Tesfaye Shiferaw Sida
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Tesfaye Shiferaw Sida has a total of 174 co-authors. Here are the top co-authors by number of shared publications:
- Frédéric Baudron
International Maize and Wheat Improvement Center (CIMMYT), Harare, Zimbabwe
8 shared publications - Jordan Chamberlin
CIMMYT
4 shared publications Ken E. Giller
Plant Production Systems, Wageningen University, Wageningen, The Netherlands
4 shared publicationsAlain Ndoli
CIMMYT Harare Zimbabwe
3 shared publications- Ken Everlyn Giller
Plant Production Systems, Wageningen University and Research, Wageningen, The Netherlands
3 shared publications - Sieglinde Snapp
International Maize and Wheat Improvement Center, (CIMMYT), Carretera México‐Veracruz Texcoco Estado de México Mexico
3 shared publications
Latest Publications
- Open Access
- 16 October 2026
- Published by Wiley in Soil Use and Management
- Vol. 42 (4)
Soil Organic Carbon and Nitrogen Stocks and Isotopic Signatures Under Contrasting Integrated Crop‐Livestock Systems in Sub‐Humid Zimbabwe
Soil organic carbon (SOC) and soil total nitrogen (STN) are critical indicators of soil health and vital sinks in agroecosystems. However, knowledge gaps remain regarding how long‐term crop‐livestock integration impacts coupled SOC and STN stocks on commercial versus communal smallholder farms in Africa. This study quantified whole profile SOC and STN stocks under two land‐use types (croplands and grazing lands) and two livestock management facilities (congregation points and corrals), across two adjacent, distinct mixed crop‐livestock farming systems, while evaluating how these land uses influence δ 13 C and δ 15 N isotopic signatures. A commercial farm and an adjacent communal smallholder farm with similar soil types were selected. Soil samples were collected from six depth increments (0–5, 5–10, 10–20, 20–40, 40–70, 70–100 cm) to determine profile‐wide SOC and STN stocks alongside their δ 13 C and δ 15 N signatures. Land‐use and management type impacted SOC and STN stocks more significantly in the communal system ( p < 0.001) than on the commercial farm ( p < 0.05). On the commercial farm, whole‐profile SOC (Mg C ha −1 ± SE) and STN (Mg N ha −1 ± SE) stocks peaked in grazing lands (129.7 ± 12.8; 10.32 ± 1.14) and reached minimums at livestock congregation points (68.1 ± 2.7; 7.04 ± 1.16). Conversely, in the communal smallholder system, both stocks peaked sharply at livestock congregation points (150.0 ± 4.6; 12.26 ± 0.27) and were lowest in croplands (56.5 ± 1.5; 5.20 ± 0.20). Deep subsoil layers (40–100 cm) stored a greater relative proportion of total stocks on the commercial farm (36%–43% SOC; 39%–45% STN) than on the smallholder farm (28%–39% SOC; 29%–39% STN). Soil δ 15 N values increased with depth across all land‐use and management types. Lower δ 15 N values in commercial croplands compared with smallholder croplands likely reflect higher synthetic fertilizer use, while the highest δ 15 N values occurred in cattle corrals. Contrary to typical microbial fractionation—where preferential loss of 12 C during respiration enriches δ 13 C at depth—no consistent vertical increase in δ 13 C was observed; all land‐use types were enriched, averaging −17‰. Our results demonstrate that, within the narrow range of soil textures evaluated, land‐use and management serve as the primary determinants of SOC and STN stocks and their isotopic signatures. While livestock congregation points create intense, localized nutrient hotspots, their small spatial extent limits their landscape‐scale sequestration. Consequently, adopting conservation practices in commercial croplands as a proactive intervention, and optimizing the management of extensive communal grazing lands and croplands are collectively essential for enhancing regional biogeochemical storage and ecosystem resilience.
