Build up the logistic framework to allow for an efficient analysis and dissemination of information relevant for pollination and pest control.

Activities include

• Identifying available data sources for weather information, pollination and pest control

• Identifying the most critical data gaps and develop/design observational networks (based on WP1 and WP2)

• Combining relevant data sets and observations for an optimized analysis with respect to pollination and pest control

• Developing easy and effective dissemination pathways to ensure that the information reaches the right end-users in a timely manner

Potential innovation: Development of automated information collection for insect pollinators and pests. And also for the first time combined utilization of large weather information data sets in insect pollinator conservation planning and pest control method design.

Deliverables: An efficient data dissemination pathway developed

 

Task WP3.1 Weather information 

Partners: UoB DIT, MAK, UoJ

Duration: M1-M60

Description of work:

Weather has a significant influence on the pest movements and population growth [1]. For example the populations of fruit fly are found to be positively correlated with both morning and evening humidity including precipitation but negatively correlated with the prevailing air temperature [2]. Therefore, an efficient weather observational network is therefore key in understanding and managing pests in an agricultural ecosystem [1, 3]. 

The goal is to develop a weather observational and information sharing framework for aiding the monitoring and control of pests including pollinator insects in the selected agro-ecological zones.

Activities include

• Create a catalogue of available relevant observations (e.g. WIMEA network, station network of the involved national Meteorological Services; satellite data) and freely available model output (e.g. from yr.no for weather forecast, or public available seasonal forecasts)

• Investigate the interaction between various meteorological parameters on pollination and pest control (both by literature studies and hopefully also own investigations)

• Use this information in combination with available observations and weather forecasts for an estimation of the potential and risks with respect to pollination and the occurrence of pests

Deliverables:

D3.1.1 Validated weather information protocols

D3.1.2 Capacity of local climate scientists in pest control and monitoring 

Potential innovation: This work package will create a comprehensive overview over available data sources for meteorological and other relevant environmental parameters, focusing on the aspects of agricultural pest control and pollination.

Deliverables: [specify the task/wp deliverables and their month of delivery]

M9: Detailed list of weather and environmental information sources and categorization of accessibility (free/restricted, but freely accessible for research/restricted)

M15: Coding available for automatic download of the most important information sources

Interdependencies: [specify in- and out- dependencies of the task/wp with other tasks/wps]

The work in WP3.1 is a basic requirement for the scientific analysis in WP3 and the outreach and dissemination of information in WP4. Identified shortcomings, both with respect to sensors and networks will give important input for the research tasks in WP1 and WP2

Resource requirements: [specify reliance on students, field work, infrastructure, equipment

 

Task WP3.2 Data analytics for Insect pollinator monitoring

Partners: NTNU DIT, MAK, UoJ

Duration: M1-M60

Description of work:

This work focuses on documenting the insect pollinator species diversity and abundance across different agricultural landscapes. As well as their health status all aimed at contributing towards their conservation through identification of biodiversity hotspots in the region. Insect pollinators such as bees are very important in the East African cropping system since them pollinator crops and generate incomes through their product sales. A PhD and masters’ student will be recruited to collect field insect samples, analyze them in the laboratory for species and pathogen status then publish. This information shall be combined with WP1 and WP2 to generate insect population diversity map across sampled region. This information is key for planners to use in conservation of the insect pollinators.

The ultimate goal of this study is to improve insect pollinator (bee) biodiversity conservation through continuous and systematic bee species monitoring across wild and agricultural landscapes. And the specific objectives are to investigate the temporal and spatial abundance and diversity of  insect pollinator (bee, hoverflies, butterflies) communities in natural and agricultural landscapes of Uganda, to examine the effect of agricultural activity, seasonal weather patterns, insect pathogens (bee pathogens) and agrochemicals on species abundance and diversity across agricultural landscapes and to determine the most dominant plants that honeybee’s visit in the area through tracking of honeybee flights. All aimed at enabling beekeepers document and conserve the relevant plants per region.

