Kenya

The DREAMS partnership, led by the U.S. President’s Emergency Plan for AIDS Relief (PEPFAR), is an initiative committed to helping adolescent girls and young women develop into determined, resilient, empowered, AIDS-free, mentored, and safe women. This is achieved through a core package of evidence-informed approaches that address the structural drivers which directly or indirectly increase girls’ HIV risk beyond the health sector. Given the complexity and diversity in implementing these interventions, DREAMS provides multiple services from a variety of sectors to ensure that adolescent girls and young women are reached in a holistic manner. It is therefore crucial that data systems that collect various elements for each service be interoperable to facilitate seamless service delivery to DREAMS recipients. Such interoperable data systems would then be useful in optimizing data use to inform expansion and/or intensification of DREAMS interventions. 

The Global Partnership for Sustainable Development Data and PEPFAR identified the DREAMS partnership as an example that included significant experimentation across multiple countries towards interoperability for tracking delivery of layered services for adolescent girls and young women. By using Data interoperability: A practitioner’s guide to joining up data in the development sector as an analytical framework, this case study aims to distill good practices in developing and implementing interoperable data systems at the program level. The case study focused on analyzing DREAMS programs in Kenya, Uganda, and Zimbabwe. It also aimed to extract lessons learned and insights that could inform improvements across all DREAMS data systems to support similar efforts in other contexts and sectors. 

A global team of more than 30 Omdena collaborators worked with the Kenya Red Cross Society to first build deep learning models to accurately identify Kenya land cover types from satellite imagery and then quantify the damage caused by the 2020 desert locust outbreaks on croplands and pasturelands by hectares.

Our damage assessment solution was developed in a bottom-up fashion. The team broke down the problem statement in subtasks and organized it into task groups which helped to explore a lot of ideas and iterate quickly on each building block of our solution.

In this article, we will only highlight the main approaches we took for assessing the damage caused by the 2020 Kenya desert locust outbreak.

Data acquisition

We opted for Sentinel-2 images using the Google earth engine platform. Even though it doesn’t have the best spatial resolution (10m per pixel for most of the bands), Sentinel-2 has a decent temporal resolution compared to other open-source satellite imagery, with a revisit time of 5 days at the equator in cloud-free conditions. It also comes with multi-spectral bands that could provide valuable information and insights to analyze the vegetation loss, which is not captured by regular RGB channels. Thus, the Sentinel-2 source was our best option for this challenge.

Where to look

Before collecting the actual images, we first needed the coordinates for the regions that were attacked and the exact timing for those attacks. This information was available on the locust hub from FAO.

The goal was to have images before and after the attack to analyze and assess the damage the desert locust caused. We used the dates the locust hub provided to export images for each coordinate.

Here we faced another challenge — making a suitable selection of interval period for the attacks, i.e the temporal gap between before and after image at a location. This had to be done while also taking into account the temporal resolution for our source (five days in a clear weather condition), as well as the cloud conditions. We ended up choosing a range of a minimum of one day before/after an attack and up to 15 days. This gave us the best results when it came to image quality and quantity and also allowed us to ignore the seasonal effects.

Given above, are the RGB and NIR-Red visualizations of a before-after image pair of an affected location in Kenya. In this example, we can spot the loss in vegetation with our naked eyes by comparison of the normalized difference vegetation index (NDVI) values.

To consider a pair of images, we needed them both to be usable. By that we mean the images have to consist of all the bands and with cloud coverage as little as possible. This resulted in discarding a good amount of images, but in the end, we were able to have 348 pairs of images to perform our analysis.

Another example from our collected data. This time we can see the clouds in the before image.

At this point, we had something to get us started with a basic understanding of the problem, and an idea about the effect of desert locust on some of the regions. But was that enough?

Getting more data for land cover classification

The data collected so far gave us insights on the loss of vegetation overall, but we needed more details. For example, to get the loss in vegetation per land cover type, we would first need to classify our land cover types and then perform the analysis on each one of these classes. And here comes the real power of AI and machine learning to help us drive these insights.

From Sentinel-2 we had our test data, but we still needed data to train our land cover classifier model. We had six main classes to take into account, as requested by the KRC — cropland, pastureland, bare soil, open water, forestland, and other-land. Considering the limited time of eight weeks in hand, and taking into account how time-consuming annotating such data could be, we needed a labeled dataset. This dataset had to have a similar resolution to the one we previously acquired for the damage assessment analysis and also similar characteristics when it comes to the topography and land types, and it had to be also free.