- 16 October 2026
- Published by Wiley in Soil Use and Management
- Vol. 42 (4)
Open AccessSoil Organic Carbon and Nitrogen Stocks and Isotopic Signatures Under Contrasting Integrated Crop‐Livestock Systems in Sub‐Humid Zimbabwe
Soil organic carbon (SOC) and soil total nitrogen (STN) are critical indicators of soil health and vital sinks in agroecosystems. However, knowledge gaps remain regarding how long‐term crop‐livestock integration impacts coupled SOC and STN stocks on commercial versus communal smallholder farms in Africa. This study quantified whole profile SOC and STN stocks under two land‐use types (croplands and grazing lands) and two livestock management facilities (congregation points and corrals), across two adjacent, distinct mixed crop‐livestock farming systems, while evaluating how these land uses influence δ 13 C and δ 15 N isotopic signatures. A commercial farm and an adjacent communal smallholder farm with similar soil types were selected. Soil samples were collected from six depth increments (0–5, 5–10, 10–20, 20–40, 40–70, 70–100 cm) to determine profile‐wide SOC and STN stocks alongside their δ 13 C and δ 15 N signatures. Land‐use and management type impacted SOC and STN stocks more significantly in the communal system ( p < 0.001) than on the commercial farm ( p < 0.05). On the commercial farm, whole‐profile SOC (Mg C ha −1 ± SE) and STN (Mg N ha −1 ± SE) stocks peaked in grazing lands (129.7 ± 12.8; 10.32 ± 1.14) and reached minimums at livestock congregation points (68.1 ± 2.7; 7.04 ± 1.16). Conversely, in the communal smallholder system, both stocks peaked sharply at livestock congregation points (150.0 ± 4.6; 12.26 ± 0.27) and were lowest in croplands (56.5 ± 1.5; 5.20 ± 0.20). Deep subsoil layers (40–100 cm) stored a greater relative proportion of total stocks on the commercial farm (36%–43% SOC; 39%–45% STN) than on the smallholder farm (28%–39% SOC; 29%–39% STN). Soil δ 15 N values increased with depth across all land‐use and management types. Lower δ 15 N values in commercial croplands compared with smallholder croplands likely reflect higher synthetic fertilizer use, while the highest δ 15 N values occurred in cattle corrals. Contrary to typical microbial fractionation—where preferential loss of 12 C during respiration enriches δ 13 C at depth—no consistent vertical increase in δ 13 C was observed; all land‐use types were enriched, averaging −17‰. Our results demonstrate that, within the narrow range of soil textures evaluated, land‐use and management serve as the primary determinants of SOC and STN stocks and their isotopic signatures. While livestock congregation points create intense, localized nutrient hotspots, their small spatial extent limits their landscape‐scale sequestration. Consequently, adopting conservation practices in commercial croplands as a proactive intervention, and optimizing the management of extensive communal grazing lands and croplands are collectively essential for enhancing regional biogeochemical storage and ecosystem resilience.
- Open Access
- 18 August 2026
- Published by Springer Nature in Communications Earth & Environment
- Vol. 7 (1) , 670
A region-specific soil health assessment protocol and evaluation for Sub-Saharan Africa
- 18 August 2026
- Published by Springer Nature in Communications Earth & Environment
- Vol. 7 (1) , 670
Open AccessA region-specific soil health assessment protocol and evaluation for Sub-Saharan Africa
- Open Access
- 16 February 2026
- Published by Springer Nature in Environmental Monitoring and Assessment
- Vol. 198 (3) , 228
Historical trends reveal significant increase in hot-dry extremes in Mexico’s Bajío region
This study provides a first-of-its-kind comprehensive analysis of temperature and precipitation extreme trends in the Bajío region of Mexico over the past four decades, utilizing high spatial resolution reanalysis data. Our findings reveal a compelling and consistent warming trend, characterized by a significant increase in warm days, warm nights, warm spell duration, and summer days, alongside a notable decrease in cold extremes. This widespread warming has profound implications for agricultural sustainability, primarily through increased heat stress and evapotranspiration. In contrast, precipitation trends exhibit a complex and spatially variable picture, with a predominant decrease in intense 5-day precipitation, consecutive wet days, and total wet-day precipitation, while 1-day extremes and the frequency of heavy rainfall days show more mixed or stable patterns. The exploration of a large set of global drivers provides a broader context for understanding these regional changes. Despite some limitations, such as potentially unbalanced data in composite analysis and the masking of certain extremes by reanalysis data, this research is crucial for understanding regional climate impacts and informing future studies. The observed shifts necessitate targeted adaptation strategies for water resource management and agricultural planning to ensure the long-term sustainability of the Bajío region.