Methods

A combination of field and laboratory data collection. At the start candidates (PhD and Master student) shall be trained on field insect collection methods as well as identification. Then a team sent to the three selected agro-ecological zones for insect sampling. These will be citrus/mango, coffee and conserved natural habitats (near game reserves). The three landscapes have been purposively selected due to the economic value of their crops, citrus landscape represents intensive farming area, use of agrochemicals but also reduced diversity of pollination sources, coffee is Uganda’s most important cash crop and export and conserved natural habitats selected for standard comparison.

Phase 1: The data for these shall entail two phases; field and laboratory. First PhD candidate and research assistant shall be re-tooled on proper insect sampling. Insects shall be collected using diverse methods by use of hand nets, pan traps, malaise traps as well as trap nets. This initial collection is to obtain counts, take quality pictures for the technology team and obtain baseline data on species diversity across landscapes. The collected insect samples shall be identified using morphometric (structural features such as wing venation) and then further confirmed through molecular analysis (DNA).

Phase 2: Insect pollinators in Uganda are diverse for example central Uganda alone is 630 bee species and about 330 butterfly species. Implying that tracking all bee species in Uganda is impossible. For this study the team shall track honeybees in managed and wild nests to monitor their behaviour in terms of plants visited, flight patterns (since some bees exposed to pesticides loose orientation), disease (wing deformations), colony strength (more bees can be indicated by weight, but also high sound as many bees). Generally images of changes inside a beehive across seasons and landscapes. This information added with seasonal disease laboratory screening shall for the first time improve understanding of the level or effect of varied agricultural landscapes on bee colony health and survival in the Ugandan context.

Two species of stingless bees i.e. Hypotrigona gribodoi and Meliponula ferruginea shall be tracked to document their nesting structure, colony cycle (when they have honey, much propolis), track the plants they visit for ricin collection. This two species have been selected due to their progressive documentation in Uganda. Whereas the biology of honeybees is well understood, few studies have explored the study of stingless bees yet we need that information to domesticate these essential insects.

Activities include:

1. Identification of suitable student candidate for taxonomy and insect ecology studies  

2. Training of the students in basic insect taxonomy and DNA analysis of the insect species

3. Field insect sample collection from the four ecological landscapes

4. Photographic capture of the insects- all life stages

5. Sample collection for archiving or biobanking

6. Sample collection for species confirmation using DNA

Deliverables:

• Capacity building of one PhD and Masters student in the field of insect pollinator ecology by year 5 of the project.

• Pollinator insect species distribution map showing biodiversity hotspots produced – year 3 of the project

• Effect of different agricultural landscapes on insect pollinator species richness understood and published.-year 4 of the project.

 Budget for WP 3.

 

Task WP3.3 Data analytics for pest control 

Partners: NTNU DIT, MAK, UoJ

Duration: M1-M60

Description of work:

Insects are estimated to destroy between 18-20% of the annual global crop production which is estimated at a value of US$470 billion (1,2). Threatening livelihoods of the resource poor smallholder farmers mainly situated in Africa (3)The frequency of occurrence coupled with emergence of new insect pest threats have is expected to increase due to climate change (3).For example recent outbreaks of locust and fall armyworm and locus invasion (4,5)

This project therefore aims at documenting the insect pest species across different cropping systems in Uganda, assessing their population dynamics across seasons and pest management approaches and finally generating information on predictive population dynamics and models for fruit flies so as to support farmers with timely and accurate information they can use to design control strategies. All the above is aimed at enabling the government and other stakeholders in crop protection develop robust insect pest management programme for Uganda and compliment the efforts of other east African countries

This project shall target cereal (maize farms), fruits (citrus and mangoes) and other cropping systems. The team shall also develop automated pest control mechanism using technology that can link farmers to quick information of detecting insect pests on their farmers and obtaining quick advice on how to manage the observed insect pest. Since Uganda has a rich diversity of fruit flies (up to 102) that cannot be exhausted in this study, the research students shall focus on the oriental fruit fly monitoring (Bactrocera dorsalis (hendel) because it’s the major pest accounting for 99% of all fruit flies in Uganda

Potential innovation: This project will combine weather and insect pest population dynamics in predicting the effectiveness of the pest control strategies. Combination of artificial intelligence (technology automated systems in monitoring) will be the first in the selected landscapes.