LandCoverNet Dataset

Not only was the data annotated, but it also has Africa as one of the main covered regions, and was collected from Sentinel-2. Again this was important because we would use this data to train our model and later test it on our before/after data, so they had to share the same properties to have consistent results, so this was perfect!

Here, it is worth mentioning that before LandCoverNet, we tried another labeled dataset which was Eurosat. The dataset was also taken from Sentinel-2, but the downfall was that it covered only regions within Europe, which proved to have quite different characteristics when testing the trained classifier on our collected images.

Supervised Learning

To build land cover classification ML models, we explored two approaches in the supervised learning category — image classification and semantic segmentation. Such classification models are first trained on a set of labeled images, to learn the class mappings with respect to image features and can then be used to predict the classes in test image inputs. In our case, the 6 classes of interest were provided by the KRC — cropland, pastureland, bare soil, open water, forestland, and otherland. The labeled data used in both the approaches were adapted to this set.

1. CNN image classification

Using transfer learning, we fine-tuned Resnet50 CNNs on the Eurosat dataset. Although it achieved a high accuracy of 98% on the Eurosat test data, the inference of this model performed poorly to our Sentinel-2 imagery of Kenya’s affected regions. One of the biggest plausible reasons for this behavior is the big gap between European land cover type and Kenyan vegetation. To bridge this difference, we had performed a label adaptation step between Eurosat land cover classes and expected classes in Kenya, which in turn, introduced fresh challenges. For example, all the images belonging to Industrial/Highway/Residential/Herbaceous Vegetation were mapped to one single class ‘Otherland’ (KRC label). This mapping process created a highly unbalanced dataset for training, and the resultant model predicted most of our images as ‘otherland’. Additionally, Eurosat did not have a ‘bare soil’ class, which is a common land cover in Kenya. As a result, Eurosat turned out impractical for our challenge.

2. Semantic Segmentation

To overcome the above challenges, we moved to land cover modeling using the LandcoverNet dataset which has classes similar to the African vegetation cover. This data has pixel-level labels. This required us to build a semantic segmentation model instead of the image classification. Segmentation is the process of classifying each pixel and assigning it to a particular label. Here we don’t separate instances of two same classes as we care only about each pixel category. It is mainly used to locate objects and boundaries in the images. Also, we can have an accurate view of an image as we can understand what is exactly in the image at pixel level view in a single class.

Moreover, semantic segmentation is quite useful in detecting and classifying the object in an image when we have more than one class in an image. This helped in our case as well, as our Sentinel-2 images had multiple classes within the same image.

3. U-Net Model

We used a pre-trained U-Net model for performing semantic segmentation. Originally introduced for biomedical image segmentation, U-net can perform fast and precise segmentation of images. Processing of 512×512 images takes less than a second on a GPU, which makes it very practical to use. We had used the Fast.AI framework, using the Google Colab platform. For training, we used 1980 images from the collected LandCoverNet data and used only 3 spectral bands (R, G, B). The original labels were also adapted to fit the KRC classes as shown below.

• Unknown, Artificial ground, Snow/ice = Otherland
• Cultivated Vegetation = Cropland
• (Semi) Natural Vegetation = Pasture land
• Natural Ground = Bare soil
• Water = Open water
• Woody Vegetation = Forestland

Due to the similarity between the labels, no significant imbalance was introduced. The U-net model achieved an impressive 81% accuracy in classifying the LandCoverNet data. Below, we can observe two targets/prediction pairs:

Having confidence in this modeling, we now applied it to classify our Sentinel-2 imagery. A sample predicted segmentation is shown below.

The segmentation masks and their prediction confidence were sufficiently high for us to next use them for the damage assessment section. However, before discussing them, we list here some ideas with which the segmentation modeling could be further improved in the future:

• Increasing the number of input images
• Fine-tuning model parameters
• Adding more spectral bands
• Increasing the number of training epochs

Unsupervised Learning

Before moving to the damage assessment section, we wanted to briefly touch upon the experiments we carried out during the initial phases of the project when labeled datasets for African vegetation were not yet available. During this time, we applied k-means clustering with various feature extraction techniques to the Sentinel-2 imagery of the affected regions. The images or pixels, which end up in the same cluster, would be more alike than those in other clusters. The different clusters could be considered to be representative of the different land-cover types and then be used for damage assessment.