- 16 February 2026
- Published by Springer Nature in Environmental Monitoring and Assessment
- Vol. 198 (3) , 228
Open AccessHistorical trends reveal significant increase in hot-dry extremes in Mexico’s Bajío region
This study provides a first-of-its-kind comprehensive analysis of temperature and precipitation extreme trends in the Bajío region of Mexico over the past four decades, utilizing high spatial resolution reanalysis data. Our findings reveal a compelling and consistent warming trend, characterized by a significant increase in warm days, warm nights, warm spell duration, and summer days, alongside a notable decrease in cold extremes. This widespread warming has profound implications for agricultural sustainability, primarily through increased heat stress and evapotranspiration. In contrast, precipitation trends exhibit a complex and spatially variable picture, with a predominant decrease in intense 5-day precipitation, consecutive wet days, and total wet-day precipitation, while 1-day extremes and the frequency of heavy rainfall days show more mixed or stable patterns. The exploration of a large set of global drivers provides a broader context for understanding these regional changes. Despite some limitations, such as potentially unbalanced data in composite analysis and the masking of certain extremes by reanalysis data, this research is crucial for understanding regional climate impacts and informing future studies. The observed shifts necessitate targeted adaptation strategies for water resource management and agricultural planning to ensure the long-term sustainability of the Bajío region.
- Top Cited PapersOpen Access
- 30 January 2026
- Published by Springer Nature in Plant and Soil
Integrating legumes to enhance cereal production: The relative inputs of fertiliser nitrogen and legume biological nitrogen fixation in major wheat and maize producing countries
Background: Cereal crops dominate arable agriculture and underpin global food supply. Cereal grain yield is closely related to crop nitrogen (N) uptake. However, excessive inputs of N fertiliser in cereal production systems can have negative environmental consequences and represent high financial risk for farmers. Scope: This review explores opportunities to reduce the current reliance of the global staples, wheat ( Triticum aestivum ) and maize ( Zea mays ), on N fertiliser by diversifying cereal-based cropping systems with legumes, in both time (rotation) and space (via intercropping). Increases in cereal grain yield and N uptake resulting from the inclusion of legumes in wheat and maize cropping systems are examined across different geographic regions. The role of grain legumes and legume-based forages, cover-crops and green manures is described for the 15 countries responsible for producing > 70% of the world’s wheat and maize and consuming > 80% of the global applications of N fertiliser to these two cereals. Estimates of the contributions from biological N 2 fixation and legume residual N to cropping soils are provided for individual countries and compared to the N offtake in wheat and maize grain and the quantities of N fertiliser applied. The potential increased area of legumes required for the returns of legume N to match the current amounts of N removed in grain or supplied as fertiliser is assessed for each country, and the strategies that might be necessary to encourage farmers to increase the frequency of legumes used in cereal-dominated cropping systems are outlined.
