The study selected Vihiga and Kilifi counties based on prior research assessing geospatial data usage among small-scale farmers in four counties (Nyeri, Vihiga, Kilifi, Wajir) [12] [13] . Wajir was excluded due to security risks, sparse agriculture, and inconsistent farming practices. Nyeri had some active crop production but limited geospatial data adoption by local agriculture officers, who lacked advanced skills in data interpretation. In contrast, Vihiga and Kilifi counties offered diverse farming practices and greater experience in using geospatial data, making them more suitable for the study.

The study selected two sub-counties from each of the two counties. In Vihiga County, Vihiga served as the experimental sub-county and Emuhaya as the control. In Kilifi County, Kaloleni was the experimental sub-county, while Magarini was the control as shown in Figure 2 .

A participatory process was conducted with the farmers, moderated by the sub-County agricultural officer to identify crop production indicators for use in the experimental survey in March 2023. Plate 1 shows a photo of farmers in one of the discussion groups in Vihiga County. The identified indicators were subsequently used in the development of a questionnaire which was used for the baseline survey.

Farmers measured productivity mainly by harvest quantity and quality. Other key indicators included reduced losses from pests, disease, or post-harvest damage, and income from the harvest. These metrics align with those used by organizations like FAO, USAID, and Feed the Future, which assess productivity through average yield per unit of land, crop quantity per household, and production value [15] - [17] .

In group discussions, farmers shared their preferred types of information and the frequency of updates. They also identified specific networks for information sharing, including champion farmers, local radio stations, and local administrators. The data and information identified were then shared with farmers during the two seasons in 2023.

a) Farmer characteristics

i. Age groups

In Emuhaya, the largest group of farmers (37%) were in the 26 - 35 age range, while in Vihiga, the highest proportion (38%) were in the 36 - 45 age group. In Kilifi, Kaloleni had 34% of farmers in the 36 - 45 age range, and Magarini had 28% in the 26 - 35 age group.

ii. Sex

The highest proportion of farmers in all the 4 sub-counties were male; Kaloleni had 62% male farmers, Magarini 53%, Emuhaya 59% and Vihiga. 57%.

iii. Level of education

Vihiga sub-County had the highest proportion of farmers with secondary education (48%), while the other three sub-counties had the highest proportion with primary education: Emuhaya (54%), Kaloleni (44%), and Magarini (47%).

iv. Size of farms

In all the four sub-Counties, the farm size of most farmers was less than 5 acres (Magarini 97%, Kaloleni 89%, Emuhaya 100% and Vihiga 77%).

b) Advisory information

In Vihiga sub-County, 75% of farmers received advisory information during the 2023 OND season, up from 60% in 2022. In Kaloleni sub-County, 42% received advisories in 2023, down from 68% in 2022. In the control sub-counties, 42% of farmers in Emuhaya received advisories in 2023, compared to 48% in 2022, while in Magarini, 31% received advisories in 2023, down from 41% in 2022. Both Magarini and Kaloleni experienced notable decreases.

Most farmers who received advisory information felt it was adequate, timely, and accurate, except in Kaloleni, where only 7% found it accurate. The results are shown in Figure 10 .

i. The type of information received by farmers

In the OND 2022 season, the top three types of information received by farmers were rainfall, crop suitability, and pest control as shown in Figure 11 . Vihiga sub-County had a higher proportion of farmers receiving soil information compared to the other sub-counties. While soil information was shared with the SCAO in Kaloleni, it was not effectively disseminated to farmers due to challenges in movement and coordination by WAOs.

ii. Sources of information

In the OND 2022 season, the most common information sources for farmers were fellow farmers, Radio/TV, and friends in Kaloleni. In Vihiga, farmers primarily relied on Radio/TV, fellow farmers, and County extension officers. Magarini’s top sources were fellow farmers, County extension officers, and Radio/TV, while Emuhaya’s were fellow farmers, County extension officers, and friends. KMD was not a popular source in any sub-county, and Kaloleni had a notably low proportion of farmers receiving information from extension officers compared to the other sub-counties ( Figure 12 ).

