AMPLIFY  VOL. 37, NO. 7
  

The agriculture sector faces numerous challenges from climate change and global warming. Climate catastrophes have degraded billions of hectares of land and depleted water reservoirs, resulting in greater use of chemicals and overexploitation of natural resources.¹ Meanwhile, emissions from farm activities contribute considerably to global warming. The recursive cause-and-effect relationship between agriculture and climate change makes it increasingly difficult to achieve the goals of food security and sustainable agriculture.²

AI is expected to transform the agricultural sector. Farm organizations can use AI-based tools to manage resources more efficiently and support climate resilience. Climate-resilient agriculture (CRA) practices aim to enhance the long-term health of agricultural systems and the social systems that depend on them by combining traditional agricultural knowledge with modern techniques.³

The capacity of AI to analyze data on weather conditions, soil conditions, and crop genetics can improve CRA practices by facilitating the development of climate-resistant seeds and enabling informed decisions on harvesting and planting times, water management, and chemical use.⁴ However, the transformation to CRA practices is hindered by complex sociocultural, technical, and ecological barriers.^5,6

In this article, we discuss the barriers smallholder farmers face in using AI tools to manage the impacts of climate change on arable farming (crops and fruit) practices. Smallholder farmers represent 95% of the world’s farmers and produce 45% of global food.⁷ They rely on rain-fed agriculture (little or no irrigation), cultivate in marginal areas, lack access to technical or financial support, and face increased vulnerability due to climate change.⁸ We present three case studies from India, highlighting how these farm organizations leverage their traditional knowledge and practices to overcome socio-technological-ecological barriers and maximize the value of AI tools for CRA.

CRA & AI

AI-based technologies are increasingly being harnessed to develop agricultural practices that can more easily adapt to climate change. Technologies such as intelligent robots and drones, deep learning, and generative AI are being used to create resilient seed varieties, predict climate patterns, and recommend crop varieties based on climate forecasts.⁹

Intelligent technologies are also being employed to monitor environmental conditions that contribute to the spread of diseases and pests, which are further worsened by climate change. For example, by analyzing soil moisture levels and weather forecasts, AI-based tools can optimize irrigation schedules to conserve water while ensuring that crops receive the optimal amount of water, even in water-stressed areas.¹⁰ Further, by integrating AI with other agricultural technologies, such as electric tractors and drip irrigation, farmers can use critical resources, including fertilizers and pesticides, more efficiently to minimize soil degradation and harmful environmental impacts.¹¹

Traditional Knowledge in CRA

Traditional farming knowledge, developed over generations, includes a location-specific understanding of climate, plant varieties, soil fertility, and seasonal weather calendars.¹² This includes experience and wisdom achieved through direct (own experiences) or indirect (others’ experiences) observations. Smallholder farmers regularly rely on traditional knowledge in their agricultural operations as a way to ensure agricultural diversity, meaningful livelihoods, food security, ecological health, and biodiversity.¹³

For instance, farmers can use traditional knowledge formed from observations of bird behavior, soil conditions, and atmospheric events to predict precipitation and weather patterns. This knowledge can inform crop selection, planting times, and harvesting activities. Other traditional practices, such as seed storage in facilities on farms and homes, help preserve viable seeds for food security and biodiversity.

Despite the value of traditional knowledge in dealing with short-term weather variability, severe climate change effects are threatening long-term efficacy and increasing the need for support from modern information technologies.¹⁴ Traditional knowledge combined with AI and real-time data enables better decision-making as well as enhanced prediction, planning, and preparation for potential climate and weather-related shocks to protect communities and agricultural systems.¹⁵ Still, to take full advantage of AI-based technologies, farm organizations must overcome sociocultural, technical, and ecological barriers.¹⁶

AI in Agriculture: 3 Case Studies

We have researched the use of AI in sustainable agriculture over the past three years, speaking with farmers and IT providers to understand their challenges and experiences. Our conversations reveal how farm organizations integrate traditional farming knowledge with AI-based tools for climate-smart crop management and sustainable agricultural outcomes. Below, we present three exemplar cases (pseudonyms have been used for the real farm names).

1. Organic Orchard

Organic Orchard is located in the western part of India. This farm grows fruit and operates a plant nursery. The owner is passionate about organic agriculture and promotes organic farming practices that prioritize soil health and fertility.

In 2021, Organic Orchard implemented an AI-based disease-detection app to support low- or no-chemical fruit plantation management. Using the application involved uploading photos of infected plants to the app to receive AI-generated diagnoses of potential diseases and nutrient deficiencies, along with disease management recommendations.

The owner explained: “The app suggests everything: chemical management, biological management, and mechanical management. Our focus is on organic farming, so whatever information is available regarding organic farming, we apply that.” Once Organic Orchard received information on nutrient deficiencies from the app, workers supplemented the app’s recommendations with traditional ways of fertilizing (e.g., oilseed cake to treat nitrogen deficiencies).