- 30 January 2026
- Published by Springer Nature in Plant and Soil
Top Cited PapersOpen AccessIntegrating legumes to enhance cereal production: The relative inputs of fertiliser nitrogen and legume biological nitrogen fixation in major wheat and maize producing countries
Background: Cereal crops dominate arable agriculture and underpin global food supply. Cereal grain yield is closely related to crop nitrogen (N) uptake. However, excessive inputs of N fertiliser in cereal production systems can have negative environmental consequences and represent high financial risk for farmers. Scope: This review explores opportunities to reduce the current reliance of the global staples, wheat ( Triticum aestivum ) and maize ( Zea mays ), on N fertiliser by diversifying cereal-based cropping systems with legumes, in both time (rotation) and space (via intercropping). Increases in cereal grain yield and N uptake resulting from the inclusion of legumes in wheat and maize cropping systems are examined across different geographic regions. The role of grain legumes and legume-based forages, cover-crops and green manures is described for the 15 countries responsible for producing > 70% of the world’s wheat and maize and consuming > 80% of the global applications of N fertiliser to these two cereals. Estimates of the contributions from biological N 2 fixation and legume residual N to cropping soils are provided for individual countries and compared to the N offtake in wheat and maize grain and the quantities of N fertiliser applied. The potential increased area of legumes required for the returns of legume N to match the current amounts of N removed in grain or supplied as fertiliser is assessed for each country, and the strategies that might be necessary to encourage farmers to increase the frequency of legumes used in cereal-dominated cropping systems are outlined.
- Open Access
- 23 March 2025
- Published by Elsevier in Journal of Agriculture and Food Research
- Vol. 21
The search for best fertilizer combination to increase wheat (Triticum aestivum L.) yield in north-western Ethiopia
- 23 March 2025
- Published by Elsevier in Journal of Agriculture and Food Research
- Vol. 21
Open AccessThe search for best fertilizer combination to increase wheat (Triticum aestivum L.) yield in north-western Ethiopia
- Open Access
- 18 September 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16 (18) , 3451
Mapping Field-Level Maize Yields in Ethiopian Smallholder Systems Using Sentinel-2 Imagery
Remote sensing offers a low-cost method for estimating yields at large spatio-temporal scales. Here, we examined the ability of Sentinel-2 satellite imagery to map field-level maize yields across smallholder farms in two regions in Oromia district, Ethiopia. We evaluated how effectively different indices, the MTCI, GCVI, and NDVI, and different models, linear regression and random forest regression, can be used to map field-level yields. We also examined if models improved by adding weather and soil data and how generalizable our models were if trained in one region and applied to another region, where no data were used for model calibration. We found that random forest regression models that used monthly MTCI composites led to the highest yield prediction accuracies (R2 up to 0.63), particularly when using only localized data for training the model. These models were not very generalizable, especially when applied to regions that had significant haze remaining in the imagery. We also found that adding soil and weather data did little to improve model fit. Our results highlight the ability of Sentinel-2 imagery to map field-level yields in smallholder systems, though accuracies are limited in regions with high cloud cover and haze.
- 18 September 2024
- Published by MDPI AG in Remote Sensing
- Vol. 16 (18) , 3451
Open AccessMapping Field-Level Maize Yields in Ethiopian Smallholder Systems Using Sentinel-2 Imagery
Remote sensing offers a low-cost method for estimating yields at large spatio-temporal scales. Here, we examined the ability of Sentinel-2 satellite imagery to map field-level maize yields across smallholder farms in two regions in Oromia district, Ethiopia. We evaluated how effectively different indices, the MTCI, GCVI, and NDVI, and different models, linear regression and random forest regression, can be used to map field-level yields. We also examined if models improved by adding weather and soil data and how generalizable our models were if trained in one region and applied to another region, where no data were used for model calibration. We found that random forest regression models that used monthly MTCI composites led to the highest yield prediction accuracies (R2 up to 0.63), particularly when using only localized data for training the model. These models were not very generalizable, especially when applied to regions that had significant haze remaining in the imagery. We also found that adding soil and weather data did little to improve model fit. Our results highlight the ability of Sentinel-2 imagery to map field-level yields in smallholder systems, though accuracies are limited in regions with high cloud cover and haze.
