Unmanned aerial vehicle hierarchical detection of leaf blast in rice crops based on a specific spectral vegetation index

Guangming LI, Dongxue ZHAO, Jinpeng LI, Shuai FENG, Chunling CHEN

Front. Agr. Sci. Eng. ››

PDF(1570 KB)
Front. Agr. Sci. Eng. All Journals
PDF(1570 KB)
Front. Agr. Sci. Eng. ›› DOI: 10.15302/J-FASE-2024576
RESEARCH ARTICLE

Unmanned aerial vehicle hierarchical detection of leaf blast in rice crops based on a specific spectral vegetation index

Author information +
History +

Highlights

● A new vegetation index, rice blast index (RBI), was constructed to detect rice leaf blast.

● The disease detection performance of RBI, TVI, DDI and MTVI1 vegetation indices were compared.

● The level of leaf blast disease in the field was evaluated using the new RBI.

Abstract

Leaf blast is a significant global problem, severely affecting rice quality and yield, making swift, non-invasive detection crucial for effective field management. This study used hyperspectral remote sensing technology via an unmanned aerial vehicle to gather spectral data from rice crops. ANOVA and the Relief-F algorithm were used to identify spectral bands sensitive to the disease and developed a new vegetation index, the rice blast index (RBI). This RBI was compared with 30 established vegetation indexes, using correlation analysis and visual comparison to further shortlist six superior indexes, including RBI. These were evaluated using the K-nearest neighbor (KNN) and random forests (RF) classification models. RBI demonstrated superior detection accuracy for leaf blast in both the KNN model (95.0% overall accuracy and 93.8% kappa coefficient) and the RF model (95.1% overall accuracy and 92.5% kappa coefficient). This study highlights the significant potential of RBI as an effective tool for precise leaf blast detection, offering a powerful new mechanism and theoretical basis for enhanced disease management in rice cultivation.

Graphical abstract

Keywords

Drone remote sensing technology / hyperspectral technology / leaf blast disease / rice / vegetation index

Cite this article

Download citation ▾
Guangming LI, Dongxue ZHAO, Jinpeng LI, Shuai FENG, Chunling CHEN. Unmanned aerial vehicle hierarchical detection of leaf blast in rice crops based on a specific spectral vegetation index. Front. Agr. Sci. Eng., https://doi.org/10.15302/J-FASE-2024576

1 INTRODUCTION

Agricultural extension and advisory services (EAS) facilitate the access of farmers to knowledge, information, financial services and technologies necessary for improving farm performance[1]. However, access to EAS remains a critical issue in many rural settings[2]. This is in part due to insufficient funds for supporting public extension, lack of appropriate strategies for effective research, and limited coverage of extension services[3]. Public advisory services in particular have been criticized for their inability to provide satisfactory services to farmers due to the fact that they are generally supply-driven and do not consider the actual farmer needs[4]. On the demand size, small land holdings, lack of education and low incomes are some of the key barriers that restrict farmer access to EAS[57]. Also, empirical studies have shown that extension services have not equally benefited female and male farmers due to differentials in access to and control of production resources and participation in training programs and decision making[8]. Farmers are experiencing new and emerging farming challenges such as climate change and related extreme events and transboundary invasive insect pests and diseases. Many argue that agricultural extension services need transformation through tested and proven decision-support devices and digital revolution to improve production, reduce crop losses and increase productivity.
Digitally-enabled EAS using single or a combination of devices can potentially ameliorate the inadequate technical assistance to farmers occasioned by a lean extension staff and provide information to marginalized and hard-to-reach areas[9,10]. Digital options include radio programs using add-on features, television shows, videos shared online, mobile mediated value-added services or agricultural value-added services, digital decision-support devices, digital learning devices and the internet. Evaluation studies have shown considerable cost-effectiveness of digital EAS in the long run and the potential to deliver timely, relevant and actionable information to farmers even in remote locations, increasing the adoption of technology. Tambo et al.[11] showed that participation in the information and communication technology (ICT)-based extension campaigns significantly increased farmer knowledge about fall armyworm and stimulated the adoption of agricultural technologies and practices for the management of the pest. Silvestri et al.[12] and Hudson et al.[13] showed significant positive effects on knowledge scores and agricultural technology adoption by farmers who listened to promotional radio programs compared to those who did not. The information provided via digital platforms is also becoming diverse, ranging from specific technologies, market access, price information, weather information, application of inputs and early warning of drought, floods, and pests/diseases, allowing farmers to make more informed decisions on how to improve their agricultural practices. Other changes facilitating increased use of digital devices include the growth of mobile phone ownership among the rural population in some developing countries[14]. Increased radio coverage presents further opportunities to deliver much-needed agricultural services to smallholders using digital technology[15].
The call for effective use of digital solutions became even more pronounced during the 2020 global economic shutdown as a result of the COVID-19 pandemic[16,17]. The lockdowns across countries entailed a rise in the use of information systems and networks, with massive changes in usage patterns and behavior[16]. In a study of ICT for improving the investment readiness of small and medium agribusinesses, Valverde[18] reports that 58% of agricultural value-added service providers interviewed reported an increase in demand for their services since the beginning of the COVID-19 crisis, in particular, for their roles in facilitating cash flows and access to credit. Recently, various studies have focused on understanding farmer socio-psychological behavior and institutional services that support access to knowledge and use of improved agricultural technologies in different parts of the world[1922]. However, challenges and capacity gaps in smallholder access and utilization of digital EAS as well as the success of digital EAS in promoting behavioral change among farmers in Kenya and Uganda has yet to be explored. There are also concerns that infrastructure weaknesses (particularly in remote areas), costs of accessing digital services and digital illiteracy of already marginalized groups can exacerbate inequities.
This study aimed to assess challenges and capacity gaps in smallholder access to digital extension and advisory services. Specifically, the objectives were to (1) assess farmer access to extension and advisory services and factors affecting the likelihood of adoption of digital EAS; (2) assess farmer information and advisory services needs that could be met by digital technology; (3) assess barriers and required skills and knowledge for farmers to use digital EAS effectively; and (4) make recommendations on appropriate use of digital EAS by smallholders in Africa. The study is based on primary data gathered through household interviews in Uganda and Kenya.
The results show that over 88% of farmers (96% in Kenya and 79% in Uganda) received extension advice from any source in the previous year mainly from family and friends, the local community, and extension workers. Significantly more households in Kenya (92%) than in Uganda (63%) reported that they received agricultural advice through digital devices dominated by radio. Lack of access to affordable internet services, low digital literacy levels, lack of ownership and control of devices, limited technical support to use digital devices, and low awareness of digital services availability were among the key factors limiting the use of digital EAS.