Organic Orchard actively created community awareness and promoted the benefits of the app, including increased productivity and reduced chemical application. However, some farm workers and others in the community were wary of using the app. They perceived it as a threat to their long-held beliefs and local farming practices because it did not provide information on fruits commonly grown in the area. These perceived threats were exacerbated by a lack of engagement from government experts who were responsible for advising farmers on best practices for crop management and pest and disease control.

One participant pointed out: “Technology should be made to reach farmers’ fields and is the responsibility of the extension workers. But you know about the government officers (extension workers); they do not come out of their office rooms.”

To build confidence in the technology, Organic Orchard’s owner organized meetings to share his experience of integrating technology with traditional knowledge and practices. He explained: “Because we work in the fields, we have an idea about the weather. We know pest attack is probable when the clouds gather. So, we already know, and we get confirmation through the app, and we feel that we are going in the right direction.”

Over time, Organic Orchard combined traditional practices like observing atmospheric events with the AI-based tool for disease detection to improve decision-making and support CRA by limiting the application of chemical fertilizers and promoting organic farming to restore soil health.

2. Crop Farm

Crop Farm is situated in the hilly northern region of India, where it grows commercial trees, wheat, and seasonal vegetables. In the past, Crop Farm was successful in applying traditional knowledge gained from observing atmospheric conditions and estimating rainfall. However, changing weather patterns and decreased precipitation increased water problems in the already water-scarce area, rendering traditional knowledge insufficient for effective crop management.

For example, Crop Farm traditionally fertilized its crops 30 days after plantation, but shifting rainfall patterns affected fertilization efficiency, causing low yields. Similarly, the farmer followed the traditional method of harvesting wheat at the end of winter and leaving the harvested wheat grains in the fields to dry. Recently, unpredictable weather and sudden rainfall frequently damaged the grains.

Crop Farm began using AI-based weather-prediction apps to get rainfall estimates in an effort to better manage water levels in the fields. The goal was to better prepare the fields for fertilization, protect threshed grains from damage, and arrange for alternative irrigation sources as required. One farm worker reported: “At times, after threshing, the wheat was kept in the fields, and suddenly it would rain at night, and the wheat would be wet and damaged. Now, we work by watching the forecast. Harvesting and threshing are mostly done after consulting the weather predictions.”

During the implementation and use of the apps, Crop Farm encountered issues due to the remote location of the village. Lack of Internet connectivity and technology infrastructure (e.g., weather stations) created problems with weather-prediction accuracy and efficacy for the apps. Because of these technical barriers, Crop Farm returned to using traditional knowledge of weather predictions for managing farm activities. A farm worker pointed out: “When the temperature has risen to 48-49 degrees Celsius, then usually we get rain after two to three days to lower the temperature. So, we have some of these ways to predict the weather.”

Based on traditional knowledge and observations of the natural rain cycle, Crop Farm’s owners understood the importance of capturing and storing rainwater for irrigation in case of delayed monsoons. They combined their experience and reasoning with app-based weather predictions to adapt to evolving weather patterns resulting from climate change and prioritized rainwater harvesting, as one of the workers explained: “For a long time, there used to be much rain in November–December here, but this year November [and] December were dry, and it only rained in January. So, I have built a temporary dam under my house, where the rainwater gets collected, then the stored water is used for irrigation.” 

This combined learning helped them better understand changing weather patterns, implement sustainable and efficient water management methods, and reduce soil erosion.

3. Cereal Farm

Cereal Farm, a family farm in the northeastern region of India that grows cereals, also faces the impacts of climate change. Once known for its high fertility and ample rainfall, the region now experiences droughts. Climate change has led to unpredictable weather patterns, sudden and delayed rainfall, and hailstorms — leading to reduced productivity and crop quality.

To tackle these issues, Cereal Farm deployed an AI-based crop advisory app that provides weather predictions, planting and harvesting recommendations, and disease-detection and remedial recommendations. Simultaneously, as recommended by the app, the owners switched to drought-resistant hybrid maize seeds to enhance productivity and use.

However, farmers expressed concerns about problems like unreliable weather predictions, failure of suggested recommendations for fertilizer applications, unsuitability of hybrid seeds to local conditions, and loss of local seed varieties. One farmer shared: “Our ancestors saved some of the harvest to be used as seeds in the next season. When the technology came, hybrid seeds were suggested to increase crop yield. But when harvesting season came, we realized the seeds hadn’t grown properly. So, people switched from the hybrid seeds to the native seeds they used to collect every year, which led to the betterment of seed quality every year.

Despite using the app, they continued to rely on their intuition, experiences, and traditional knowledge. One farmer said: “Traditional knowledge has its own importance, but when it comes to dealing with climate change, it will be important to use technology and take advantage of it. Both should be considered side by side.”