- Open Access
Landscape and Micronutrient Fertilizer Effect on Agro-Fortified Wheat and Teff Grain Nutrient Concentration in Western Amhara
Agronomic biofortification, encompassing the use of mineral and organic nutrient resources which improve micronutrient concentrations in staple crops is a potential strategy to promote the production of and access to micronutrient-dense foods at the farm level. However, the heterogeneity of smallholder farming landscapes presents challenges on implementing agronomic biofortification. Here, we test the effects of zinc (Zn)- and selenium (Se)-containing fertilizer on micronutrient concentrations of wheat (Triticum aestivum L.) and teff (Eragrostis tef (Zucc.) Trotter) grown under different landscape positions and with different micronutrient fertilizer application methods in the western Amhara region of Ethiopia. Field experiments were established in three landscape positions at three sites, with five treatments falling into three broad categories: (1) nitrogen (N) fertilizer rate; (2) micronutrient fertilizer application method; (3) sole or co-application of Zn and Se fertilizer. Treatments were replicated across five farms per landscape position and over two cropping seasons (2018 and 2019). Grain Zn concentration ranged from 26.6 to 36.4 mg kg−1 in wheat and 28.5–31.2 mg kg−1 in teff. Grain Se concentration ranged from 0.02 to 0.59 mg kg−1 in wheat while larger concentrations of between 1.01 and 1.55 mg kg−1 were attained in teff. Larger concentrations of Zn and Se were consistently attained when a foliar fertilizer was applied. Application of ⅓ nitrogen (N) yielded significantly larger grain Se concentration in wheat compared to a recommended N application rate. A moderate landscape effect on grain Zn concentration was observed in wheat but not in teff. In contrast, strong evidence of a landscape effect was observed for wheat and teff grain Se concentration. There was no evidence for any interaction of the treatment contrasts with landscape position except in teff, where an interaction effect between landscape position and Se application was observed. Our findings indicate an effect of Zn, Se, N, landscape position, and its interaction effect with Se on grain micronutrient concentrations. Agronomic biofortification of wheat and teff with micronutrient fertilizers is influenced by landscape position, the micronutrient fertilizer application method and N fertilizer management. The complexity of smallholder environmental settings and different farmer socio-economic opportunities calls for the optimization of nutritional agronomy landscape trials. Targeted application of micronutrient fertilizers across a landscape gradient is therefore required in ongoing agronomic biofortification interventions, in addition to the micronutrient fertilizer application method and the N fertilizer management strategy.
Open AccessLandscape and Micronutrient Fertilizer Effect on Agro-Fortified Wheat and Teff Grain Nutrient Concentration in Western Amhara
Agronomic biofortification, encompassing the use of mineral and organic nutrient resources which improve micronutrient concentrations in staple crops is a potential strategy to promote the production of and access to micronutrient-dense foods at the farm level. However, the heterogeneity of smallholder farming landscapes presents challenges on implementing agronomic biofortification. Here, we test the effects of zinc (Zn)- and selenium (Se)-containing fertilizer on micronutrient concentrations of wheat (Triticum aestivum L.) and teff (Eragrostis tef (Zucc.) Trotter) grown under different landscape positions and with different micronutrient fertilizer application methods in the western Amhara region of Ethiopia. Field experiments were established in three landscape positions at three sites, with five treatments falling into three broad categories: (1) nitrogen (N) fertilizer rate; (2) micronutrient fertilizer application method; (3) sole or co-application of Zn and Se fertilizer. Treatments were replicated across five farms per landscape position and over two cropping seasons (2018 and 2019). Grain Zn concentration ranged from 26.6 to 36.4 mg kg−1 in wheat and 28.5–31.2 mg kg−1 in teff. Grain Se concentration ranged from 0.02 to 0.59 mg kg−1 in wheat while larger concentrations of between 1.01 and 1.55 mg kg−1 were attained in teff. Larger concentrations of Zn and Se were consistently attained when a foliar fertilizer was applied. Application of ⅓ nitrogen (N) yielded significantly larger grain Se concentration in wheat compared to a recommended N application rate. A moderate landscape effect on grain Zn concentration was observed in wheat but not in teff. In contrast, strong evidence of a landscape effect was observed for wheat and teff grain Se concentration. There was no evidence for any interaction of the treatment contrasts with landscape position except in teff, where an interaction effect between landscape position and Se application was observed. Our findings indicate an effect of Zn, Se, N, landscape position, and its interaction effect with Se on grain micronutrient concentrations. Agronomic biofortification of wheat and teff with micronutrient fertilizers is influenced by landscape position, the micronutrient fertilizer application method and N fertilizer management. The complexity of smallholder environmental settings and different farmer socio-economic opportunities calls for the optimization of nutritional agronomy landscape trials. Targeted application of micronutrient fertilizers across a landscape gradient is therefore required in ongoing agronomic biofortification interventions, in addition to the micronutrient fertilizer application method and the N fertilizer management strategy.