In the OND 2023 season, there was an increase in farmers receiving information from friends in Vihiga and Emuhaya sub-counties, and a sharp rise in those getting information from extension officers in Magarini. In Vihiga, the proportion of farmers receiving information from KMD increased from under 10% to nearly 50%, while those receiving information from extension officers rose from 35% to 50%. In Kaloleni, fellow farmers and Radio/TV remained the most popular sources, with over 90% of farmers using them. Emuhaya saw more than 50% of farmers getting information from friends, fellow farmers, Radio/TV, and extension officers. A notable shift in Vihiga was the rise in digital media use, increasing from 6% to 64%.

iii. Format used to share the information

In the OND 2023 season, the most popular format for sharing information with farmers was direct, verbal communication, used by over 90% of farmers in all four sub-counties ( Figure 13 ). Printed materials and soft copy materials, like brochures and bulletins, were used by fewer farmers (10% or less), with Vihiga having the highest usage at 35%. Radio/TV was widely used across all sub-counties, and digital media ranked third in usage.

iv. Challenges using information provided

The 2023 survey showed that a high proportion of farmers across all four sub-counties reported no challenges in using the provided information ( Figure 14 ). Few farmers mentioned language as a barrier, suggesting that efforts to share information effectively, such as verbal and direct communication, along with trusted sources like fellow farmers, friends, and Radio/TV, were successful.

c) Crops grown in the two study seasons and the outcomes of the harvests

Maize was grown by over 95% of farmers in all four sub-counties. In Vihiga, beans (94%) and bananas (31%) were also commonly grown. In Kaloleni, cowpeas were grown by all farmers, and cassava by 83%. Most farmers in all sub-counties grow crops for subsistence, with Vihiga at 82%, Emuhaya at 66%, Kaloleni at 94%, and Magarini at 66%.

From 2022 to 2023, Vihiga saw a significant increase in the proportion of farmers reporting a good harvest ( Figure 15 ). In Emuhaya, there was a decrease in those reporting a poor harvest and an increase in those reporting good and average harvests. In Kaloleni, the proportion of farmers reporting a good harvest decreased, while those reporting average and poor harvests increased. In Magarini, the number of farmers reporting a good harvest increased, while those reporting average and poor harvests decreased.

Farmers in all four sub-counties primarily assessed their harvests based on the quantity and quality of the crops ( Table 5 ). However, in Emuhaya, there were significant post-harvest losses in 2022, with 84% of farmers reporting a poor harvest for that season.

In Vihiga sub-County, the increase in good harvests was largely attributed to favorable weather, the advice and information provided, and the cultivation of suitable crops—information that was actively shared with farmers. While all sub-counties experienced enhanced rainfall, some poor harvests were linked to flooding, which damaged leguminous crops like cowpeas and green grams. The factors farmers attributed to their harvests are shown in Table 6 .

In 2023, Vihiga and Emuhaya had a high proportion of farmers who harvested their crops (over 90%) for each of the three commonly grown crops. However, Kaloleni and Magarini had lower proportions of successful harvests. For maize, which was grown in all sub-counties, Vihiga had 67% of farmers reporting a good harvest, while 24% had an average harvest. In Emuhaya, 19% had a good harvest, with 73% having an average harvest. Kaloleni, impacted by extreme rainfall, had only 4% with a good harvest, 50% with an average harvest, and 46% with a poor harvest. Magarini had 54% with a good harvest and 32% with an average harvest. Similar trends were observed for other crops in the sub-counties, as shown in Table 7 and Table 8 .