2 MATERIALS AND METHODS

2.1 Study areas, populations and samples

The study was conducted in Uganda and Kenya. Four local government areas were selected in each country that represent diversity in biophysical characteristics and production activities which may influence farmer agricultural knowledge seeking behaviors (Table 1). In each area the local administration helped to select at least two sub-counties that were considered to have contrasts in terms of ICT infrastructure, signal strength and accessibility (e.g., rural/remote or peri-urban/urban areas). Rural/remote locations were those more than 30 km from the main district/county town and lacking an all-weather road, or as defined by the local administration based on their accessibility indicators. From each of the selected sub-counties at least two villages were selected from where the respondents were drawn. The study population comprised all the farm households in the enumeration area, though particular attention was placed on segmenting responses from contrasting categories of farmers including those known to be already excluded such as female, older and subsistence (as opposed to commercially oriented) farmers. At the end of the exercise a total of 436 households, 228 in Kenya and 208 in Uganda, were interviewed. Table 1 shows the main biophysical characteristics and production activities of the study locations and sample sizes.
Tab.1 Main biophysical characteristics and production activities of study locations and sample sizes
Local government area and enumeration sub-counties Biophysical characteristics Production activities Sampled households
Kenya
 Baringo County   - Eldama Ravin (rural)   - Koibatek (peri-urban) Semiarid, receiving an average of 745 mm of rainfall per year Livestock farming is dominant and crop farming under irrigation schemes 56
 Kirinyaga county   - Kirinyaga East, West (rural)   - Mwea (peri-urban) The annual rainfall is 996 mm Rice production at Mwea irrigation scheme. Coffee and tea grown in the cooler areas 55
 Nakuru county   - Mangu (rural)   - Rongai (peri-urban) The rainfall is around 762 mm per year Main crops include: maize, beans, potato and wheat. Horticultural crops are fruits, vegetables and flowers 64
 Tharaka Nithi county   - Igamba ngombe (rural)   - Tharaka (peri-urban) Rainfall is around 853 mm per year, and poorly distributed on lower areas Cultivation of tea, coffee, maize, cowpeas, pigeon peas, tobacco and other food crops 53
Uganda
 Kiryandongo district   - Kigumba (rural)   - Kiryandongo town council (peri-urban) Average rainfall of 1259 mm with high variability Smallholder agriculture mainly cereal crops and sunflower. About 6.2% of the total farmland is under large scale commercial farming 53
 Luwero district   - Butuntumula (rural)   - Luwero (peri-urban) Average rainfall of 1,270 mm Small to large scale farming but majority are smallholders. Banana-coffee farming system 49
 Lyantonde district   - Mpumude (rural)   - Lyantonde town council (peri-urban) Average rainfall range of 915 mm Mainly smallholders with agro-pastoral practices 54
 Tororo district   - Merikit (rural),   - Tororo Municipality (peri-urban) Average rainfall range of 1215− 1328 mm Small-scale subsistence mainly annual crops 52
Total 436

2.2 Data collection

Data collection was done by structured questionnaire. The questionnaire was coded in Open Data Kit (ODK collect), an open-source Android application, and data collected using tablet computers by trained enumerators. Farmer interviews sought information on the extent of utilization of digital EAS; ownership of digital devices (radio, TV, mobile phone or computer); awareness and access to digital EAS; unmet needs for information, advice and decision support that could be met by digital devices; and barriers to accessing digital EAS.

2.3 Empirical model

Logistic regression analysis was done to identify drivers of digital EAS access. The dependent variable was designated as access to digital EAS, having a value of zero if a farmer did not receive any agricultural advice via digital devices (e.g., radio, SMS, TV and video), and one if a farmer received agricultural advice through any of the devices. However, this did not consider the frequency of receipt of information or whether the messages led to adoption of practices. The following explanatory variables were included in the model taken from the household survey: (1) ownership of digital devices (radio, TV or mobile phone); (2) farm size, viz., areas of land under production and fallow, in hectares, during the October–December 2020 growing season; (3) livestock assets, viz., the count of the number of livestock owned by the household from a list of common animals including cattle, goats and sheep, converted to tropical livestock units[23]; (4) education, viz., highest level of education completed by the respondent with the following categories, none (reference), primary, secondary and tertiary level; (5) location which was coded as zero for rural/remote areas and one for those in peri-urban areas or near townships (as defined in the methodology section); (6) gender, viz., women were the reference category (coded as 0); (7) farm orientation, viz., farmers who sold more than 60% of their farm produce were categorized as commercial and those with lower produce sale as subsistence (coded as 0); (8) age- chronological age of respondent in years; and (9) other household socioeconomic characteristics - age of respondent in years, household size, access to extension services and farmer engages in non-farm production activities. These variables are hypothesized to affect the dependent variable based on empirical evidence from other studies[2426].

