability categories, and validated their maps with high accuracy [23]. Similar methods in South Africa assessed underutilized crops like taro, integrating climatic, soil, and socio-economic factors; the field validation and accuracy metrics confirmed the model’s reliability [20]. Incorporating expert input and local knowledge, AHP effectively supports sustainable land use, particularly for crops with specific ecological needs like taro.

While AHP enables nuanced evaluation of land suitability for crops with specific ecological needs like taro, it struggles with uncertainty in expert judgments [24,25,26]. Fuzzy-AHP addresses this by integrating fuzzy set theory. It uses linguistic variables and triangular fuzzy numbers (TFNs) to express preference uncertainty [27]. In Fuzzy-AHP, uncertainty appears during the decision-making process. The relative significance of the main criteria and sub-criteria is assessed through pairwise comparisons, using fuzzy numerical representations of linguistic terms [22,28]. This hybrid model better handles decision ambiguity and has proven effective in agricultural studies, including crop systems and agricultural assessments [29,30,31]. Although several international studies have mapped cropping suitability with these tools, their systematic application to taro across Rwanda remains limited.

Therefore, this study focuses on enhancing taro production in Rwanda through spatial land suitability analysis. The main objective is to map its suitable areas using Fuzzy AHP-GIS. This involves incorporating climatic, topographic, and soil-related variables. Secondly, the study assesses the expansion opportunities by identifying highly and moderately suitable areas beyond currently cultivated agricultural land. Finally, building on the generated suitability map, the study estimates the production gap. Here, the production gap refers to the percentage difference between the potential taro production that can be achieved by our developed suitability map and the current taro production. These objectives directly support urgent, strategic planning. They identify actionable areas suitable for taro, a climate-resilient, underused crop. The findings of this study aim to inform policymakers, providing a clear, evidence-based foundation for updating national agricultural policies, such as integrating taro into programs like the CIP. By delineating suitable land, the research offers practical guidance for policymakers and farmers alike. This supports land use decisions, closes the production gap, and strengthens agricultural resilience and food security in Rwanda. Through spatially explicit land suitability mapping, this work underpins policy reforms. These are needed to address environmental pressures, land scarcity, and policy gaps, while promoting sustainable and inclusive agricultural development.

2. Materials and Methods

2.1. Study Area

Rwanda is a landlocked country in East Africa, situated between 1.04° and 2.51°S in latitude and between 28°45′ and 31°15′E in longitude (Figure 1), covering an area of 26,338 km². It has diverse topography dominated by hills, mountains, and valleys, with altitudes ranging from 950 m in the eastern lowlands to over 4507 m in the northwestern volcanic regions. Varied topography and climatic conditions throughout the country contribute to its considerable agro-ecological diversity, which is essential for evaluating crop land suitability. Its slopes vary from gentle undulations to steep gradients, particularly in the western and northern regions. These sloped terrains greatly influence soil formation and erosion patterns.

The country experiences a temperate tropical highland climate, influenced by its elevation. Mean annual temperatures vary between 15 and 17 °C in the western region and up to 30 °C in the eastern region [32]. Rainfall ranges from 750 to 1130 mm, with two distinct rainy seasons that favor taro cultivation [33]. The short rainy season lasts from mid-September to December, while the long rainy season lasts from March to May (Figure 2).

Rwanda’s soil, shaped by parent rock, slope, and elevation, includes Ferrallisols, Acrisols, Nitisols, and fertile volcanic soils like Andosols. Alluvial valley soils support taro due to high moisture retention, while steep slopes increase erosion risk. Soil pH ranges from 3.7 to 7.4, influenced by rainfall and nutrient leaching [34]. Land use has shifted under pressure from population growth, infrastructure, and conservation policies. Cropland covers ~59% of land, followed by forests (17.7%), water bodies (6.4%), and built-up areas (15%). Agricultural land is increasingly fragmented, highlighting the need for suitability assessments for resilient crops like taro, especially within complex biophysical landscapes [35].

Taro cultivation is nationwide, spanning the country’s diverse geography from south to north and east to west. The main grown varieties include Amayanga, Bwayisi, and Amabungubungu. Figure 3 shows the Amayanga variety cultivation in Gatsibo district, Eastern province (Figure 3a) and Muhanga district, Southern province (Figure 3b).

2.2. Overall Methodological Framework

An integrated Fuzzy-AHP-GIS framework was developed to assess land suitability for taro cultivation. The methodology, illustrated in Figure 4, comprised four primary phases: (1) data collection and standardization, (2) weighting using the Fuzzy-AHP model, (3) GIS-based overlay and suitability mapping, (4) and model validation and taro production gap analysis.

2.3. Datasets and Preprocessing

Table 1. Datasets Used for Land Suitability Analysis.

Data Type		Resolution	Source
Climatic variables
	Temperature	0.1° × 0.1°	ERA5-Land monthly averaged data
	Precipitation	0.1° × 0.1°	ERA5-Land monthly averaged data
Topographic datasets
	Digital Elevation Model (DEM)	30 m	Shuttle Radar Topography Mission (SRTM) (https://earthexplorer.usgs.gov/) accessed on 30 June 2025
	Slope (derived from DEM)	30 m
Soil physical property datasets
	Texture	250 m	ISRIC—World Soil Information—Africa SoilGrids(https://data.isric.org/geonetwork/srv/eng/catalog.search#/home) accessed on 30 June 2025
	Drainage	250 m
	Bulk density	250 m
	Soil moisture (SM)	1 km
Soil chemical property datasets
	Cation Exchange Capacity (CEC)	250 m	ISRIC—World Soil Information—Africa SoilGrids(https://data.isric.org/geonetwork/srv/eng/catalog.search#/home) accessed on 30 June 2025
	pH	250 m
	TN	250 m
	Extractable P	250 m
	Extractable K	250 m
	SOC	250 m
Accuracy evaluation
	NDVI	250 m	MODIS/006/MOD13Q1
	LULC	10 m	Sentinel-2
Productivity	Taro yield and planting area	At district level	National Institute of statistics of Rwanda

presents data obtained from various reliable sources to evaluate land suitability for taro cultivation in Rwanda. The primary datasets included climate, soil, and topographic datasets. Land use/land cover (LULC) for the land mask of non-cropland and Normalized Difference Vegetation Index (NDVI) were used as a production proxy. All datasets were reprojected, resampled, and clipped to a consistent spatial framework at the resolution of 250 m.