We started with the simple Principal Component Analysis (PCA) for feature extraction. In this example shown, k-means was applied to the top three bands found to be the principal components in a sentinel-2 image having six bands.

However, a limitation of such segmentation was that the clusters formed in one image would be different from the clusters found in other images. This would not be very reliable as we would like to identify the same classes across images. A possible strategy to overcome this was to perform PCA (followed by k-means) on a single large image covering the whole of Kenya. But, this turned out to be computationally very expensive and this approach was not pursued. Therefore, next, we moved to attempt image-level clustering. Three main techniques applied were — the SCAN algorithm, Image feature vectors from a pre-trained CNN, and auto-encoders. An example is shown below, using SCAN. Detailed results are omitted and can be read on our unsupervised article for this project.

How to use the clusters in our work

A major limitation with clustering is that although the images are grouped, they do not give any information about which cluster represents what type of land-cover. Therefore the next step was to annotate the clusters and map them to a land-cover class. In the initial stages, we attempted manual annotation, by visual observation of the cluster images. Later, we obtained the Land Use Land Cover (LULC) map of Kenya, 2019 (which differed from the land-cover types given KRC but was a good representation of Kenya land-cover type). As each Sentinel-2 image in any given cluster referred to geolocation, we plotted points of all the clusters on this map. Using majority voting of the land-cover type that most of the images in a cluster belonged to, clusters were assigned an overall class label. A sample plotting of images from the clusters of the VGG16 features are shown below:

In the annotation step with the LULC map, we observed that more than one cluster was sometimes assigned the same label, therefore, indicating that clusters could be further merged. It implied that our database of the attack locations of Kenya does not contain six different land-cover types as expected by KRC. The clustering is thus expected to improve if carried out with an optimal value of k, using elbow methods, etc. Additionally, K-means being a hard clustering technique, any given image could belong to only a single cluster. As a result, the entire two km of land covered in an image in our project, would be categorized as a single land-cover type, whereas the ground reality could be that multiple land-cover types exist in that land covered in a single image. In future works, soft clustering methods (such as Fuzzy clustering) could be used. Therefore, in the later stages of the project when Land-Cover-Net labeled data was obtained, the U-net supervised modeling was pursued to perform semantic segmentation which overcame this limitation of image-level clustering.

Damage Assessment

Once the land cover types were classified the next step was to assess the impact of damage caused by desert locusts in these areas. Traditionally, damage assessment is made by collecting crop acreage, data from field surveys, farmers, etc., which is time-consuming and cost-intensive. We researched the current approaches that are used for impact assessment and how to incorporate them effectively in our existing ML pipeline.

How did we approach this?

Remote sensing for vegetation provides a cost-effective solution for vegetation condition monitoring by providing multi-temporal images. It is mainly done by obtaining the electromagnetic wave reflectance information from canopies using passive sensors. The reflectance of light from plants changes with its type, the water content in its tissues, and other intrinsic factors. Indices extracted from these spectra can be used for the vigor quantification of plants.

Vegetation indices

Vegetation interacts with solar radiation differently when compared to other natural materials. The vegetation spectrum reflects the green wavelength, strongly reflects the near-infrared wavelength (NIR), and absorbs red and blue wavelengths. The variation across the spectrum is further caused by carbon content, water content, nitrogen content, pigment, etc. The relationship to one another is known as vegetation indices.

Using vegetation indices we can get information about plant health, growth. It also helps to categorize different land covers as shown in the image above. Such Indices are based on three light spectra: UV region, Visible spectra & Near and Mid-infrared band. The three main vegetation indices which we considered are:

1. NDVI (Normalised Difference Vegetation Index)

A widely used index to measure vegetation condition or health. In other words, it is a simple indicator of photosynthetically active biomass.

How does NDVI work?

During photosynthesis plant, cells use carbon dioxide and sunlight to produce oxygen and sugar molecules. The primary pigment used in photosynthesis is chlorophyll and this absorbs visible light and reflects near-infrared light(NIR). By comparing NIR with visible light we can differentiate healthy plants from unhealthy ones. NDVI is used in this calculation, by measuring the difference between near-infrared light(NIR) and red light.

Very low values (<0.1) indicate barren rock or snow, water and high values (generally around 0.6 to 0.9) indicate dense vegetation, such as forests and crops. In the image below, we can see the evolution of NDVI values before and after the locust attacks. The after image clearly shows the degradation of vegetation, indicated by lower values of NDVI.