2.4 Data analysis

The final data sets from the household survey were downloaded from the Open Data Kit aggregate server as CSV files and exported to STATA 16 software for analysis. Descriptive analysis was done by calculating frequencies, means and percentages to understand farmer awareness and access to digital extension services, information and knowledge gaps, and barriers to utilization of digital EAS by farmers.

3 RESULTS

3.1 Respondent household characteristics

The majority of respondents were male with an average age of 45 years and an average household size of six people (Table 2). At least 55% of the respondents had secondary and tertiary level education. About 4% indicated that they had not received any formal education. At least 77% of the respondents owned a radio (92% in Kenya and 62% in Uganda), 77% owned a feature phone (69% in Kenya and 86% in Uganda), while TV and smartphone ownership was mentioned by 51% and 43%, respectively. Significantly more households in Kenya owned digital devices (radio, TV and smartphone) than in Uganda. In particular, ownership of TVs and smartphones in Uganda was limited at 30% and 19%, respectively.
Tab.2 Socio-economic and demographic characteristics of respondents in Kenya and Uganda
Descriptive Kenya Uganda Overall
Mean Std. Dev. Mean Std. Dev. Mean Std. Dev.
Respondent sex (male = 1) 0.56 0.50 0.54 0.50 0.55 0.50
Respondent age (years) 44.89 15.52 45.04 15.36 44.96 15.43
Household size 5.16 2.14 7.21*** 3.28 6.14 2.93
Location (peri-urban = 1) 0.48 0.50 0.56** 0.50 0.52 0.50
Farm size (ha) 1.19 1.36 2.36*** 5.35 1.75 3.86
Tropical livestock units 2.52 0.48 2.28*** 0.26 2.39 0.27
Education level
  Primary 0.36 0.48 0.47** 0.50 0.41 0.49
  Secondary/vocational 0.41 0.49 0.38 0.49 0.39 0.49
  Tertiary 0.22 0.41 0.09*** 0.29 0.16 0.37
  None 0.02 0.13 0.06** 0.24 0.04 0.19
Ownership of digital devices (yes = 1)
  Radio 0.92 0.28 0.62*** 0.49 0.77 0.42
  TV 0.70 0.46 0.30*** 0.46 0.51 0.50
  Feature phone 0.69 0.46 0.86*** 0.35 0.77 0.42
  Smart phone 0.64 0.48 0.19*** 0.39 0.43 0.50
Farm orientation (commercial = 1) 0.17 0.37 0.05*** 0.21 0.11 0.31
Primary activity
  Farming 0.91 0.29 0.90 0.30 0.90 0.30
  Business 0.02 0.13 0.02 0.15 0.02 0.14
  Salaried employment 0.05 0.21 0.06 0.24 0.06 0.23
  Other 0.03 0.16 0.01 0.12 0.02 0.12
Agribusiness activities (yes = 1)
  Produce aggregation/transportation 0.43 0.47 0.10*** 0.23 0.29 0.40
  Produce selling 0.78 0.42 0.25*** 0.43 0.52 0.50
  Value addition and processing 0.03 0.16 0.04 0.20 0.03 0.18
  Service delivery 0.06 0.23 0.07 0.25 0.06 0.24
  Input sales 0.03 0.17 0.00** 0.00 0.02 0.13

Note: ***, **, indicate 1% and 5% levels of significance, respectively, between Kenya and Uganda.

The majority of farmers (90%) engaged in farm production and were categorized as subsistence (> 60% of food grown for home consumption). Significantly more farmers in Kenya than Uganda were categorized as commercial despite the fact that they had significantly smaller farm sizes compared to Uganda. Maize and beans were mostly grown in the sample counties in Kenya, while crop farming in Uganda was more diversified with maize, sorghum, banana, beans and coffee mentioned as key food and cash crops. Livestock production was represented in small proportions averaging 2.4 tropical livestock units. Poultry was the most common (81% in Kenya and 74% in Uganda). Cattle farming, especially dairy cattle, was more pronounced in Kenya than in Uganda, while small ruminants were represented in similar proportions across Kenya and Uganda (44% in Kenya and 43% in Uganda). Produce selling (78% in Kenya and 25% in Uganda), produce transportation (32% in Kenya and 6% in Uganda) and produce aggregation (11% in Kenya and 10% in Uganda) were the most commonly mentioned off-farm agribusiness activities. Significantly more farmers in Kenya than Uganda engaged in produce aggregation and selling, and transportation.

3.2 Farmer access to extension and advisory services

Farmers were asked if they had received extension advice in the last 12 months considering both non-digital and digital approaches (Table 3). Over 88% (96% in Kenya and 79% in Uganda) received extension advice from any source, while 12% did not receive any advice irrespective of the source. At least 67% (75% in Kenya and 60% in Uganda) received extension from conventional extension approaches dominated by family and friends, local community, extension workers or plant doctors, agricultural input dealers and farmer groups/cooperatives. There were significant differences in farmer access to advice from extension services, agricultural input dealers and places of worship between Kenya and Uganda. A higher proportion of farmers in Uganda mentioned receiving information from extension workers compared to Kenya, while the reverse was true for Kenya in terms of accessing information from agricultural input dealers and through places of worship.
Tab.3 Sources of agricultural advice as mentioned by farmers in Kenya and Uganda (%)
Variable Kenya Uganda Total
Farmer accessed extension services 96 79*** 88
Farmer did not access any extension advice 4 21*** 12
Source of extension advice
  Digital extension 92 63*** 78
  Conventional extension 75 60*** 67
  Both digital and conventional 96 79*** 88
Conventional extension
  Friends/family 52 48 50
  Local community 41 36 39
  Extension 26 54*** 38
  Agricultural input dealer 43 23*** 35
  Farmer cooperative 16 20 18
  Worship places 10 2** 6
  Print materials 4 4 4
Digital extension and devices
  Radio 84 76*** 80
  Television 58 36*** 47
  Smartphone 23 5*** 14
  Feature phone 9 5 7
  Computer 5 1** 3
  Community radio 1 3 2

Note: ***, **, indicate 1% and 5% levels of significance, respectively, between Uganda and Kenya.