ERA5-Land monthly average data from 2001 to 2024, provided temperature and annual total precipitation information essential for assessing thermal and moisture conditions relevant to taro climatic sensitivity [36]. Elevation and slope data were derived from SRTM. Elevation influences temperature, rainfall, and soil formation, while slope affects erosion risk and drainage, both key to taro growth (Figure 5).

Africa SoilGrids soil datasets were collected from ISRIC—World Soil Information [37]. Soil chemical parameters critical for plant growth, including CEC, pH, SOC, TN, P, and K, were included. The spatial distribution and reclassified suitability scores for these soil chemical properties are presented in Figure 6.

Soil physical parameters influence root development and water management. These include texture, drainage, bulk density, and soil moisture, defined as plant available water holding capacity (v%) of the soil fine earth fraction, with field capacity defined at h = 200 cm or pF 2.3, aggregated over rootable depth and the top 30 cm, mapped at 1 km resolution.

The spatial distributions of reclassified suitability scores for these soil physical properties are presented in Figure 7.

Finally, land use land cover (LULC) and NDVI data from a recent five-year period (2020–2024) were incorporated to focus the analysis on areas under agricultural use, identify zones for future expansion, and serve as a validation proxy [38]. The LULC data were derived from 10 m resolution Sentinel-2 satellite imagery, providing detailed annual land cover classifications [https://livingatlas.arcgis.com/landcover/, accessed on 1 July 2025]. To represent crop vigor, we computed a mean NDVI value for a composite growing window (November to May) using Google Earth Engine. This period was strategically selected to capture the months of peak photosynthetic activity by excluding the early establishment phase (September–October) and the water-stressed dry season (June–August). Consequently, this period captures the sustained, vigorous growth phase of taro in Rwanda, accounting for varietal differences in peak growth timing. The NDVI data were obtained from the MODIS/006/MOD13Q1 product, accessed on 1 July 2025.

2.4. Criteria Identification and Standardization

2.4.1. Criteria Identification

Identifying criteria was a critical first step. We used a systematic process to ensure all factors were ecologically relevant and scientifically justified for taro cultivation. This involved synthesizing knowledge from three key sources: the crop’s well-documented physiological needs, a comprehensive review of the scientific literature, and established frameworks from previous land suitability studies. The relevant factors were grouped into four categories: (i) climate (temperature, rainfall); (ii) topography (elevation, slope); (iii) soil physical properties (texture, drainage, bulk density, rootzone soil moisture); and (iv) soil chemical properties (pH, CEC, N, P, K, SOC).

The selection of these specific criteria was based on the optimal growth requirements for taro, as established in the literature. Taro requires an optimal temperature of 25 °C to 30 °C, making temperature a vital factor in assessing its thermal suitability [17]. A minimum of 1000 mm of annual rainfall is necessary to meet the crop’s water requirements, with the optimal requirement of 1500 to 2000 mm highlighting the importance of rainfall in the suitability analysis [39]. Drainage and soil texture are also key considerations [40], as taro thrives in soils with good moisture retention and moderate drainage, particularly in fertile, loamy soils. Soil pH is another important factor, as taro prefers slightly acidic to neutral soils (5.0 to 7.0) for optimal growth [9]. Elevation plays a significant role, with the crop being best suited for areas ranging from sea level to 1500 m above sea level, where temperature and moisture conditions are ideal [41]. Slope is also considered, as steep terrains may hinder mechanized farming and increase the risk of erosion. Soil drainage is important, as it allows for better root development, supporting healthy taro growth [17].

2.4.2. Criteria Classification

Once the key criteria were identified, the next step involved data classification. Each dataset was classified based on its suitability for taro cultivation. For example, soil types and drainage were classified into suitability categories based on their fertility or drainage potential. Each raster factor was then classified into different suitability classes based on predefined score thresholds, as shown in Figure 5, Figure 6 and Figure 7.

2.5. Weights Determination by Fuzzy-AHP

A Multi-Criteria Decision Analysis using Fuzzy-AHP was employed to integrate climatic, topographic, and soil-related factors affecting taro cultivation in Rwanda [42].

2.5.1. Fuzzy Pairwise Comparison Matrix

Fuzzy set theory, introduced by Zadeh [27], enables the modeling of linguistic vagueness through partial membership values [25]. In multi-criteria decision-making contexts, TFNs are commonly applied to represent subjective preferences and handle imprecision in decision criteria [22,25]. The comparison was determined primarily from evidence synthesized in prior research and literature reviews. The criteria were then evaluated using the linguistic terms shown in Supplementary Table S1, and the resulting pairwise comparison matrices were systematically converted into TFNs, as detailed in

Table 2. Fuzzy-AHP Main Criteria Pairwise Comparison Matrix.