The combination of traditional knowledge with recommendations from the crop advisory app was the best strategy for planting and harvesting decisions at Cereal Farm. As another farmer put it: “The things that have been carried out through generations about what should be planted in which season, how the field should be prepared, and how things should be done, these are the traditional knowledge. What new do we add to this? When will it rain? Should we irrigate or not? When should the crop be harvested, and where should the products be sold? This is what is new, and with app tools, we get to know how these should be done.”

Common Barriers to AI Use in CRA Practices

These case studies show the range of sociocultural, technical, and ecological barriers farmers face in using AI for CRA:

• Sociocultural barriers — challenges related to the social and cultural contexts in which AI-based technologies are deployed:

  • Insufficient knowledge of AI within farm organizations, shortage of domain experts, and inadequate engagement of stakeholders (e.g., government officers) can limit organizational awareness of the value and benefits of AI for CRA.

  • Culturally specific factors, such as community norms and practices and workers’ reluctance to try new practices, can decrease AI’s potential for CRA practices.

• Technical barriers — real and perceived problems associated with AI-based technologies and tools:

  • AI-based tools trained on global data and non-representative farm-specific data can lead to inconsistent and even conflicting predictions. These inconsistencies (e.g., weather predictions) can increase organizations’ reluctance to use the tools.

  • A lack of technological infrastructure within rural farming regions (e.g., weather stations) can amplify prediction inaccuracies.

• Ecological barriers — arise from the dynamic conditions in the natural environment in which agricultural activities take place:

  • Lack of location-specific, contextual awareness of AI-based tools can create problems in achieving desired results because recommendations can fail in local conditions.

  • Reliance on AI-based technology for CRA practices (e.g., climate-resistant and hybrid seeds) can raise ecological risks of losing local seed diversity.

Potential of Conjoint Learning Approach for CRA

Overcoming the aforementioned barriers will require additional development of AI-based technologies. It will also require listening to — and learning from — agricultural workers. A conjoint learning approach that combines traditional knowledge with AI tools can enhance AI performance and help overcome the barriers discussed above. Specifically:

• Traditional knowledge combined with direct observations can be useful in addressing technical barriers. Climate change makes it difficult for AI tools to predict microclimatic shifts.¹⁷ By incorporating local knowledge with satellite imaging and remote sensing, AI tools can provide more precise weather forecasting and improve adaptation of agricultural practices at a local scale.

• Social networks, indirect observation, and the integration of traditional farming knowledge with AI can help resolve sociocultural barriers. Traditional knowledge emphasizes cultural sensitivity and respect for nature, which can be integrated into AI tools to prevent bias and ensure cultural consciousness. Incorporating information on local culture, crop varieties, and languages through conjoint learning can increase trust in AI tools. It can also make AI tools relevant and useful for smallholder farmers who do not grow commodity crops or who practice organic or regenerative farming.¹⁸

• When combined with traditional ecological knowledge, which includes the relationship between plants, animals, land, and people, AI-based tools are more context-aware. By enhancing the understanding of ecological and atmospheric events, they can support better crop management, water and resource management, and chemical usage. Traditional ecological knowledge, when combined with AI can assist in ecological restoration by analyzing historical data, identifying native species, and predicting suitable areas for reforestation, habitat restoration, and ecosystem rehabilitation.¹⁹

By broadening the view of AI use in CRA from a technical perspective to a socio-cultural-technical-ecological one,²⁰ conjoint learning offers a more holistic way to advance AI-based tools for CRA and encourages more reflection on how agricultural practices can be adapted based on the vulnerabilities of a specific farming system.^21,22 Figure 1 presents a framework for understanding these relationships.

Recommendations

Conjoint learning offers ways to improve AI-based agricultural solutions, ultimately leading to better outcomes for CRA:

• Organic Orchard actively shared its successful experiences of conjoint learning in managing local crops with community members to raise awareness and build trust in AI-based systems.

• Crop Farm gained valuable insights into changing local weather patterns and used app-based predictions to add value to traditional farm practices.

• Cereal Farm combined its traditional ecological knowledge with app-based weather predictions to enhance planting and harvesting decisions.

To reap the full benefits of conjoint learning to enhance AI effectiveness for CRA, agriculture technology organizations should prioritize conjoint learning in their own activities, from development to deployment. This would involve understanding farmers’ expectations for using AI to respond to climate change impacts and developing respectful ways to collect and organize traditional knowledge.

For example, to increase context awareness and prediction accuracies of AI-based technologies, the developers of the crop advisory app used by Cereal Farm organized monthly meetings with the farmers in their fields to collect real-time data. This helped the developers gain a deeper appreciation of farmers’ needs and the value of their traditional knowledge.