At least 78% (92% in Kenya and 63% in Uganda) reported that they also accessed agricultural advice through digital devices. Radio was the prevalent digital platform used by farmers to access agricultural advice in both Kenya (84%) and Uganda (76%), followed by television (58% in Kenya and 36% in Uganda) and mobile phones (32% in Kenya and 10% in Uganda) (Table 3). Use of computers and community radio were less popular and were reported by 5% and 1% in Kenya and 1% and 3% in Uganda, respectively.
Female and male farmer access to digital EAS differed between the two countries (Table 4). In Kenya there was no significant difference in the proportion of female (90%) and male farmers (93%) receiving agricultural advice via digital devices. In Uganda there was a significant difference between female (46%) and male (77%) farmers receiving information via digital devices. In terms of location of farmer, the data did not reveal significant differences in access to information via digital devices by farmers in very remote areas and those close to towns. Male farmers, irrespective of location, were more likely to access digital EAS than their female counterparts. When subdivided by digital device used, significantly more male farmers than female farmers in Kenya were likely to use TV and mobile phone to received agricultural advice, while use of radio was not significantly different across gender (Table 4).
Tab.4 Access to digital EAS by female and male farmers by country and location (%)
Variable Kenya Uganda Remote areas Townships/peri-urban
Female Male Female Male Female Male Female Male
Access to digital EAS 90 93 46 77*** 69 85*** 76 91***
Digital devices used to access EAS
  TV 46 60** 20 25 31 40 42 55*
  Radio 76 78 32 61*** 56 72** 58 71**
  Mobile phone 22 35** 2 9** 14 22 13 30***

Note: ***, **, indicate 1% and 5% levels of significance, respectively, between male and female farmers by country and location.

Farmers receiving information through digital devices indicated that they received information mainly on crop/livestock pest and disease management, weather information, markets for inputs and products and market price information. The trend of information flow was very similar across Kenya and Uganda, with a focus on advisory. These messages are simple and can be easily transmitted through digital devices (Table 5; Table 6).
Tab.5 Information accessed from digital devices by farmers in Kenya (%)
Information type TV Radio Feature phone Smart phone
Managing crop pests and diseases 59 54 17 47
Managing livestock vectors and diseases 57 57 39 45
Weather information 53 48 28 30
Livestock production 36 31 0 9
Where to buy seed, fertilizers, pesticides etc. 29 37 11 30
What type of seed to use 27 34 17 28
Market pricing information 22 21 22 17
Crop agronomy (GAPs) 19 17 6 6
Credit services 9 15 0 6
Alerts on agricultural activities e.g., time of planting 8 12 17 11
Purchase and sale of produce 8 6 6 17
Processing and value addition 2 2 0 6
Tab.6 Information accessed from digital devices by farmers in Uganda (%)
Information type TV Radio Feature phone Smart phone
Managing crop pests and diseases 72 63 40 50
Managing livestock pests and diseases 46 27 20 67
What type of seed to use 46 42 40 0
Alerts on agricultural activities e.g., time of planting 41 38 20 0
Crop agronomy (GAPs) 35 22 60 0
Livestock production 33 18 20 33
Where to buy seed, fertilizers, pesticides etc. 24 42 0 50
Weather information 22 43 0 17
Market pricing information 20 20 0 33
Processing and value addition 13 8 20 0
Purchase and sale of produce 11 9 0 0
Credit services 2 2 0 0
At least 65%, 59% and 47% of the farmers using radio, TV and mobile phone, respectively, rated them better than conventional face-to-face extension methods. The main reasons given for their rating were: ease of access, ease of understanding the advice, information is considered relevant to farming activities, information is of good quality and is time-saving. However, those who did not use any digital devices indicated that they were not aware of the digital services (52%) and as such relied on other sources of information (Table 7). Other reasons given were that services are expensive, lack of knowledge on how to use the services and lack of ownership of digital devices. There were also issues related to ability to read and comprehend the messages sent especially via mobile phones, language barriers and information relevance. These challenges were more pronounced in Uganda than in Kenya.
Tab.7 Reasons for not using digital extension devices (%)
Reason Kenya Uganda Total
I am not aware of these services 21 59 52
The services are too expensive 37 42 41
I do not know how to use these types of services 16 44 38
I do not own a phone/radio to access these services 16 42 37
These types of services are not available in my area 16 24 23
I do not have the time to use them 16 23 22
I have trouble reading the content 26 18 20
The services are not available on my phone network 11 18 16
There is no network coverage in my area 0 17 13
The content is not in a language I understand 11 10 10
The content is not relevant to me 16 1 4