	Climatic	Topographic	Soil Physical	Soil Chemical	Fuzzy Weights
Climatic	(1,1,1)	(1,1,1)	(1,2,3)	(1,2,3)	(0.198, 0.341, 0.519)
Topographic	(1,1,1)	(1,1,1)	(1,1,1)	(1,2,3)	(0.198, 0.287, 0.394)
Soil Physical	(1/3,1/2,1)	(1,1,1)	(1,1,1)	(1,1,1)	(0.151, 0.203, 0.300)
Soil Chemical	(1/3,1/2,1)	(1/3,1/2,1)	(1,1,1)	(1,1,1)	(0.114, 0.170, 0.300)
λ_max = 4.0606, n = 4, CI = 0.0202, RI = 0.90, CR = 0.0225

and Supplementary Table S3. A TFN is defined as

$$Ã\left({\alpha }_{ij}\right)=\left[{\displaystyle \genfrac{}{}{0pt}{}{\genfrac{}{}{0pt}{}{(\mathrm{1,1},1)}{\left(l_1,m_1,u_1\right)}}{\genfrac{}{}{0pt}{}{⋮}{\left(l_{n1},m_{n1},u_{n1}\right)}}} {\displaystyle \genfrac{}{}{0pt}{}{\genfrac{}{}{0pt}{}{{(l}_{12},m_{12},u_{12})}{\left(\mathrm{1,1},1\right)}}{\genfrac{}{}{0pt}{}{⋮}{\left(l_{n2},m_{n2},u_{n2}\right)}}} {\displaystyle \genfrac{}{}{0pt}{}{\genfrac{}{}{0pt}{}{\dots }{\dots }}{\genfrac{}{}{0pt}{}{\ddots }{\dots }}} {\displaystyle \genfrac{}{}{0pt}{}{\genfrac{}{}{0pt}{}{{(l}_{n1},m_{n1},u_{n1})}{\left(l_{n2},m_{n2},u_{n2}\right)}}{\genfrac{}{}{0pt}{}{⋮}{\left(\mathrm{1,1},1\right)}}}\right]$$

(1)

where ${\alpha }_{ij}=\left(l_{ij},m_{ij},u_{ij}\right) \mathit{and} {\alpha ij}^{-1}=\left({\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$u_{ij}$}\right.}{u}_{ij},{\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$m_{ij}$}\right.},{\displaystyle \raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$l_{ij}$}\right.}\right)$

For $i,j=1,\dots ,n$ and $i\ne j$, $l$, $m$ and $u$ stands for lower, middle and upper bound of fuzzy member $\mathrm{Ã}$ (Table S1).

2.5.2. Consistency Index (CI) and Consistency Ratio (CR)

Prior to computing the consistency index, the fuzzy matrix was defuzzified into a crisp matrix using the following simple mean method:

$${crisp}_{ij}=m_{ij}$$

(2)

Then, following Saaty [43], the consistency index can be calculated as

$$CI={\displaystyle \frac{{\lambda }_{max}-n}{n-1}}$$

(3)

where ${\lambda }_{max}$ is the maximum eigenvalue of the crisp matrix. In a perfectly consistent matrix, the condition ${\lambda }_{max}=n$ holds. This index quantifies how far the matrix is from perfect consistency. To further interpret this inconsistency in relation to random judgments, the Consistency Ratio (CR) is calculated as

$$CR={\displaystyle \frac{CI}{RI}}$$

(4)

where RI is the Random Index, an average consistency index obtained from randomly generated reciprocal matrices of order n, as provided by Saaty (Supplementary Table S2). CR was computed to verify judgment reliability, with CR < 0.1 confirming acceptable consistency.

2.5.3. Computing Fuzzy Weights

The geometric mean method [44] was used to calculate the fuzzy weights ${\mathfrak{w}}_i$ for each criterion:

$${ř}_i={\left(\prod _{i = j}^n{ã}_{ij}\right)}^{1/n}i=\mathrm{1,2},3\dots ,n$$

(5)

$${\mathfrak{w}}_i={ř}_{i\otimes }{\left(\sum _{k=1}^nřk\right)}^{-1}$$

(6)

where ${\mathrm{ã}}_{ij}$ is the fuzzy comparison value between criterion $i$ and $j$. ${\mathrm{ř}}_i$ is a triangular fuzzy number calculated via fuzzy multiplication and fuzzy power (raising to $1/n$). Once all fuzzy geometric means ${\mathrm{ř}}_i$ are computed, the fuzzy weights ${\mathfrak{w}}_i$ are derived by normalizing. ⊗ denotes fuzzy multiplication. The inverse ${\left(\sum _{k=1}^n\mathrm{ř}k\right)}^{-1}$ is the multiplicative inverse of the fuzzy sum. Each ${\mathfrak{w}}_i$ is a normalized triangular fuzzy number representing the relative weight of criterion $i$. The fuzzy weights obtained were represented as TFN and needed to be clarified and transformed into precise numerical values. This defuzzification was carried out using the center of area (COA) method, a widely used technique for this purpose [24]. The optimal best non-fuzzy performance (BNP) value for the fuzzy weight ${\mathfrak{w}}_i$ was then computed as follows:

$${BNP}_i= {lw}_i + {\displaystyle \frac{\left({uw}_i-{lw}_i\right)+\left({mw}_i-{lw}_i\right)}{3}}, {\forall }_{\mathit{i}}$$

(7)

where

${lw}_i$ = lower bound of the fuzzy weight $w_i$.

${mw}_i$ = middle (modal) value of the fuzzy weight.

${uw}_i$ = upper bound of the fuzzy weight.

The final normalized weights $w_i$ were derived by dividing each $BNPi$ by the sum of all $BNPi$ values. Table 2 and

Table 3. Global Weights derived from Fuzzy-AHP.

Main Criteria	Normalized Weights	Sub-Criteria	Normalized Weights	Combined Weight
Climatic factors	0.3351	Temperature	0.500	0.1675
		Precipitation	0.500	0.1675
Topographic factors	0.2792	Slope	0.500	0.1396
		Elevation	0.500	0.1396
Soil Physical factors	0.2049	Texture	0.341	0.0699
		Drainage	0.341	0.0699
		Soil moisture	0.203	0.0415
		Bulk Density	0.115	0.0236
Soil Chemical factors	0.1808	CEC	0.262	0.0473
		pH	0.262	0.0473
		SOC	0.153	0.0276
		TN	0.153	0.0276
		Potassium	0.085	0.0154
		Phosphorus	0.085	0.0154

, respectively, present main criteria and global weights pairwise comparison matrix.