Analysis of farmers who received advisory information during the OND 2023 season and reported a good harvest showed that those who received advisories generally had better harvests. The results, as shown in Figure 16 , indicated higher proportions of farmers with good harvests in Vihiga and Kaloleni compared to Emuhaya and Magarini.

a) Weather conditions

The seasonal forecasts for both Kilifi and Vihiga for the short rains season showed enhanced and well-distributed rainfall across the four sub-counties. In Vihiga, the forecast predicted above-normal rainfall, exceeding the 30-year average by over 20 mm in most areas. Similarly, Kilifi was expected to experience above-normal rainfall. Kaloleni sub-County, within a high rainfall area, was forecasted to receive the highest amount of rainfall in Kilifi County during the season. Monthly and weekly forecasts from KMD were also shared.

Weekly weather advisories for Vihiga and Kilifi sub-counties, produced by KMD, were shared with the SCAOs, who then communicated the information to farmers, emphasizing its impact on farming activities. In Vihiga, for example, advisories included guidance on farm management practices based on the expected rainfall, such as measures to prevent soil erosion and clear drainage channels to prevent flooding in cropped areas.

b) Crop Conditions and Yield estimates from Remote Sensing Data

Crop condition maps produced by MoALD and GEOGLAM Crop Monitors were used to monitor crop performance over the duration of the season. Satellite data and other ancillary data used to produce the maps include: Crop calendars for the short and long rain seasons, cropland maps, crop specific masks, NDVI, NDVI anomaly, precipitation and precipitation anomalies, temperature, soil moisture and evapotranspiration. These were supplemented with farm observation data from within the crop growing areas collected by the SCAOs and WAOs. These maps are also used at Global level to produce the GEOGLAM Crop Monitors which provide open, timely, and science-driven information on crop conditions in support of market transparency and early warning of production shortfalls [18] [19] . Crop condition maps for some of the months of the study season, were retrieved and they indicated Watch² and Favourable conditions for maize throughout the season in the two study sub-Counties (see e.g., Figure 17 and Figure 18 ). The indicators showed positive outcomes at the end of the season towards the harvest with 75% favourable conditions for maize in the country.

The areas marked in yellow and labelled Watch include those that were affected by floods where 36 counties affected by floods which destroyed over 17,600 areas of farmland in the country [20] .

Enhanced Vegetation Index (EVI) and NDVI trends were calculated using Landsat8 and Landsat9 images for the four sub-counties, with Google Earth Engine (GEE) used to extract data for analysis between January 2022 and December 2023. NDVI and EVI were calculated using specific bands from the satellite images. The formulae used for NDVI and EVI are shown in Equations (1) and (2) respectively with the monthly average NDVI and monthly average EVI trends shown in Figure 19 and Figure 20 respectively, the results revealed a marked increase in both indices during the OND 2023 season compared to OND 2022, indicating improved vegetation health.

$NDVI=\frac{NIR-RED}{NIR+RED}$

Equation (1): NDVI formula (Source: USGS— https://www.usgs.gov/ )

Where

NIR = Near InfraRed band (SR_B5) value

Red = Red band (SR_B4) value

$EVI=2.5\ast \frac{\left(NIR-RED\right)}{\left(NIR+6\right)\ast \left(RED-7.5\right)\ast \left(Blue+1\right)}$

Equation (2): EVI formula (Source: USGS— https://www.usgs.gov/ )

Where

NIR = Near InfraRed band (SR_B5) value

Red = red band (SR_B4) value

Blue = blue band (SR_B2) value

c) Yield Data from MoALD

The Ministry of Agriculture collects crop production data monthly through County agricultural officers, which is forwarded to the National Government’s Statistics Unit to produce crop monitoring reports. At the end of the season, the production data is also used to estimate the yield for the season for various crops. In some of the Counties, such as those that produce staple crops in large scale, ground validation of the yield is conducted through physical crop cuts and weighing.

Table 9 shows that in the OND seasons of 2022 and 2023, maize yields significantly increased in both Kilifi and Vihiga Counties. Kilifi saw a more than 100% increase in cowpea yields, while Vihiga experienced a notable decrease in bean yields. The reduced bean yield in Vihiga was attributed to heavy rainfall, which caused flooding and prolonged high soil moisture, damaging the crop.