3.3 Logistic regression results

We used a binary logistic generalized linear model to test the likelihood that a respondent adopts digital EAS. We estimated the odds ratio the dichotomous dependent variable: access to information via digital devices (Table 8). Ownership of a radio, TV or mobile phone was associated with a higher likelihood of accessing digital EAS in Kenya. In fact, radio, TV or phone ownership was associated with 23, 58 and seven times as likely as non-owners to use digital EAS. In Uganda, only TV ownership was significantly associated with the likelihood of using digital EAS, almost six times more likely than non-owners. Farmer engagement in non-farm production activities (post-production), e.g., transportation and service delivery, was associated with a higher likelihood of adopting digital EAS compared to farmers engaged only in farm production activities in Kenya. In both countries, access to extension services was associated with a high likelihood of adopting EAS. The coefficient on being a male respondent was positive in both Kenya and Uganda but showed significant effects in Uganda. Males were 3.2 times more likely to use digital EAS in Uganda than their female counterparts.
Tab.8 Logistic regression results of access to digital EAS by farmers
Explanatory variable Kenya Uganda
Odds ratio Std. Err. Odds ratio Std. Err.
Location (remote = 0) 2.06 1.60 0.99 0.40
Respondent gender (female = 0) 0.60 0.53 3.19** 1.42
Respondent age (chronological are in years) 1.00 0.03 0.98 0.01
Education level: primary 132** 321 0.64 0.48
Education level: secondary/vocational 11.53 27.11 1.21 1.00
Education level: tertiary 3.73 8.94 1.50 1.88
Household size (# of household members) 0.95 0.24 1.08 0.08
Radio ownership (yes = 1) 5.34 62.9 8.55 3.68
TV ownership (yes = 1) 23.1*** 23.4 5.55*** 3.10
Feature phone ownership (yes = 1) 58.4** 4.79 0.72 0.44
Smart phone ownership (yes = 1) 7.26** 7.44 1.48 1.00
Farm size (hectares) 0.98 0.13 1.01 0.04
Tropical livestock units 1.25 0.34 0.94 0.05
Farm orientation (commercial = 1) 0.92 0.90 1.00
Farmer engages in non-farm production (yes = 1) 0.06** 0.08 1.16 0.57
Extension service access (other sources) 7.05** 5.76 2.48** 1.04
Constant 0.01** 0.02 0.22** 0.31
Observations 228 208
Chi-square 67.03 91.55
Probability 0.000 0.000
Pseudo R 2 0.512 0.344

Note: ***, **, indicate 1% and 5% levels of significance, respectively.

3.4 Farmer information and advisory service needs that could be met by digital technology

Respondents were asked the type of information and advice they would need but found difficult to obtain. The highest proportion of respondents (46% in Kenya and 25% in Uganda) mentioned crop pest and disease management (Fig. 1). Farmers indicated that the most commonly shared information was in crop/livestock pest and disease management but they also expressed information gaps especially in pest/disease identification, prevention, control practices and products. In particular, farmers require information on how to identify and diagnose pests and diseases, how to distinguish diseases with similar symptoms, how to use biological pest and disease control methods, and the appropriate stage at which to control pests and diseases. Similarly, farmers expressed information gaps in recommended pesticides, when and how to spray, where to get quality pesticides, and safe handling of pesticides. In terms of livestock production, the knowledge gaps were reported on general animal husbandry, diagnosis of livestock diseases, feeding dairy cattle, proper breeding, recommended vaccines (especially for poultry), control of livestock diseases, and how to maximize profits from livestock production. Other reported information gaps were as follows.
Fig.1 Information that is difficult for smallholders to access in Kenya and Uganda.

Full size|PPT slide

• Markets: information on market prices, how to access markets and where to get good markets.
• Fertilizer use: safe use of fertilizers, where to get affordable fertilizers, application rates, effective/recommended fertilizers, how to obtain subsidized fertilizers, how to make organic fertilizers and the types of fertilizer to use on different crops.
• Credit facilities: how/where to access credit facilities.
• Quality seed: best type of seed to grow in their area, where to obtain quality/certified seed, information on quality/certified seed, how to distinguish quality seed from ‘fake’ seed and best cultivars to plant.
• Soil fertility management: soil pH testing services, how to increase soil fertility, best soils for different crops and advice on soil conservation.
• Value addition: how to do value addition and processing of milk and fruits and postharvest storage.
• Water management: who can help in installation of piped water and how to deal with too much rainwater
• General information on farming: landscaping, how to practice crop rotation, how to do organic farming, increasing production, spacing, how to increase yields, and weather.
Factors that contribute to difficulty in obtaining the information needed by smallholders included: limited access to extension services, and limited farmer resources to proactively search for this information, e.g., when travel is required. Farmers also reported that government extension officers are few and not easy to access, while private ones are expensive to hire. There were also perceptions that the available cadre of government extension officers are not interested/willing to train farmers, do not have adequate technical skills and are unable to help farmers, especially during pest/disease outbreaks. However, the study noted current efforts in Uganda to recruit more extension workers to bridge this gap. Farmers also mentioned that the recommendations extension officers give to them sometimes do not work and extension worker have preference for large scale farmers who can pay for their services. Other challenges mentioned by farmers include a lack of resources to acquire recommended inputs, poor availability of inputs such as quality seed and lack of access to soil testing services.