2.6. GIS Integration and Land Suitability Index Development

Fuzzy-AHP-derived weights were applied to standardized input layers in a GIS-MCDA framework to generate a composite suitability index for taro cultivation.

The composite index was calculated by aggregating the weighted criterion values, resulting in a spatial suitability map, following this formula:

$$S=\sum _{i=1}^n{W}_i{X}_i$$

(8)

where $S$ is a final suitability map, $n$ is the number of evaluated criteria, $W_i$ is the weight of a factor $i$, and $X_i$ is the standardized values of factor $i$.

Following FAO guidelines, land was classified into four categories: highly suitable (S1), moderately suitable (S2), marginally suitable (S3), and not suitable (N1) [45]. In the next phase, standardization and transformation of the datasets were conducted to ensure comparability across the different criteria. Standardization involved transforming each factor into a common scale ranging from 0 to 10, allowing them to be combined into a single composite suitability index [46]. For instance, each factor was reclassified into suitability scores as optimum, good, poor and not good with 10, 7.5, 5 and 2.5 scores, respectively (Figure 5, Figure 6 and Figure 7).

2.7. Sensitivity Analysis and Validation

2.7.1. Sensitivity Analysis

A one-at-a-time (OAT) sensitivity analysis was used to assess how varying criterion weights (±20% to ±100%) affected spatial suitability, while keeping total weights normalized [28,47]. The adjusted weights were calculated as

$${\mathcal{m}\mathcal{w}}_j\left(c\right)=\left(1+c\right)\times {\mathcal{w}}_j$$

(9)

where ${\mathcal{m}\mathcal{w}}_j\left(c\right)$ is the modified weight of criterion $j$, $c$ is the change rate (±20% to ±100%), and ${\mathcal{w}}_j$ is the original weight of the j^-th criteria from Fuzzy-AHP. The weights of other criteria were recalculated to preserve normalization:

$${\mathcal{m}\mathcal{w}}_i\left(c\right)={\mathcal{w}}_i\times {\displaystyle \frac{1-{\mathcal{m}\mathcal{w}}_j}{1-{\mathcal{w}}_j}}$$

(10)

${\mathcal{m}\mathcal{w}}_i\left(c\right)$ represents the adjusted weight corresponding to ${\mathcal{m}\mathcal{w}}_j$, where $i\ne j$; ${\mathcal{w}}_i$ is the initial weight of the $i$^-th criterion derived from the Fuzzy-AHP.

For each simulation, the Land Suitability Index (LSI) was recomputed as

$$R\left({\mathcal{w}}_{j,c}\right)={\mathcal{m}\mathcal{w}}_j\times {\mathcal{x}}_j+\sum _{i\ne j}^{n}m{\mathcal{w}}_i\times {\mathcal{x}}_i$$

(11)

where $R\left({\mathcal{w}}_{j,c}\right)$ is simulated LSI ${\mathcal{x}}_i$ and ${\mathcal{x}}_j$ are the scores of criteria $i$ and $j$, respectively, in Table S2.

The change rate of LSI for each pixel (${\Delta E}_k$) was derived as

$${\Delta E}_k\left({\mathcal{w}}_{j,c}\right)={\displaystyle \frac{R_k\left({\mathcal{w}}_{j,c}\right)-R_0}{R_0}}\times 100\%$$

(12)

where ${\Delta E}_k\left({\mathcal{w}}_{j,c}\right)$ represents the rate of change in the LSI result for the $k$^-th pixel in response to a change rate in ${\mathcal{m}\mathcal{w}}_j$. $R_k\left({\mathcal{w}}_{j,c}\right)$ denotes the simulated land suitability results for the $k$-th pixel, computed using Equation (12), while $R_0$ is the original LSI result obtained from Equation (8). The index $k$ refers to the sequence of randomly selected pixels.

Overall, the Mean Absolute Change Rate (MACR) of LSI was then calculated to quantify sensitivity as follows:

$$MACR\left({\mathcal{w}}_{j,c}\right)=\sum _{k=1}^N{\displaystyle \frac{1}{N}}\times \left|{\displaystyle \frac{R_k\left({\mathcal{w}}_{j,c}\right)-R_0}{R_0}}\right|\times 100\%$$

(13)

where $MACR\left({\mathcal{w}}_{j,c}\right)$ represents the MACR using ${\mathcal{m}\mathcal{w}}_j$ as the rate of change, and N denotes the total number of pixels.

2.7.2. Model Validation

The model was validated in two ways. First, NDVI acted as a proxy for crop productivity, commonly used in data-scarce regions [38]. The Kruskal–Wallis H test, a statistical method for comparing groups, assessed whether NDVI distributions differed significantly across suitability classes, with Mann–Whitney U tests, another statistical tool, for pairwise comparisons. Next, the final map was overlaid with Land Use/Land Cover (LULC) data for visual validation. This tested the logical expectation that suitable land should largely coincide with existing agricultural areas, thereby grounding the results in Rwanda’s actual landscape. Taken together, these methods provided complementary biophysical and contextual validation of the model outputs.

2.8. Estimation of Production Gap

Yield production estimation followed the FAO land productivity index concept, which assesses the physical productivity of land relative to the best land [48]. The yield potential of S1 land is assumed to vary between 80 and 100% (0.8–1.0 relative index), with a 20% reduction for each subsequent suitability class. For instance, 1.0, 0.8, 0.6, and 0.4 are assigned as relative productivity factors for S1, S2, S3, and N1, respectively.

To define attainable yield, we used the 90th percentile of observed district-level yields, calculated via linear interpolation between the top-ranked values. This method follows Van Ittersum et al. [49], who recommend using high-percentile yields to represent well-managed conditions, and applies the interpolation approach of Hyndman and Fan [50].