Policymakers and government agencies that bear responsibility for promoting and supporting CRA can also benefit from, and contribute to, conjoint learning. As we saw with Organic Orchard, lack of involvement from government representatives can impede the adoption of technology by smallholder farmers, delaying CRA goal achievement. Note that as policymakers start to use AI-based tools to support their work in devising climate-resilience strategies, they will need to integrate their own traditional knowledge.

Engagement among various stakeholders (farmers, developers, and policymakers) can create a conducive environment for efficient development and use of AI-based tools in addressing climate change impacts:

• Farming organizations should establish connections with other farmers and domain experts, gather information about agricultural innovations, collaborate with technology providers in improving AI-based solutions, and embrace innovation in their farming practices.

• Agricultural AI providers should engage in collaborative research to understand farmers’ needs, develop ways to weave traditional knowledge into their solutions, and involve farmers in development processes to enhance AI-based tools and support effective deployment and implementation.

• Policymakers and research organizations should educate farming organizations on the use and benefits of AI, create farming knowledge hubs, and facilitate open discussions with technology providers to promote awareness of and trust in AI-based tools.

Conclusion

When combined with traditional knowledge, AI-based technologies have the potential to address the sociocultural, technical, and ecological barriers of AI use for CRA. A conjoint learning approach can lead to more accurate and inclusive AI solutions, benefiting diverse cultures and farm environments. Operationalizing this approach will ensure the effectiveness of our response to the existential challenges of climate change.

References

¹ Sartori, Martina, et al. “Remaining Loyal to Our Soil: A Prospective Integrated Assessment of Soil Erosion on Global Food Security.” Ecological Economics, Vol. 219, May 2024.

² “Climate Smart Agriculture Sourcebook.” Food and Agriculture Organization of the United Nations (FAO), accessed July 2024.

³ Okoronkwo, David John, et al. “Climate Smart Agriculture? Adaptation Strategies of Traditional Agriculture to Climate Change in Sub-Saharan Africa.” Frontiers in Climate, Vol. 6, January 2024.

⁴ Hall, Curt. “Sustainable, Intelligent & Connected: Electric Tractors for Precision Agriculture.” Cutter Consortium Sustainability Advisor, 29 May 2024.

⁵ Mana, Aali, et al. “Sustainable AI-Based Production Agriculture: Exploring AI Applications and Implications in Agricultural Practices.” Smart Agricultural Technology, Vol. 7, No. 7664, February 2024.

⁶ Lakshmi, Vijaya, and Jacqueline Corbett. “Using AI to Improve Sustainable Agricultural Practices: A Literature Review and Research Agenda.” Communications of the Association for Information Systems, Vol. 53, 2023.

⁷ Heldreth, Courtney, et al. “What Does AI Mean for Smallholder Farmers? A Proposal for Farmer-Centered AI Research.” Interactions, Issue 28, No. 4, July–August 2021.

⁸ Duchicela, Sisimac A., et al. “Microclimatic Warming Leads to a Decrease in Species and Growth Form Diversity: Insights from a Tropical Alpine Grassland.” Frontiers in Ecology and Evolution, Vol. 9, September 2021.

⁹ Mana et al. (see 5).

¹⁰ Heldreth et al. (see 7).

¹¹ Duchicela et al. (see 8).

¹² Abbasi, Rabiya, Pablo Martinez, and Rafiq Ahmad. “The Digitization of Agricultural Industry — A Systematic Literature Review on Agriculture 4.0.” Smart Agricultural Technology, Vol. 2, December 2022.

¹³ Abbasi et al. (see 12).

¹⁴ Imoro, Ziblim Abukari, et al. “Harnessing Indigenous Technologies for Sustainable Management of Land, Water, and Food Resources Amidst Climate Change.” Frontiers in Sustainable Food Systems, Vol. 5, August 2021.

¹⁵ Imoro et al. (see 14).

¹⁶ Lakshmi and Corbett (see 6).

¹⁷ Duchicela et al. (see 8).

¹⁸ Martinescu, Livia. “AI for Climate Change: Using Artificial and Indigenous Intelligence to Fight Climate Change.” Oxford Insights, 4 December 2023.

¹⁹ Abbasi et al. (see 12).

²⁰ Rahman, Saeed, Natalie Slawinski, and Monika Winn. “How Ecological Knowledge Can Catalyze System-Level Change: Lessons from Agriculture & Beyond.” Amplify, Vol. 1, No. 5, 2022. 

²¹ Lakshmi and Corbett (see 6).

²² Zheng, Hongyun, Wanglin Ma, and Quan He. “Climate-Smart Agricultural Practices for Enhanced Farm Productivity, Income, Resilience, and Greenhouse Gas Mitigation: A Comprehensive Review.” Mitigation and Adaptation Strategies for Global Change, Vol. 29, No. 8, March 2024.