3.5 Barriers and required skills and knowledge for farmers to use digital EAS effectively

Smallholders faced various barriers in accessing digital devices and this differed by gender and age category of farmers. Sixty-one percent of respondents (43% in Kenya and 81% in Uganda) indicated that older farmers faced barriers in accessing mobile phones, followed by female farmers: 42% of respondents (21% in Kenya and 66% in Uganda). The main reasons given for limited access to mobile phone services included: low literacy levels, lack of ownership and control of digital devices, calling prohibitively expensive, and subscription fees for some services unacceptably high. The difference between Kenya and Uganda was statistically significant (P < 0.001). In particular respondents in Kenya were more likely to mention higher calling and subscription rates than those in Uganda, while those in Uganda were more likely to report not owning digital devices.
For radio access the majority of respondents did not perceive barriers across gender and age category. However, 37% of respondents (18% in Kenya and 58% in Uganda) indicated that women faced barriers in accessing radio compared to men 33% (12% in Kenya and 56% in Uganda). The difference between the two countries was statistically significant (P < 0.001) as more farmers in Uganda were more likely to report facing barriers with respect to accessing radio than those in Kenya. The main reasons given were a lack of time to tune into suitable programs, timing of programs coinciding with other activities, lack of ownership and control of digital devices. Data also show significant differences with respect to specific barriers faced between farmers in Kenya and Uganda. Farmers in Uganda were more likely to report lack of ownership of digital devices (80%) compared to Kenya (28%) while more farmers in Kenya were more likely to report lack of time (73%) compared to Uganda (40%). For video screenings and computer access, all categories of farmers faced barriers, particularly attributed to low literacy levels, lack of ownership of digital devices and lack of awareness. This is also related to the fact that those who actually used or accessed these sources of information were relatively few compared to other means, namely radio, TV and phone in particular.
Based on the barriers faced, more than 50% of the respondents mentioned they would like skills on how to access digital devices, how to use these devices to access the agricultural information of interest to them, and how to subscribe to receive information on SMS, including subscription codes. They would also like access to hotlines and the internet, as well as knowledge on applications that give farming information and training. A few farmers mentioned that they needed information on where to access affordable digital devices, awareness on which radio or TV stations run programs on agriculture and the timing of these programs. They would also like the information to be communicated in local languages and via multiple means.

4 DISCUSSION

This study shows that farmers accessed extension advice from various sources including both traditional and digital ones. Family and friends, extension workers, local community/neighbors, agribusinesses, and farmer groups were important face-to-face sources of information, while radio was the most prevalent digital device used by farmers to access agricultural advice, followed by television and mobile phones. Dominance of radio as the main mass media source may be attributed to the fact that ownership of radios was more widespread than other ICT devices such as phones. Other factors mentioned by farmers as limiting the use of digital devices in this study include: lack of ownership and control of digital devices, limited technical support to use digital platforms/devices, and low awareness of digital services availability. Female and elderly farmers were more likely to report these barriers than men and younger people. This is consistent with Aker at al.[27]. who note that access to mobile phones, as well as other ICT, is often unequally distributed, which may aggravate information asymmetries between men and women, or older and younger farmers. Kansiime et al.[28] also noted a digital divide in access to extension services in Tanzania as men were more familiar with digital approaches such as radio, than women. The differences between female and male access to digital EAS are related to differences in socioeconomic and cultural factors that may affect ownership of ICT devices or participation in extension activities. In addition, lower levels of literacy, household duties and workloads, social norms and limited disposable income, all intersect to reduce female participation in extension programs but more so mobile phone or mobile internet use for agricultural advice[8].
In relation to specific digital devices, several overarching challenges were found. For example, while radio was the most prevalent source of digital EAS, farmers reported a lack of time to tune into aired programs which potentially limits the use of radio or TV to access extension services. This limitation has also been highlighted in other studies attributed to the fact that programs on radio or TV are often aired at the time when farmers are busy with farm work or household chores limiting their participation in such programs[27,29]. For mobile phones, lack of ownership of digital devices, high cost of internet access and low digital literacy were the main challenges. Consequently, the use of different but linked communication channels, including digital and analog ones would ensure more farmers are reached in a way that maximizes accountability and increases impact[12,19].
The logistic regression model results show that ownership of digital devices (radio, TV or mobile phone), farmer engagement in post-production activities, and access to extension services were associated with a high likelihood of using digital EAS. The correlation between ownership of digital devices and likelihood of accessing digital EAS is fairly obvious, as ownership of the digital devices facilitates easy access to information including agricultural advice, and ensures connectivity with extension and fellow farmers[24]. Post-production activities included service delivery, produce marketing and value addition. It is assumed that farmers engaged in such activities are more exposed and more likely to know and use digital devices to access advice. Farmers who have access to extension services know the value and are always seeking better ways to remain connected to technical personnel, and are thus more likely to adopt digital EAS devices. Contrary to expectation, farmer location (whether a farmer is in a remote or peri-urban area) did not show significant effects on the likelihood of them using digital EAS, and was also not mentioned by farmers as a key barrier. This may imply that the study areas were fairly equally connected to radio or TV, the most commonly mentioned digital devices, or it may reflect the general lack of awareness and access to services that have limited location bearing.
Though the relative importance of and demand for different types of information varied across the farmers surveyed, there was a consistent demand for information on crop pest and disease diagnosis and management, types of fertilizers for a given soil type, pesticide use and safety, type of seed for a given agroecology, new crop cultivars, credit and market information, weather advisories and livestock production (pest control products, pest and disease management, breeding). This is consistent with literature on farmer information needs[30,31]. These information types also represent areas where information needs to be context-specific to support farmer decision-making. Participatory design methodologies need to be used that consider new insights about local information needs, user preferences and capacities[32]. This process ensures that both the digital platform and content are adapted for end-users that are often of different genders, ages, wealth groups, literacy, languages and agroecological zones.