In this study, highly and moderately suitable areas were considered as suitable land for taro. In a situation where S1 land in a certain district was larger than the current land under taro farming, only S1 land was used in production gap estimation. However, where S1 was smaller than the current taro land, we first estimated the potential yield of available S1, then added the yield potential of the remaining area as S2’s relative yield potential. For instance, potential yield production ($Yp$) was computed following this formula:

$$\mathrm{Y}\mathrm{p}=\left\{\begin{array}{c}A\mathrm{c}\mathrm{u}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{t}\times Y\mathrm{a}\mathrm{t}\mathrm{t}\times 1.0\hfill \\ \left(A\mathrm{S}1\times Y\mathrm{a}\mathrm{t}\mathrm{t}\right)+\left(\mathrm{m}\mathrm{i}\mathrm{n}\left(A\mathrm{c}\mathrm{u}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{t}-A\mathrm{S}1,A\mathrm{S}2\right)\times 0.8\times Y\mathrm{a}\mathrm{t}\mathrm{t}\right) \end{array} \begin{array}{c}\mathrm{i}\mathrm{f} A\mathrm{c}\mathrm{u}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{t}<A\mathrm{S}1\\ \mathrm{i}\mathrm{f} A\mathrm{c}\mathrm{u}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{t}>A\mathrm{S}1\end{array}\right.$$

(14)

where $\mathrm{Y}\mathrm{p}$ is the estimated potential yield production; $A\mathrm{c}\mathrm{u}\mathrm{r}\mathrm{r}\mathrm{e}\mathrm{n}\mathrm{t}$ is the current area under taro cultivation; $Y\mathrm{a}\mathrm{t}\mathrm{t}$ is the attainable yield; $AS1$ and $\mathrm{A}\mathrm{S}2$ are, respectively, highly suitable and moderately suitable areas from developed taro land suitability; and $1.0$ and $0.8$ are, respectively, the yield factors applied to the S1 and S2 area, reflecting their relative productivity as defined by FAO.

The current yield production was obtained from annual agriculture statistics of Rwanda published by National Institute of Statistics of Rwanda (NISR) [51]. The production gap was then calculated as a percentage of the potential production, using the following formula:

$$YpG=\left({\displaystyle \frac{Yp-ActYp}{Yp}}\right)\times 100$$

(15)

where $YpG$ is the taro production gap (%); $Yp$ is the estimated taro production; and $ActYp$ is the Actual taro production at the district level. This approach helps quantify the extent of unrealized production potential and identifies areas where productivity improvements could be most impactful.

3. Results

3.1. Spatial Distribution of Variables and Their Influence on the Land Suitability Index

Table 4. Variables distribution base on their suitability reclassification (optimum to not good) range.

Variable	Range	Area (km2)	(%)	Variables	Range	Area (km2)	(%)
Temperature	25–30	15,436	34.3	Drainage	Well drained	22,366	93.35
(°C)	22–25	4967	11.04	(classes)	Moderate	783	3.27
	20–22	4747	10.55		Imperfect and poor	763	3.18
	<20	201	0.45		Very poor	47.81	0.2
Precipitation	1500–2100	3453	7.67	pH	≥5.5	10,202	42.58
(mm)	1000–1500	18,987	42.19		5.0–5.5	12,910	53.88
	900–1000	2193	4.87		4.5–5.0	847	3.54
	<900/>2100	309	0.69		3.7–4.5	0.0	0.0
Elevation	≤1500	9781	38.59	TN	≥5	143	0.6
(m)	1500–2000	10,906	43.03	(g/kg)	2–5	8981	37.48
	2000–2500	3940	15.54		1–2	10,581	44.16
	>2500	720	2.84		0.24–1	4256	17.76
Slope	≤3	4393	17.35	CEC	≥30	1235	5.16
(degrees)	3–8	6436	25.41	(cmol(+)/kg)	15–30	13,384	55.86
	8–15	6183	24.41		8–15	9335	38.96
	>15	8313	32.83		4–8	5	0.02
SOC	≥50	1410	5.89	P	>50	76	0.32
(g/kg)	20–50	12,414	51.81	(mg/kg)	20–50	5296	22.02
	10–20	9387	39.18		10–20	13,462	55.97
	<10	749	3.12		<10	5218	21.69
Bulk density	≤1250	5657	26.61	K	≥300	1587	6.6
(kg/m3)	1250–1400	15,899	66.36	(mg/kg)	150–300	20,488	85.18
	1400–1500	2224	9.28		100–150	1976	8.22
	>1500	180	0.75		<100	2	0.01
Soil texture	Loam, Silt loam, and Clay loam	10,915	45.56	SM	≥12	580	2.38
(classes)	Sand clay loam, Silt clay loam	4591	19.16	(v%)	9–12	15,337	62.93
	Sandy clay and Clay	8453	35.28		7–9	8441	34.64
	Light clay	0.0	0.0		6–7	13	0.05

shows the area and percentage distribution of sub-criteria ranges across the study area. Figure 5, Figure 6 and Figure 7 illustrate their spatial distribution based on their assigned scoring threshold (10 for optimum to 2.5 for not good). The optimum temperature range (25–30 °C) covered 34.3% of the area, located primarily in the eastern part of Rwanda (Figure 5a). Precipitation in the range of 1000–1500 mm occurred over 42.2% of the area, mainly in the north-western parts (Figure 5b). Topography was predominantly between 1500 and 2000 m in elevation (43%), with higher elevations mainly in the north-western part (Figure 5c). Gentle slopes (<15°) occupied 32.8% of the area, mostly distributed in the northeastern part (Figure 5d).