2. EVI (Enhanced Vegetation Index)

To address the soil and atmospheric limitations with NDVI, the Enhanced Vegetation Index (EVI) was introduced. Liu and Huete, after a detailed study, concluded that due to the interaction between soil and atmosphere increasing one may reduce the other. To simultaneously correct the soil and atmospheric effects, they built a parameter in this algorithm. EVI ranges from -1.0 to +1.0, with healthy vegetation generally between 0.20 to 0.80.

3. SAVI (Soil Adjusted Vegetation Index)

In areas where the soil is exposed and vegetation cover is low, NDVI values can be influenced by soil reflectance. SAVI was introduced to tackle this soil brightness factor. It is simply a modification of NDVI with an adjustment factor L for soil brightness. The default value for L is 0.5 and works well mostly. When L=0, NDVI=SAVI.

Calculating the degree of damage per land cover type

We obtained land-cover segmentation masks from the images taken before the locusts attack, using the U-net model. Then, we compared the vegetation indices of these segmented images with their temporal counterparts, i.e the after attack images. This analysis of index changes, allowed us to compute the total area damaged in hectares, the total area damaged per county, and vegetation cover damage per county. In the plots below, we report the NDVI changes observed. A similar trend of behavior was observed across EVI & SAVI indices as well, which validated our loss assessment.

NDVI analysis

The vegetation indices comparison illustrates that in terms of the amount of land (in hectares) affected, the most damaged land-cover type was pastureland, followed by forestland and then cropland. Laikipia county suffered the biggest loss in cropland, while the most loss in pastureland was seen in Turkana county. The counties that faced major damages combining the effect across all 3 land cover types were —Turkana, Marsabit, Samburu, Isiolo & Laikipia. However, in terms of the degree of damage suffered, most counties faced low damage (NDVI difference below 0.2) except Kirinyaga and Laikipia, where some high damage was seen.

The following heatmaps highlight the geolocations of the damage. It is important to note in these plots that the resolution of the images is only 2km around each point, so there are possible gaps between the collected data points that were not part of this analysis. In the future, having bigger footprints could help improve this analysis.

Conclusion

The challenge was to leverage machine learning power and create a damage assessment tool for the desert locust attack. Data was one of the major obstacles, and even though we managed to overcome that, having a higher image resolution would have given far better insights as is the case with all machine learning problems. There is the “Copernicus Global Land Cover maps” dataset that we came across by the end of the challenge, and it could be used in the future, for training a classifier and comparing the different results.

There is also the possibility of trying other classification algorithms to enrich the outcome. And in fact, we did have some time to begin such an attempt by working on a U-net model that incorporates the NDVI band. After a few difficulties with the preparation of the dataset and the model to accept the fourth band, the model was finally working, but unfortunately time was not on our side and we couldn’t tune the model to have better results. Hence, we continued with our original U-net RGB model classifier. Another interesting approach was made by a collaborator was to predict the change in NDVI values for future attacks using the LSTM model. Working on different spectral indices could also provide more insights in this case.

This work was part of the Kenyan Red Cross efforts to offer relief for those affected by the large scale attacks. Being part of such a humanitarian mission, made us feel highly responsible and honored, motivating us to keep going to overcome the challenges along the way. We hope this work would provide first insights into the locust attack impact as was originally intended and pave the way for more future analysis in this direction.

Kenya’s Agriculture Sector Transformation and Growth Strategy (ASTGS) 2019-2029 commits to ensuring that agriculture data is available, usable, timely, and interoperable. It recognizes the importance of having both traditional data such as censuses and surveys in addition to other innovative data sources. 

The need was further exacerbated with the COVID-19 pandemic. The government needed to make decisions in real time and needed accurate, reliable and timely data for these decisions. With a reliable network of stakeholders and existing partnerships in the agriculture sector, the government looked to existing systems, institutions, and networks to mobilize action.