- Open Access
- 21 July 2023
- Published by Elsevier in Computers and Electronics in Agriculture
- Vol. 212
Failure to scale in digital agronomy: An analysis of site-specific nutrient management decision-support tools in developing countries
- 21 July 2023
- Published by Elsevier in Computers and Electronics in Agriculture
- Vol. 212
Open AccessFailure to scale in digital agronomy: An analysis of site-specific nutrient management decision-support tools in developing countries
- Top Cited PapersOpen Access
- 29 June 2023
- Published by Springer Nature in Nature Sustainability
- Vol. 6 (10) , 1268-1278
Spatially differentiated nitrogen supply is key in a global food–fertilizer price crisis
A regional geopolitical conflict and sudden massive supply disruptions have revealed vulnerabilities in our global fuel–fertilizer–food nexus. As nitrogen (N) fertilizer price spikes threaten food security, differentiated responses are required to maintain staple cereal yields across over- and underfertilized agricultural systems. Through integrated management of organic and inorganic N sources in high- to low-input cereal production systems, we estimate potential total N-fertilizer savings of 11% in India, 49% in Ethiopia and 44% in Malawi. Shifting to more cost-effective, high-N fertilizer (such as urea), combined with compost and integration of legumes, can optimize N in N-deficient systems. Better targeted and more efficient N-fertilizer use will benefit systems with surplus N. Geospatially differentiated fertilization strategies should prioritize high-N fertilizer supply to low-yield, N-deficient locations and balanced fertilization of N, P, K and micronutrients in high-yield systems. Nationally, governments can invest in extension and realign subsidies to enable and incentivize improved N management at the farm level.
- 29 June 2023
- Published by Springer Nature in Nature Sustainability
- Vol. 6 (10) , 1268-1278
Top Cited PapersOpen AccessSpatially differentiated nitrogen supply is key in a global food–fertilizer price crisis
A regional geopolitical conflict and sudden massive supply disruptions have revealed vulnerabilities in our global fuel–fertilizer–food nexus. As nitrogen (N) fertilizer price spikes threaten food security, differentiated responses are required to maintain staple cereal yields across over- and underfertilized agricultural systems. Through integrated management of organic and inorganic N sources in high- to low-input cereal production systems, we estimate potential total N-fertilizer savings of 11% in India, 49% in Ethiopia and 44% in Malawi. Shifting to more cost-effective, high-N fertilizer (such as urea), combined with compost and integration of legumes, can optimize N in N-deficient systems. Better targeted and more efficient N-fertilizer use will benefit systems with surplus N. Geospatially differentiated fertilization strategies should prioritize high-N fertilizer supply to low-yield, N-deficient locations and balanced fertilization of N, P, K and micronutrients in high-yield systems. Nationally, governments can invest in extension and realign subsidies to enable and incentivize improved N management at the farm level.