Clay loam (44%) was primarily located in the middle-southern part, while clay soil (28%) was found in the northern part (Figure 7a), together representing the most common soil texture classes. Soil bulk density was in the range of 1250–1400 kg/m³ for 66.4% of the area (Figure 6c). Organic carbon was in the range of 20–50 g/kg for 51.8% of the area and was present in all regions (Figure 6a). Soil pH was between 5.0 and 7.4 for 96% of the area (Figure 6b). The area classified with the highest score for individual soil chemical properties (e.g., TN, CEC, P, K) was below 8% for each; the majority of the area for each property was classified within the moderate scoring range (Table 4). Drainage was classified as well-drained in 93.4% of the area (Figure 7b). Root zone soil moisture was in the 9–12 v% range for 62.9% of the area (Figure 7d).

3.2. Accuracy Assessment of Land Suitability and Composite Taro Suitability Map

The one-at-a-time sensitivity analysis showed that slope had the highest Mean Absolute Change Rate (MACR) at 9.14%, followed by temperature and precipitation. Soil texture, drainage, and elevation showed intermediate MACR values, while phosphorus, potassium, and soil moisture had the lowest values (Figure 8). The MACR values for all sub-criteria remained below 10%, with an average of 2.14%.

Results from both the Kruskal–Wallis H test for the relationship between land suitability classes and the Mann–Whitney U test to test pairwise comparisons were statistically significant (p < 0.001) and showed that S1 areas had higher NDVI values than S2 and S3 classes (Figure 9). After overlaying the S1 class with LULC, our study showed that 43.68% of S1 falls within croplands and 35.82% within rangelands, and 17.7% and 3.03% overlapped with forest and flooded vegetation, respectively.

A composite land suitability map for taro cultivation was developed using 14 sub-criteria. The results show that 22.8% of the area is S1, 55.7% S2, 0.1% S3, and 21.4% N (Figure 10). S1 zones are mainly concentrated in the eastern and southern provinces. S2 areas are widespread across the region, while S3 and N zones are mainly located in central and edge areas, which are characterized by steep terrain.

3.3. Analysis of Taro Land Suitability at the Sub-National Level

The previous section presented the spatial distribution of land suitability classes across the entire land area. Building on this, the suitability classes were next examined within existing cropland to assess how this potential aligns with current agricultural activity. This focused comparison clarifies the spatial relationship between land suitability and contemporary land use for taro cultivation.

By focusing on current agricultural land, the suitability distribution was 69.7% S2, 30.2% S1, and 0.11% S3 (Figure 11). Notably, the Eastern province had the largest positive differences between the area of highly suitable (S1) land and the area currently cultivated with taro. The districts with the largest such differences were Gatsibo (representing 10.1% of national S1 agricultural land), Kayonza (9.4%), Ngoma (9.2%), Nyagatare (7.6%), and Kirehe (7.3%) (Figure 12). Together, these districts account for 13.1% of Rwanda’s agricultural land. For example, Gatsibo district cultivates 479 ha of taro and contains over 45,000 ha of S1 land (Supplementary Table S4). Other districts in the Southern, Western, and Northern provinces also showed positive differences. In contrast, Ngororero and Rutsiro districts showed a negative difference, with the cultivated area exceeding the area of high-suitability land by 117 ha and 989 ha, respectively.

3.4. Estimation of Taro Production Gap at the Sub-National Level

The estimated taro production gap varied widely across Rwanda’s districts. The gap ranged from 4% to 62.5% (Supplementary Table S5). All 30 districts had a production gap greater than 10%, with 17 districts (56.7%) showing gaps exceeding 30%.

Spatial analysis revealed distinct geographical patterns in the production gap distribution (Figure 13). The highest gaps (>50%) were concentrated in the Western province, particularly in Nyabihu district. Moderate to high gaps (30–50%) were widespread across districts in the Southern and Eastern provinces, including Bugesera, Rwamagana, and Nyagatare. The lowest gaps (4–20%) were primarily found in districts scattered throughout the Northern and parts of the Western province.

The top ten districts with high production gaps ranged from 38.5% to 62.5% (Figure 14). Nyabihu district showed the highest gap at 62.5%, followed by Bugesera (56.4%) and Rwamagana (51.9%). Other districts in the top ten included Nyagatare (49.1%), Burera (49.0%), Gicumbi (44.4%), Nyarugenge (43.4%), Rulindo (39.5%), Ruhango (38.8%), and Gisagara (38.5%).

4. Discussions

4.1. The Uniqueness of This Study

This study provides the first comprehensive, national-scale spatial assessment of land suitability for taro in Rwanda. It addresses a critical paradox: although Rwanda ranks among the top 10 global taro producer countries [3], 31% of its population faces food insecurity [8]. Despite this, taro is systematically excluded from key agricultural policies like the Crop Intensification Program [4,5]. Our study’s evidence-based land suitability analysis reposes taro from a marginalized “orphan crop” into a strategic asset for enhancing food security and climate resilience. The methodological innovation of integrating Fuzzy AHP with GIS represents a significant advancement for evaluating underutilized crops. By relying on globally available remote sensing data, this framework provides a replicable and cost-effective blueprint that is highly portable to other data-scarce regions [24,25,26]. This makes it a powerful tool for assessing a wide range of crops, directly contributing to strategic agricultural diversification and enhanced global food security.

4.2. Why Is Taro Grown Primarily on Marginal Land Instead of Highly Suitable Prime Land?

4.2.1. From the Policy Perspective

Over the past two decades, Rwanda’s agricultural policies have shifted toward improving smallholder profitability, commercialization, and food security through “Imihigo” performance contracts [52]. Imihigo refers to the multilevel, top-down performance framework in which national priorities cascade through all administrative levels down to individual farmers, where each part is required to meet strict agricultural targets aligned with national priorities. The Crop Intensification Program, launched in 2007 and scaled nationally in 2009, promotes land consolidation, high-value crops, and subsidized inputs such as improved seeds and fertilizer. Currently, CIP prioritizes eight crops, namely maize, rice, beans, Irish potato, wheat, soybeans, cassava, and vegetables [53].