The Food Security War Room (FSWR) was established to rapidly support Kenya’s COVID-19 response, and ensure that despite the pandemic, three key objectives (all guided by reliable and timely data) were met:

1. Ensure the availability, accessibility and affordability of food and water (e.g., maintain flow of produce to markets including imports, minimize disruptions to market operations, and monitor price volatility of food)
2. Support subsistence farmers, livestock farmers and fisherfolk (e.g., maintain availability and affordability of inputs, mitigate impact of the locust invasion)
3. Maintain agricultural output and value addition (e.g., support operations of large farms and processors, and limit disruptions to markets including for export)

To propel the FSWR, stakeholders in the agriculture sector organized themselves and created a forum for experts to brainstorm and advise the government on the most appropriate solutions. From the onset, there was concurrence that the country did not have reliable information on available food stocks. While the national government and counties (sub-national governments) collect information on the food balance sheet on a regular basis through the Digital Food Balance sheet data on the food balance sheet across the East African region is viewed as unreliable—yet a large proportion of the population is constantly food insecure.

The FSWR posited that accessing information on the current food stocks would be their number one priority to guide their decision making.

A resilient system: A key success factor

The overall factor that led to the success of this exercise was the resilience and readiness of the Ministry of Agriculture Livestock Fisheries and Cooperatives (MoALFC). The use of existing mechanisms meant that these could be deployed at short notice using systems that are stable and robust, and players who were able to work in a coordinated effort. 

• The Agriculture Sector Development Support Programme II (ASDSP II) provided an existing mechanism for mobilizing and coordinating value-chain actors at the county level, and submitting the data verified by County Government officials to the National Government. With the programme already in operation in all 47 counties, there were resources (human and financial) to deploy this work, with a well-established mechanism for sharing information between the national and sub-national governments and with key stakeholders.
• ESRI already had existing solutions for data management, data visualization through dashboards as well as data collection through the ArcGIS for Survey123 app that were possible to configure in a short timeframe. Together with the support of other players, and as part of its COVID19 disaster response program, ESRI was able to offer free licenses to the government to enable this exercise. The choice of a geospatial platform would enable understanding location differences and also signifies that this solution can be used in future and can be regularized.
• MoAFLC had an existing relationship with ESRI making mobilization easier.
• There was a ‘ready’ network of stakeholders who were able to provide technical and practical support to ensure this worked. Key industry leaders were convened on a WhatsApp group (the Joint Agriculture Action Group-JAAG), where daily progress was monitored and specific organizations called upon to provide technical and other kinds of support.

The solution

To enable timely data collection in all the 47 counties, a team to execute this assignment was availed by the ASDSP II. This program’s Monitoring and Evaluation (M&E) officers, who are present in all the counties and have existing national coordination mechanisms as well as solid linkages with the county governments, were engaged. The data was to be used in designing strategies to support the efforts of the FSWR during COVID-19 and to guide ASDSP II in their subsequent projects.

The M&E officers and programme coordinators were virtually trained by ESRI on the mechanisms of implementation of the survey. These M&E officers identified their enumerators at the county level and trained them virtually too.

Data collection exercise commenced on 4th May 2020 and ended on 25th June 2020. In less than two months, data on the availability and pricing of food staples was available for the entire country.

I wish to thank the ASDSP II team for the dedication and support for this initiative. This is going to be a major learning curve in the efforts towards improving agricultural data collection. The success of this initiative will inform the beginning of real time national and county data collection and reporting. Let's all document observation and lessons for further use. - MoALFC

For the full table of key players and organizations that supported this process, see the PDF download below.

What did the data show?

A total of 26,134 respondents were reached during the survey over the two months. On average, each county collected information from 556 respondents. All the data was collected and aggregated in a Food Staples Dashboard, which provides analytics by staple, quantity, price, and the location and geographic distribution of the produce.

From figures 2 and 3, the prices of some staples changed significantly in the two months. For example, the price of maize, rice, and green grams rose significantly indicating a high demand for these products. Given that households were feeling the effects of the pandemic on the economy, these staples were becoming popular as they were affordable, but this was also affecting its supply and prices.

A snapshot of two counties as shown in figures 4 and 5 also provides sub-national situation analysis on food availability and pricing. In Nairobi for example, in the capital city, food was readily available. In Nandi county, food prices were quite fair compared to available quantities. While it may not be possible to compare across counties because of context specific differences, these sub-national snapshots guided the decision makers at the county level.

Impact: Decisions made using the data

I. Guidelines on Food Availability and Food Prices:

With the up-to date  data collected, the Ministry of Agriculture was able to develop guidelines on food availability and food prices that were rolled out across the country. The guidelines provided protocols for managing food availability and guidance on the recommended food prices for the various commodities. Among other things, the protocols included:

1. Encouraging County Governments to continuously sensitize workforce on the need to sustain food supplies in various markets on COVID-19 to prevent panic and enable them to perform market functions in the long run.
2. Suspension of market access charges for three months to cushion food stockists.
3. Encouraging food suppliers to adopt e-commerce to minimize contact and to allow households in urban areas to access food.
4. As an early warning mechanism, food distributors were encouraged to share with the government any price volatility or food rationing.  