5 CONCLUSION AND RECOMMENDATIONS

We investigated challenges and capacity gaps in smallholder access to digital EAS using household survey data from Kenya and Uganda. Farmers accessed EAS from various long-established extension services and digital sources dominated by radio. Dominance of radio as the main mass media source is attributed to the majority of the respondents owning a radio in comparison to other digital devices. In relation to specific digital devices, several overarching challenges were found. Lack of time to listen to aired programs was the major limitation for accessing EAS on radio or TV, while low or lack of ownership of digital devices, high cost of internet access, and low digital literacy were the key challenges for use of digital services in general. Our results further highlight that ownership of digital devices, participation in post-production activities, and access to extension are key drivers of digital EAS use. These factors reflect differences in smallholder technological capabilities, farming objectives and exposure, which should be taken into consideration in the design of innovations to aid appropriate use of digital EAS by farmers. Across farmers surveyed there was a consistent demand for information on crop pest and disease diagnosis and management, fertilizer application, pesticide use and safety, quality seed, new cultivars, market information, weather advisories and livestock production. These subjects represent areas where farmers need to make decisions based on agroecology and farmer asset base. We therefore make the following recommendations for policy and practice.
Recommendations for policy: (1) farmer profiling to understand the different needs of smallholders to provide targeted information and advisory services; (2) enhancing farmer digital innovation capacity and creating farmer awareness of available digital EAS to help agricultural extension services to tap the full potential of the digital revolution; and (3) enhancing physical infrastructure development for digital access and reducing costs associated with access to internet and digital devices to enhance inclusion by smallholders.
Recommendations for practice: (1) continued development, testing and evaluation of ICT for various farmer categories and suitability to pass on information on promoted technologies; (2) integration of digital communication within multimode advisory services that use different but linked communication channels, for inclusive scaling of extension activities; (3) inclusion of bundled agricultural production services (e.g., insurance, credit and inputs) in digital EAS delivery to inspire participation of smallholders; and (4) content development that addresses farmer-identified information needs, and which is adaptable to various digital devices to enhance dissemination.