Input subsidies reinforce this hierarchy. Fertilizer and seed subsidies now follow a targeted approach that directs resources toward agro-ecological zones, farmer categories, and commodities with strong market potential. In 2022/23, subsidy spending represented 64% of district-level agricultural budgets, illustrating the scale of policy influence [54]. District governments implement these subsidies via approved agrodealers, further institutionalizing crop prioritization. As resources and technical support are concentrated on a small number of priority crops, farmers have strong incentives to use prime land for those crops. As a result, crops not favored by policy, such as taro, despite its cultural and nutritional importance are pushed onto marginal lands that receive less support and fewer market opportunities.

Under the Imihigo contract framework, farmers can face penalties for non-compliance. Consequently, poor performance on Imihigo indicators leads to negative evaluations for responsible officials. This creates strong incentives to grow government-prioritized crops to avoid losing access to subsidies and other benefits. As a result, even if taro is biophysically suitable for prime land, the risk of crop marginalization leads farmers to opt for government-supported crops.

4.2.2. From the Economic Perspective

Economic profitability is a significant determinant of crop choice. Farmers are keenly aware of the financial implications of different crops, particularly regarding input costs and market access. Subsidies play a crucial role in making priority crops more economically viable. For instance, prioritized crops benefit from government input subsidies (e.g., fertilizers and improved seeds), which reduce financial barriers and enhance profitability. Despite the agronomic suitability of taro for prime land, farmers often allocate their most productive land to subsidized crops, as this offers better returns on investment. The disparity in subsidies and technical support between prioritized and non-prioritized crops like taro reinforces this trend. Recent profitability analyses show that while smallholder farmers often achieve positive economic margins, the commercialization potential of their crops is primarily shaped by market access and government support [55]. Farmers growing priority crops that benefit from subsidies and technical support experience higher market integration and reduced production costs. This further reinforce crop choices, as these crops have both local and export market potential.

In contrast, marginal crops like taro, typically cultivated for subsistence, lack the necessary support and market access to be commercially viable. As a result, taro is often relegated to marginal lands, where lower productivity and limited market opportunities constrain its profitability (Figure 15a). Consequently, it has become the most widely consumed crop at the household level, as farmers primarily grow it for their own consumption rather than for the market (Figure 15b). This policy-driven allocation of land creates socio-economic divides, as farmers growing government-supported crops experience improved incomes, whereas those cultivating traditional, non-prioritized crops like taro face reduced commercial opportunities. Consequently, the emphasis on market-oriented crops not only limits crop diversity but also contributes to rural disparities, with some farmers benefiting from government incentives while others are left behind.

4.3. Comparison with Other Studies

4.3.1. Taro’s Persistent Status as an Orphan Crop

The substantial production gaps in several districts highlight persistent challenges within Rwanda’s taro production systems. These gaps reflect taro’s persistent status as an “orphan crop” [56,57], largely excluded from mainstream agricultural research and policy support despite its nutritional value and climate resilience. The spatial variability in these gaps aligns with findings from across sub-Saharan Africa, where similar challenges in root and tuber systems are linked to suboptimal farming practices and resource constraints [58,59]. Critically, the concentration of the largest gaps in districts with abundant suitable land suggests that biophysical potential is a necessary but insufficient condition for high productivity. This pattern underscores that realizing this potential is contingent on corresponding investments in extension services, improved inputs, and market development. Therefore, our findings reinforce the critical need to integrate geospatial land suitability assessments with targeted agricultural interventions, particularly in high-potential districts where closing these gaps could substantially enhance national food security [13].

4.3.2. Underutilized Suitable Land Alongside Cultivation on Less Suitable Margins

The concentration of highly suitable land (S1) in the eastern regions, combined with its current underutilization, identifies a clear priority for agricultural resource allocation. These underutilized S1 areas also occur in forest and flooded-vegetation zones, creating a dual challenge of identifying feasible expansion areas while respecting ecological and policy constraints. In addition, the pervasiveness of substantial production gaps in districts like Nyabihu and Bugesera strongly suggests that institutional and technical factors, rather than biophysical limitations, are the primary constraints to taro production. This inference is supported by the concurrence of high suitability and low output, a well-documented challenge for orphan crops in sub-Saharan Africa. Such challenges are typically attributed to limited policy support, inadequate extension services, and poor market access [60]. Building on this, our study provides empirical, spatially explicit evidence for this pattern in Rwanda, quantifying the extent of underutilized suitable land. Moreover, the similarity of our findings to a land suitability study in Hubei province, China [61] which also identified underutilized suitable land alongside cultivation on less suitable margins indicates that this may be a common inefficiency in root and tuber crop systems across diverse contexts. Together, this recurrence reinforces the critical need to integrate geospatial suitability assessments into agricultural planning to guide targeted investments and avoid such resource mismatches.

4.3.3. Sensitivity

Spatial patterns revealed by our analysis provide actionable insights for addressing Rwanda’s food security challenges. Building on these findings, our sensitivity analysis identified slope, temperature, and precipitation as the most influential factors. These results are consistent with, and spatially contextualizes, the body of agronomic research highlighting the importance of these specific criteria for crop suitability assessment in general [14,62,63], and for taro in particular. For instance, our identification of optimal temperature and precipitation ranges directly corroborates the physiological thresholds established by Zhang et al. [17] and the high water requirements demonstrated by Juang et al. [64] and other studies [42,43]. Most significantly, our result that slope was the most critical factor (9.14% MACR) quantitatively extends the findings of Talwana et.al, which revealed that topography had a clear impact on taro yield, with low-lying areas producing double the yield of uplands [65]. While soil fertility management is known to enhance taro growth and yield [66,67], it is noteworthy that its parameters appeared as less influential factors in our macro-scale spatial model. This contrast highlights a key insight from our study: at the national planning scale, terrain and climate are the primary screening factors for taro suitability, while soil chemistry becomes a secondary consideration for management within suitable zones.