II. Improving Two-Way Communication between the Government and the Public:

1. With accurate data, the Ministry was able to frequently equip the media with accurate information and detailed analysis to inform the public on the availability and access of food, agricultural services and commodities. This was crucial in minimizing speculation and public misinformation as specific bottlenecks to the food-system were openly highlighted, alongside respective actions undertaken by both levels of Government. The public was also updated when problems were solved (e.g., re-opening of markets, reduction of food prices);
2. A Food Security Hotline to respond to urgent enquiries from stakeholder and members of the public was set up. In the initial days, most calls to the hotline raised concerns over access to market during the lock-down of the Capital, Nairobi and its neighboring Counties, and Mombasa in the Coast, by farmers and agricultural vendors and to report on problem areas that needed the attention of the FSWR (e.g., corruption by traffic police, misinterpretation of COVID-19 agricultural markets’ guidelines and protocols etc.);
3. The private sector and other key stakeholders had an opportunity to query data (specifically data on food quantities and prices) and to raise concerns where there was doubt through the FSWR weekly meetings. These were addressed by both national and county government officials;
4. Avenues for leveraging resources were identified by development partners, who made efforts to adjust/align programme budgets to address food system issues highlighted in the weekly reports. For instance, the World Bank projects joined the ASDSP II to support data collection efforts and verification efforts in counties, enabling the food balance sheet to cover more value chains;
5. The One-Million Kitchen Gardens Campaign was launched to address the shortage of fruits and vegetables in urban areas. This Government-led campaign was designed to educate the public on the importance of maintaining a balanced diet during the COVID-19 pandemic and on inexpensive methods of producing and preserving vitamin-rich foods in their homes.  

III. Data was used by County War Rooms to determine areas of food deficit and surplus, and trigger distribution of relief food.

IV. Extension Staff used the data to determine and inform value chain actors on opportunities for trade.

V. The success of the system has informed programme and projects on areas to consider on market information, key in point, Kenya Cereal Enhancement Programme Climate Resilient Agricultural Livelihoods Window (KCEP-CRAL)  and the Kenya Climate Smart Agriculture Project (KCSAP) which have now developed comprehensive data systems within their scopes.

“By August 2020, ~185 issues directed to the hotline (0800 724 891) have been resolved.”- FSWR August 2020 Bulletin

The data enabled us to know that there is a great disparity between national and county governments, and we are able to know what exactly the challenges were. For example, availability of enumerators, devices to collect information and digital tools for collection of data. - MoALFC

For ASDSP II, who were instrumental in the delivery of this activity, this data was instrumental in planning the next phase of their project.

This was a great opportunity for the programme to be entrusted by the Ministry to collect and share data on available surplus food stocks in all the 47 counties during this COVID19 pandemic period and beyond. This enabled us to harness the capacity and experience among various actors to collaboratively collect and share data in real time between the National Government and County Governments” - ASDSP II

What next: Transformations for strengthening agriculture data post-COVID-19

The food staples survey demonstrated some key lessons for government and other agriculture stakeholders. It enabled government to re-think how data can and should underpin agriculture transformation and what it would take to do this.

Transforming agriculture data through better coordination and lesson sharing

This exercise initiated a conversation on how the country can truly transform the agriculture sector’s data systems and the importance of a geospatial infrastructure. This should inform the government for the years to come, as it works to align and strengthen its data systems and transform the Agriculture sector, and to realize the goals outlined in the 10-year ASTGS. The MoAFLC is keen to sustain these lessons, but importantly, to consolidate and map all the efforts on agriculture data so that this is coordinated and allows for better partnerships and build a data-use and share culture among stakeholders. This is a role that the ATO - under Flagship 8 of the ASTGS - is mandated to lead on.