Activities:

1. PhD and Master student selection

2. Field data collection

3. Laboratory sample analysis

Deliverables:

1. D3.3.1: Distribution and pest risk map over study areas

2. D3.3.2: Built capacity in using weather information for managing and control of pests

3. D3.3.3: Developed framework to estimate the probable losses arising due to pest / insect inversion 

Interdependencies: WP1 and WP2

 

Task WP3.4 Predictive models for monitoring the growth dynamics of pest / insect populations

Partners: NTNU DIT, MAK, UoJ

Duration: M1-M?

Description of work: The goal is to develop AI based predictive models for monitoring the growth dynamics of pest /insect populations over the study areas. 

Activities include

Activity 1: Developing a distribution and risk mapping framework for the pest / pollinator insects. In order to achieve this, a series of steps will be followed

a.   The first step is in understanding the different types and dynamics of pollinators/pests, species, sex and stages of development. This will be guided by literature, surveys and expert knowledge from entomologists. We shall also benefit from insect pollinator data and information from WP2 and WP3.3

b.    Using knowledge gained from (a), Artificial Intelligence (AI) will be used to automate the processes of pest/pollinator identification, species identification, gender identification, stage of development etc. The data used in this step will involve phone captured images, videos and sound from audio signatures. Computer vision, image processing, machine learning and signal processing techniques will be applied to automate the processes involved.

c.     Develop both web-based and phone-based applications for automatic identification of pests/pollinators. Make these systems location-aware for effective real-time risk mapping.

d.    Engage farmer networks to not only collect location aware data but also for in-field testing of the system. This will go in-hand with development of a data repository framework to enable generation of realistic risk maps.

Outputs

• • AI based pollinator/pest identification models

  • Pest/pollinator distribution maps

  • Data repositories

  • Farmer-network engagement plan.

Activity 2: Assessing climate change impacts and adaptation options for pest management and control.

This will involve;

a.     Developing a real time tracking system to assess the changes in pollinator/pest volumes in relation to climatic changes in both spatial and temporal spaces. The real time system will be done in two stages;

• An AI based quantification module for specifying the volumes of pests/pollinators over a specific period of time will be developed. The quantification module will involve an automated count based on the results obtained from activity 1.

• A reporting module to show how climatic changes correlate with volumes of pollinators and pests in both time and space.

b.    Developing a real-time advisory module to effectively guide relevant stakeholders on pest management basing on outputs from 2(a) and expert knowledge.

c. Develop a pest/pollinator mobility module to inform movement of these insects. Using drones to capture their movements. Predictive models on their next destination of attack can be developed.

d.     Real time tracking of stages of development of pollinators/pests, their impact and how climatic changes impact on their growth. 

Outputs;

• • Pest /pollination quantification model

  • Spatial temporal models showing effect of climatic changes on pollinator/pest volumes.

  • Pest control advisory platforms

  • Pest mobility model

  • Spatial temporal models showing effect of climatic changes on pollinator/pest stages of development.

Activity 3: Estimating probable losses due to pest inversion 

To estimate losses, a series of steps will be followed;

a.     Quantify severity of the effect of pest damage on crops: This will involve collecting crop data such as images of affected parts of a crop (leaves, stems or root images) with their corresponding levels of damage. From data collected, a series of machine learning techniques will be applied to develop automated models to quantify levels of pest damage on the crop.

b.    Engaging farmer networks to not only collect location-aware crop data but also for in-field testing of the system. This will go in-hand with development of a data repository framework. Analyses from such data will ably inform probable losses in different farm fields.

Outputs;

• • Severity quantification models

  • Reports on quantified crop damages.