The regional disparities in suitability and production gaps could be attributed to a combination of biophysical and socio-economic factors. The model identified slope as the most sensitive criterion. Slope is known to have a pronounced physical impact on erosion risk. This could be the contributing reason to the high production gap in the western highlands, such as Nyabihu. This finding is consistent with previous research underscoring topography as a decisive factor for taro cultivation [65]. Conversely, the underutilization of highly suitable land in the Eastern province, which possesses optimal temperatures and gentle slopes, indicates that non-biophysical factors are likely the dominant constraint. This pattern aligns with the established literature on orphan crops, which identifies limited market access, entrenched farming traditions, and a lack of crop-specific extension services as critical barriers to utilizing their biophysical potential [60,62,63].

4.4. Uncertainties

Based on the previous discussions, the current biophysically suitable area for taro is likely to be occupied by prioritized crops from the CIP. We could not be specific to directly point to a particular crop (e.g., maize, cassava, or sweet potatoes…) as crops can be seasonally interchanged through crop rotation, and all these go with the previously discussed Imihigo policy. However, we can add a critical point which might be the concern of our current study innovation. Indeed, improving the biophysical suitability assessment would necessitate a geospatial analysis of land suitability by crop type. Unfortunately, except for the ongoing “Crop Type Mapping for Rwanda project”, crop-specific suitability mapping studies in Rwanda remain limited, which makes our study a part of innovation. The ongoing remote sensing crop type mapping project focuses on selected crops and districts, and its spatial suitability distribution maps have not yet been released [68]. Referring to all of the points discussed above, we could say that crop prioritization or marginalization in Rwanda is defined by government policy through the CIP. Initially, the program started by prioritizing six food crops [69] and progressively added others later [53]. From this concept, we believe that other crops currently not on the list such as taro, can also be candidates for this CIP inclusion opportunity.

Our study relied on the NDVI and LULC as validation tools and provided a robust framework for taro suitability assessment. NDVI has been widely used to predict crop yield and other crop growth parameters. For example, an 11-year study conducted in 48 U.S. states found that using high-resolution NDVI with cropland masks significantly improved yield estimates [70], and it has also been effectively used to validate various crops, including root and tuber crop yield predictions [38,71,72]. However, we acknowledge the limitations of the use of NDVI as a production proxy. As a measure of canopy greenness and above-ground biomass, it serves as an excellent but indirect proxy for the ultimate yield of taro corms, an underground organ. Furthermore, while this study quantifies the production gap from biophysical factors, a full assessment of taro’s potential impact on its national food security is critical. This would require a comparative analysis of the nutritional value versus the crops it could replace or interchanged with on suitable land, which remains an important area for future research.

4.5. Strategy Recommendations

Therefore, to build on this work, we recommend a multi-faceted strategy for future research. First, targeted field sampling within the identified high-suitability zones is essential to calibrate the relationship between our suitability index and the actual corm yield, thereby addressing the key limitation of NDVI-based validation. Second, to explain the model’s finding of underutilized suitable land, targeted socio-economic surveys in districts with a high production gap are crucial to definitively uncover the causal mechanisms limiting cultivation. Third, a logical and critical extension of this work is a comprehensive multi-crop suitability analysis to identify areas unsuitable for other staple crops but highly suitable for taro, which would provide a more nuanced and powerful evidence base for crop substitution and diversification policies. Finally, future models should integrate climate projections to assess how suitability zones may shift over time. The integration of these data streams, ground truth yield measurements, socio-economic drivers, multi-crop synergies, and climate forecasts will enable the development of robust, decision-ready models to support broader agricultural diversification and climate adaptation strategies in Rwanda.

5. Conclusions

This study provides a comprehensive spatial assessment of taro (Colocasia esculenta (L.) Schott) suitability in Rwanda, applying a Fuzzy AHP-GIS to identify and quantify the ecological potential of the crop across the country. The findings reveal that 22.8% of Rwanda’s agricultural land is highly suitable and 55.7% moderately suitable for taro cultivation. The substantial availability of suitable land presents a clear opportunity to expand cultivation in areas where environmental conditions can naturally support high yields, enabling more efficient use of national land and resources.

In contrast, some districts show evidence of overuse, where taro is being cultivated beyond the boundaries of ecological suitability, possibly on marginal or unsuitable land. Such practices not only result in lower productivity but also threaten the sustainability of agricultural landscapes over time. These findings emphasize the need to realign crop distribution with biophysical realities, ensuring that cultivation practices are environmentally appropriate and economically rational.

A critical insight from this study is the national production gap of 32.4%, with over half of Rwanda’s districts exhibiting gaps exceeding 30% and some surpassing 50%. These gaps reflect both underutilization of high-potential land and inefficiencies in current agronomic practices. Addressing this gap could lead to a substantial increase in national taro production, improving rural livelihoods and directly contributing to national food security, particularly given taro’s high nutritional value and resilience under changing climate conditions.

By offering spatially detailed, ecologically grounded insights, this study equips agricultural planners with practical tools to optimize land allocation. The integration of fuzzy logic with GIS enhances precision in land use decisions, allowing interventions to be tailored to specific local contexts. These insights can support more adaptive and climate-smart planning at both the local and national levels. Moreover, the data-driven evidence presented here has direct relevance for policy-making, especially in revisiting crop prioritization frameworks and expanding support to underutilized yet promising crops like taro.

Ultimately, aligning crop choice with land suitability, while also closing the production gap through targeted extension, input provision, and investment, has the potential to transform taro from a marginal crop into a cornerstone of Rwanda’s climate-resilient agricultural future. In doing so, this study contributes not only to sustainable land management and economic opportunity but also to the broader goals of agricultural diversification and long-term food system resilience.