The ATO is keen to consolidate all the lessons learned from previous interventions so that it moves beyond the implementer and the donor and become meaningful and accessible to everyone in the agriculture data ecosystem. - MoALFC

The FSWR has now transitioned to a Food Security Monitoring Committee (as envisaged in the ASTGS) to ensure continuous surveillance on food affordability, accessibility, and availability. This requires appropriate funding and policies, and the momentum needs to carry on beyond COVID-19. The Government needs to be able to sustainably support its ministries, departments, and agencies to collect and use data and information for decision making by investing in domestic finances for data. Sustaining these activities requires a mix of skills, culture shift, resilience, and adaptive capacity (i.e. agility to work with multiple stakeholders, strong internal systems that can accommodate the influx of tech and data tools and solutions particularly during a crisis), to create an enabling environment that fosters partnerships to strengthen the access and use of correct and timely data in the sector.

The digital food balance sheet was initially collecting information on maize, now the Ministry is keen to expand to other 12 value chains across 256 markets in the country through the Kenya Climate Smart Agriculture Project (KCSAP) Market Information App. We need to be able to make sure that this is managed and sustained. - MoALFC

Transforming agriculture data needs more and better financing

The Government through the ATO is committed to building and consolidating digital tools and building skills as it moves into operationalizing the ASTGS. While the challenges exist, the government understands that for more and better resources to be channeled to agriculture data, government structures and policies would need a re-think. For example, current financing to Ministry of Agriculture should have a dedicated line item for agriculture statistics. In addition, more and better funding should be channeled to traditional data such as surveys and censuses as they would form the foundation for innovation and digital tools.

Innovative methods of financing need to be explored such as index insurance, public private partnerships to allow for through-flow of innovations to and within government and realigning of domestic budgets towards creating a robust digitally enabled workplace.

The bottom line about data for the agricultural sector is having a structured and sustainable funding approach to meet the need for data collection, analysis and sharing. Data is a social good. The challenge is for the Government to make this a national priority with a long-term view to having a structured approach to data management, while at the same time engage donors for short-term seed funding for capacity building, systems development, etc. - MoALFC

Conclusion

The COVID-19 pandemic accelerated the data revolution globally, and Kenya’s agriculture sector was not left behind. As governments needed accurate, real time information to make decisions about people’s livelihoods, the government of Kenya was able to take advantage of existing mechanisms, solutions and networks, to generate and use data for decision making on food security. This demonstrated a resilient system, but it also posed as a learning point for the government.

To truly transform the sector’s data needs will require deliberate action. Two ingredients are key: better coordination of the players and systems in the country and more and better financing for agriculture data.

As the country moves into the post-COVID-19 era, measuring the impact of these lessons would be important. The Ministry of Agriculture understands what is needed to fully transform the sector and the data generated in the sector. As the ATO becomes fully operational, this understanding, lessons and experiences would be put to test. ATO is spearheading four main activities to strengthen research and use of data for better decision-making and performance management in the agriculture sector:

1. Digitization of existing data, research and other performance information held by MoALFC and associated agencies; 
2. Creation of an enabling environment for research and innovation with clear linkages between data, research, and innovation;
3. Defining data laws and set up open data platforms for agricultural data at national and county levels to accelerate the launch of the research and data flagship;
4. Launching data use cases.

Discussions are underway with the Agriculture and Rural Development donor group and other partners, to build on the successes and lessons from this initiative, to enhance the Ministry’s M&E capacity through the ATO in line with the ASTGS Flagships 7 (i.e. launch three knowledge and skills-building programmes for ~200 national and county government leaders and flagship implementers and establish a digitally enabled extension programme led by ~3,000 county-based youth extension agents) and 8 (i.e. strengthen research and innovation as launch priority digital and data use cases to better drive decision making and performance management). As the ATO develops a programme performance dashboard to strengthen the role of M&E and use of data to improve service delivery and inform decision making. In so doing, the ATO seeks to address three key challenges the sector faces to be able to realize real-time operational and strategic improvements.  First catalyzing the research and innovation space in agriculture, including around use of big data and advanced analytics. Second, enabling more reliable access to useable and shareable data. And finally, demand for quality analyses to support evidence-based decisions on performance management, M&E, research and policy.

This initiative is enough proof that Kenya’s journey to transformation aboard the data revolution bus is on the right track. 

This case study benefitted from the review of Thule Lenneiye (ATO), Tom Dienya (MoALFC), Neema Grace Mutemi (Food Security War Room), Heri Mwagonah (MoALFC), Richard Ndegwa (ASDSP II), Bernard Mwangangi (ASDSP II), Philip Thigo (Thunderbird School of Global Management), Clifford Okembo (ESRI), Stanley Mbagathi (SustaiNet) and Mikael Segerros (NIRAS).