- Open Access
- 1 September 2022
- Published by MDPI AG in Agriculture
- Vol. 12 (9) , 1352
Enhancing Smallholder Wheat Yield Prediction through Sensor Fusion and Phenology with Machine Learning and Deep Learning Methods
Field-scale prediction methods that use remote sensing are significant in many global projects; however, the existing methods have several limitations. In particular, the characteristics of smallholder systems pose a unique challenge in the development of reliable prediction methods. Therefore, in this study, a fast and reproducible new approach to wheat prediction is developed by combining predictors derived from optical (Sentinel-2) and radar (Sentinel-1) sensors using a diverse set of machine learning and deep learning methods under a small dataset domain. This study takes place in the wheat belt region of Ethiopia and evaluates forty-two predictors that represent the major vegetation index categories of green, water, chlorophyll, dry biomass, and VH polarization SAR indices. The study also applies field-collected agronomic data from 165 farm fields for training and validation. According to results, compared to other methods, a combined automated machine learning (AutoML) approach with a generalized linear model (GLM) showed higher performance. AutoML, which reduces training time, delivered ten influential parameters. For the combined approach, the mean RMSE of wheat yield was from 0.84 to 0.98 ton/ha using ten predictors from the test dataset, achieving a 99% confidence interval. It also showed a correlation coefficient as high as 0.69 between the estimated yield and measured yield, and it was less sensitive to the small datasets used for model training and validation. A deep neural network with three hidden layers using the ten influential parameters was the second model. For this model, the mean RMSE of wheat yield was between 1.31 and 1.36 ton/ha on the test dataset, achieving a 99% confidence interval. This model used 55 neurons with respective values of 0.1, 0.5, and 1 × 10−4 for the hidden dropout ratio, input dropout ratio, and l2 regularization. The approaches implemented in this study are fast and reproducible and beneficial to predict yield at scale. These approaches could be adapted to predict grain yields of other cereal crops grown under smallholder systems in similar global production systems.
- 1 September 2022
- Published by MDPI AG in Agriculture
- Vol. 12 (9) , 1352
Open AccessEnhancing Smallholder Wheat Yield Prediction through Sensor Fusion and Phenology with Machine Learning and Deep Learning Methods
Field-scale prediction methods that use remote sensing are significant in many global projects; however, the existing methods have several limitations. In particular, the characteristics of smallholder systems pose a unique challenge in the development of reliable prediction methods. Therefore, in this study, a fast and reproducible new approach to wheat prediction is developed by combining predictors derived from optical (Sentinel-2) and radar (Sentinel-1) sensors using a diverse set of machine learning and deep learning methods under a small dataset domain. This study takes place in the wheat belt region of Ethiopia and evaluates forty-two predictors that represent the major vegetation index categories of green, water, chlorophyll, dry biomass, and VH polarization SAR indices. The study also applies field-collected agronomic data from 165 farm fields for training and validation. According to results, compared to other methods, a combined automated machine learning (AutoML) approach with a generalized linear model (GLM) showed higher performance. AutoML, which reduces training time, delivered ten influential parameters. For the combined approach, the mean RMSE of wheat yield was from 0.84 to 0.98 ton/ha using ten predictors from the test dataset, achieving a 99% confidence interval. It also showed a correlation coefficient as high as 0.69 between the estimated yield and measured yield, and it was less sensitive to the small datasets used for model training and validation. A deep neural network with three hidden layers using the ten influential parameters was the second model. For this model, the mean RMSE of wheat yield was between 1.31 and 1.36 ton/ha on the test dataset, achieving a 99% confidence interval. This model used 55 neurons with respective values of 0.1, 0.5, and 1 × 10−4 for the hidden dropout ratio, input dropout ratio, and l2 regularization. The approaches implemented in this study are fast and reproducible and beneficial to predict yield at scale. These approaches could be adapted to predict grain yields of other cereal crops grown under smallholder systems in similar global production systems.