  • Risk maps of crop damages 

Deliverables:

1. Distribution and pest risk map over study areas

2. Built capacity in using weather information for managing and control of pests

3. Developed framework to estimate the probable losses arising due to pest / insect 

inversion    

Potential innovation: This project will combine weather, pest/pollinator and crop data dynamics to build AI-based predictive models for monitoring the growth dynamics of pest / insect populations over the study areas. 

Interdependencies: [Wp 2 and WP3.3]

Resource requirements: [specify reliance on students, field work, infrastructure, equipment]

Deliverables:

Task WP3.4 deliverables and activities and the duration of delivery are specified in the gantt chart attached below.

1. 1. 1. Bibliography

[1] Kawakita, S., Ichikawa, K. Automated classification of bees and hornet using acoustic analysis of their flight sounds, Apidologie(2019) 50:71-79, Springer Nature 2019

[2] Noda, J.J., et al. Acoustic Classification of Singing Insects Based on MFCC/LFCC Fusion. Appl. Sci. 2019, 9, 4097; doi:10.3390/app9194097

[3] Xia, D. et al. Insect Detection and Classification Based on an Improved Convolutional Neural Network. Sensors (Basel). 2018 Dec; 18(12): 4169. doi: 10.3390/s18124169

[4]https://repository.lboro.ac.uk/articles/Locust_recognition_and_detection_via_aggregate_channel_features/9222596/files/16801931.pdf

[5] Valan, M. et al. Automated Taxonomic Identification of Insects with Expert-Level Accuracy Using Effective Feature Transfer from Convolutional Networks. Systematic Biology, 68(6), 876–895, 2019. https://doi.org/10.1093/sysbio/syz014

[6] Nanni, L., Maguolo, G., Pancino, F. Insect pest image detection and recognition based on bio-inspired methods. Ecological Informatics, 57, 101089, 2020. Doi: https://doi.org/10.1016/j.ecoinf.2020.101089

[7] Hansen, O.L.P, et al. Species‐level image classification with convolutional neural network enables insect identification from habitus images. Ecology and Evolution, 10(2), 2020. doi:https://doi.org/10.1002/ece3.5921

[8] Maggiora, R., Saccani, M., Milanesio, D. et al. An Innovative Harmonic Radar to Track Flying Insects: the Case of Vespa velutina. Sci Rep 9, 11964 (2019). https://doi.org/10.1038/s41598-019-48511-8

[9] https://lib.dr.iastate.edu/cgi/viewcontent.cgi?article=1260&context=ent_pubs

[10] VHF Radar Based Telemetric tracking of insects. https://atstrack.com/animal-class/insects.aspx

[11] De Souza, P., et al. Low-Cost Electronic Tagging System for Bee Monitoring. Sensors (Basel). 2018 Jul; 18(7): 2124. doi: 10.3390/s18072124

[12] Barlow, S.E., O’Neill, M.A., Pavlik, B.M. A prototype RFID tag for detecting bumblebee visitations within fragmented landscapes. J Biol Eng. 2019; 13: 13. doi: 10.1186/s13036-019-0143-x

[13] https://gtr.ukri.org/projects?ref=NE%2FL012359%2F1

[14] https://www.noldus.com/track3d/insects

[15] WIMEA-ICT Project. https://wimea-ict.net/

[16] https://www.raspberrypi.org/products/raspberry-pi-4-model-b/

[17] https://www.seeedstudio.com/Raspberry-Pi-High-Quality-Cam-p-4463.html

[18]https://www.bluedesigns.com/products/snowball-ice/

[19]http://dx.doi.org/10.5061/dryad.20ch6p5

[20] http://idigbio.org

[21]https://www.ipmimages.org/

[22] https://data.mendeley.com/

[23] Redmon J, Farhadi A (2016) YOLO9000: Better, Faster, Stronger.

arXiv:1612.08242

[24] Yi Lin et al. Focal Loss for Dense Object Detection (2018). arXiv:1708.02002v2