Africa RISING ESA Project Partners meeting 1 July 2021

From africa-rising-wiki
Jump to: navigation, search

Africa RISING ESA Project Partners Meeting
1 July 2021
Virtual via Ms TEAMS
[edit | edit source]


  1. A. Kimaro, ICRAF
  2. B. Zemadim, ICRISAT
  3. C. Azzarri, IFPRI
  4. C. Thierfelder, CIMMYT
  5. D. Mgalla, IITA
  6. F. Kizito, IITA
  7. F. Muthoni, IITA
  8. G. Fischer, IITA
  9. J. Groot, WUR
  10. J. Manda, IITA
  11. J. Odhong, IITA
  12. L. Claessens, IITA
  13. M. Bekunda, IITA
  14. M. Mutenje, IITA (consultant)
  15. P. Okori, ICRISAT
  16. R. Chikowo, MSU


  • Pushing the boundaries with spectroscopy for precision agriculture on heterogeneous farms- Drs. Regis Chikowo (MSU), Sieg Snapp (MSU) and Dan TerAvest (OurSci)
  • Rip tillage: a gendered village case study in Kiteto, Tanzania- Dr. Gundula Fischer (IITA)

File:ChikowoSnappTerAvest Seminar 1 July 2021.pptx - Drs. Regis Chikowo (MSU), Sieg Snapp (MSU) and Dan TerAvest (OurSci)

  • Download presentation from link in the title above.


  • Francis. Regis, it’s better to apply iSDA soil layers with 30m resolution (farm-scale) instead of SOILGrids with 250 m.
    • Sieg. Francis, we used the isDA soil layers which are reported to be 30 meters. Over that, there is none on sampling is based on report sensing and some updates from sampling. Also, we haven’t found a much higher resolution. However, we were preparing the paper on that. It is a good point, and of course, as we go forward and now that isDA is available, we are using that in our correlation and as the way to help. Also, the handout supplements extend and improve our ability to support the decision on the ground, but it compliments the IsDA. Besides, remote sensing is internet-accessible. Similarly, the soil grade types information is at different resolutions even if it says in 30 meters it based on interpolations. It needs proximal local information to grate it up.
      • Francis. Regis,why do you assume a linear relationship between spectra and soil properties?, also, It will be interesting to know the RMSE of the linear predictions. Probably non-linear model will be more intuitive. The samples 1156 seem few considering the high spatial variation of soil properties.
        • Sieg. Francis the liner prediction has been used and literature and seems to have been working very well.
      • Francis. Referring to the last point relating to the linear relations, I thought if we look at the work done by Hegu and the team, the study would better be done by non-linear relations. It requires many soil samples from different agronomical ecology zone/ soil profiles/types of soil. Then, you can do the non-liner model or machine learning. Non-liner relations have proven to be much better than assuming liner-relations which is very simplistic.
        • Regis. Thank you, Francis. Different models can be deployed for final predictions, but we trained the, with more than thousands soil. 50 and 70% of them were used to get the model. We also tested 30% of those based on the data from the lab. So, we have a good level of confidence because we have physically and chemically analyzed the soil, we used thousands of spans.
  • Sieg. To add to Regis’s statement, we found a better correlation than the key spearheaded mid-range infrared specula, which is technical. The mid-range infrared specula should be better in predicting carbon because specula residence is closely related to carbon. But, because our training was hyperlocal, we complemented other efforts than the continental scale. It is one of the first in-expensive hyperlocal because we trained them to hyperlocal both Tanzania and Malawi. The prediction is around 80% compared to conventional dry carbon. This is good, and it is better than the key spearheaded specula from our same soil sample, and it is working for our conditions (very dry field condition).
  • Lieven. About reflectometer: how do you consider differences in the field vs lab conditions (where calibration is done), especially concerning moisture content (which can influence color/reflectance/'translation' to SOC a lot....)?
  • Regis. Reflectors are heavily influenced by moisture content that absorbs a light if there is the right wavering, so we normalize with dry soil throughout to make the condition similar.
    • Sieg. To add to Regis's statement, due to the dry field conditions, sampling from the same semi-arid tropics is collected. Since we have long dry seasons, it is perfect for changing in the field measure, and then we are calibrating it with the samples that we are taking to the lab, but the calibration is around 85%, very linear and predictive. So, they are good enough for what is being done, and it will keep going better.
  • Mateete.Who owns "Our Scikit" and when would it be available to development partners?
  • Sieg. Google - this the lab developing these sensors and an associated electronic survey kit like - this is all open source - you can produce your own using the info on the website or if anyone is interested in purchasing for about $400 each (keeping in mind that a large effort including 100s of samples over a year of calibration will be required). Also, we have some reflectometers in Malawi and would be happy to do training - and advise on calibration.

File:ESARipTillagePresJuly2021.pptx - Gundula Fischer (IITA)

  • Download presentation from link in the title above.


  • Mateete. Gundula, can some farmer observations, e.g. rainfall trends, be validated by available or national rainfall data records?
    • Gundula. Yes Mateete, we can do that. It is not just the section, but we will relate this to a rainfall trend such as recorded in national statistics.
  • Mateete. In the previous presentations, you have been linking gender to all the SI domains, when you are presenting, for example in this the ability to experiment how is it being influenced by other domains, the differences in agro-ecology between the two villages discussed, whether they are equally accepted in term of RIP tillage. Would such a thing affect gender responses?
  • Gundula. I am not sure there are interrelations at this point where you look at the agro-ecology. Swai started in Njoro and went to Kiperesa, and his first choice was Njoro, although there was no implement. Later he went to the village where there were implements and the other NGO works. I assume both ecologies are equally suitable, but it would be better Swai to provide more details.

Interactions with other SI domains for example, at the social domain; what happens in the community level if implements are shared, one could look at what men and women can produce that could influence what happens in terms of the capacity to experiments in terms of capturing opportunities to produce and to sell. But usually, yes, I mainstreamed this through all the domains, and this will be done more strongly in the development of the paper on this village case studies.

  • Sieg. Gundula. How or in what way to envision better feedback as we go forward to incorporate this kind of gender-aware and continue to adopt, support, and refine options for farmers. How do you envision it to happen in this case, and what might be next?
  • Gundula. With the Africa RISING project, we are almost at the end but, what can be done in the future, the kind of bundling of technical technologies? We should very early integrate social matches to turn this bundle into more social technically bundles meaning, we have gender awareness training is provided together with the more technical training of farmers. Other cigar centers have started to pilot these approaches. Moreover, measuring how far gender awareness influences the development outcome, we change corporations in terms of labor and other aspects in the long run. We should start much earlier, and we should use SIAF more as a planning instrument to incorporate more of the social element and monitor what is going and adjust as early as possible.
  • Comment. Mateete. Presentations from the seminar will be converted and presented in the form of a manuscript.