References

[1]
Zhuang Y H, Ruan S H, Zhang L, Chen J R, Li S S, Wen W J, Liu H B . Effects and potential of optimized fertilization practices for rice production in China. Agronomy for Sustainable Development, 2022, 42(2): 32
CrossRef Google scholar
[2]
Sen S, Chakraborty R, Kalita P . Rice—not just a staple food: a comprehensive review on its phytochemicals and therapeutic potential. Trends in Food Science & Technology, 2020, 97: 265–285
CrossRef Google scholar
[3]
Younas M U, Wang G D, Du H B, Zhang Y, Ahmad I, Rajput N, Li M Y, Feng Z M, Hu K M, Khan N U, Xie W Y, Qasim M, Chen Z X, Zuo S M . Approaches to reduce rice blast disease using knowledge from host resistance and pathogen pathogenicity. International Journal of Molecular Sciences, 2023, 24(5): 4985
CrossRef Google scholar
[4]
Dean R, Van Kan J A L, Pretorius Z A, Hammond-Kosack K, Di Pietro A, Spanu P D, Rudd J J, Dickman M, Kahmann R, Ellis J, Foster G D . The top 10 fungal pathogens in molecular plant pathology. Molecular Plant Pathology, 2012, 13(4): 414–430
CrossRef Google scholar
[5]
Tao H L, Feng H K, Xu L J, Miao M K, Long H L, Yue J B, Li Z H, Yang G J, Yang X D, Fan L L . Estimation of crop growth parameters using UAV-based hyperspectral remote sensing data. Sensors, 2020, 20(5): 1296
CrossRef Google scholar
[6]
Agilandeeswari L, Prabukumar M, Radhesyam V, Phaneendra K L N B, Farhan A . Crop classification for agricultural applications in hyperspectral remote sensing images. Applied Sciences, 2022, 12(3): 1670
CrossRef Google scholar
[7]
Marang I J, Filippi P, Weaver T B, Evans B J, Whelan B M, Bishop T F A, Murad M O F, Al-Shammari D, Roth G . Machine learning optimised hyperspectral remote sensing retrieves cotton nitrogen status. Remote Sensing, 2021, 13(8): 1428
CrossRef Google scholar
[8]
Deng X L, Zhu Z H, Yang J C, Zheng Z, Huang Z X, Yin X B, Wei S J, Lan Y B . Detection of Citrus Huanglongbing based on multi-input neural network model of UAV hyperspectral remote sensing. Remote Sensing, 2020, 12(17): 2678
CrossRef Google scholar
[9]
Zhang N, Yang G J, Pan Y C, Yang X D, Chen L P, Zhao C J . A review of advanced technologies and development for hyperspectral-based plant disease detection in the past three decades. Remote Sensing, 2020, 12(19): 3188
CrossRef Google scholar
[10]
Gao Z M, Khot L R, Naidu R A, Zhang Q . Early detection of grapevine leafroll disease in a red-berried wine grape cultivar using hyperspectral imaging. Computers and Electronics in Agriculture, 2020, 179: 105807
CrossRef Google scholar
[11]
Li N W, Huo L, Zhang X L . Classification of pine wilt disease at different infection stages by diagnostic hyperspectral bands. Ecological Indicators, 2022, 142: 109198
CrossRef Google scholar
[12]
Zhao X H, Zhang J C, Huang Y B, Tian Y Y, Yuan L . Detection and discrimination of disease and insect stress of tea plants using hyperspectral imaging combined with wavelet analysis. Computers and Electronics in Agriculture, 2022, 193: 106717
CrossRef Google scholar
[13]
Randive P U, Deshmukh R R, Janse P V, Gupta R S. Discrimination Between Healthy and Diseased Cotton Plant by Using Hyperspectral Reflectance Data. In: Recent Trends in Image Processing and Pattern Recognition. Singapore: Springer, 2019, 342–351
[14]
Shukla K S, Birah A, Nigam R, Kanojia A K, Km Khokhar M, Bhattacharya B K, Chander S . Selection of sensitive bands for assessing Alernaria blight diseased severity grades in mustard crops using hyperspectral reflectance. Journal of Agrometeorology, 2023, 25(2): 274–279
CrossRef Google scholar
[15]
Zhang N N, Zhang X, Shang P, Ma R, Yuan X T, Li L, Bai T C . Detection of cotton Verticillium wilt disease severity based on hyperspectrum and GWO-SVM. Remote Sensing, 2023, 15(13): 3373
CrossRef Google scholar
[16]
Khanal S, Kc K, Fulton J P, Shearer S, Ozkan E . Remote sensing in agriculture—Accomplishments, limitations, and opportunities. Remote Sensing, 2020, 12(22): 3783
CrossRef Google scholar
[17]
Raj R, Kar S, Nandan R, Jagarlapudi A. Precision agriculture and unmanned aerial vehicles (UAVs). In: Avtar R, Watanabe T, eds. Unmanned Aerial Vehicle: Applications in Agriculture and Environment. Springer, 2020, 7–23
[18]
Ma H Q, Huang W J, Dong Y Y, Liu L Y, Guo A T . Using UAV-based hyperspectral imagery to detect winter wheat fusarium head blight. Remote Sensing, 2021, 13(15): 3024
CrossRef Google scholar
[19]
Moriya É A S, Imai N N, Tommaselli A M G, Berveglieri A, Santos G H, Soares M A, Marino M, Reis T T . Detection and mapping of trees infected with citrus gummosis using UAV hyperspectral data. Computers and Electronics in Agriculture, 2021, 188: 106298
CrossRef Google scholar
[20]
Zheng Q, Huang W J, Ye H C, Dong Y Y, Shi Y, Chen S S . Using continous wavelet analysis for monitoring wheat yellow rust in different infestation stages based on unmanned aerial vehicle hyperspectral images. Applied Optics, 2020, 59(26): 8003–8013
CrossRef Google scholar
[21]
Abdulridha J, Min A, Rouse M N, Kianian S, Isler V, Yang C . Evaluation of stem rust disease in wheat fields by drone hyperspectral imaging. Sensors, 2023, 23(8): 4154
CrossRef Google scholar
[22]
Liu T, Shi T Z, Zhang H, Wu C . Detection of rise damage by leaf folder (Cnaphalocrocis medinalis) using unmanned aerial vehicle based hyperspectral data. Sustainability, 2020, 12(22): 9343
CrossRef Google scholar
[23]
Ashourloo D, Mobasheri M, Huete A . Evaluating the effect of different wheat rust disease symptoms on vegetation indices using hyperspectral measurements. Remote Sensing, 2014, 6(6): 5107–5123
CrossRef Google scholar
[24]
Heim R H J, Wright I J, Allen A P, Geedicke I, Oldeland J . Developing a spectral disease index for myrtle rust (Austropuccinia psidii). Plant Pathology, 2019, 68(4): 738–745
CrossRef Google scholar
[25]
Kira K, Rendell L A. The Feature Selection Problem: Traditional Methods and a New Algorithm. In: Proceedings of the Tenth National Conference on Artificial Intelligence—AAAI’92. San Jose, California: AAAI Press, 1992, 129–134
[26]
Huang W J, Guan Q S, Luo J H, Zhang J C, Zhao J J, Liang D, Huang L S, Zhang D Y . New optimized spectral indices for identifying and monitoring winter wheat diseases. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2014, 7(6): 2516–2524
CrossRef Google scholar
[27]
Mandal N, Das D K, Sahoo R N, Adak S, Kumar A, Viswanathan C, Mukherjee J, Rajashekara H, Ranjan R, Das B . Assessing rice blast disease severity through hyperspectral remote sensing. Journal of Agrometeorology, 2022, 24(3): 241–248
CrossRef Google scholar
[28]
Zheng Q, Huang W J, Xia Q, Dong Y Y, Ye H C, Jiang H, Chen S S, Huang S Y . Remote sensing monitoring of rice diseases and pests from different data sources: a review. Agronomy, 2023, 13(7): 1851
CrossRef Google scholar
[29]
Wan L, Li H, Li C S, Wang A C, Yang Y H, Wang P . Hyperspectral sensing of plant diseases: principle and methods. Agronomy, 2022, 12(6): 1451
CrossRef Google scholar
[30]
Zheng Q, Huang W J, Cui X M, Shi Y, Liu L Y . New spectral index for detecting wheat yellow rust using sentinel-2 multispectral imagery. Sensors, 2018, 18(3): 868
CrossRef Google scholar
[31]
Kuncheva L I, Whitaker C J . Measures of diversity in classifier ensembles and their relationship with the ensemble accuracy. Machine Learning, 2003, 51(2): 181–207
CrossRef Google scholar
[32]
Cover T, Hart P . Nearest neighbor pattern classification. IEEE Transactions on Information Theory, 1967, 13(1): 21–27
CrossRef Google scholar
[33]
Caruana R, Niculescu-Mizil A. An Empirical Comparison of Supervised Learning Algorithms. In: Proceedings of the 23rd International Conference on Machine Learning—ICML’06. Pittsburgh, Pennsylvania: ACM Press, 2006, 161–168
[34]
Altman N S . An Introduction to Kernel and nearest-neighbor nonparametric regression. American Statistician, 1992, 46(3): 175–185
CrossRef Google scholar
[35]
Biau G, Scornet E . A random forest guided tour. Test, 2016, 25(2): 197–227
CrossRef Google scholar

Compliance with ethics guidelines

Guangming Li, Dongxue Zhao, Jinpeng Li, Shuai Feng, and Chunling Chen declare that they have no conflicts of interest or financial conflicts to disclose. This article does not contain any studies with human or animal subjects performed by any of the authors.

RIGHTS & PERMISSIONS

The Author(s) 2024. Published by Higher Education Press. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0)
AI Summary AI Mindmap
PDF(1570 KB)

Accesses

Citations

Detail

Sections
Recommended

/