Show me that good grass: a call center for pastoralists in Northern Mali
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
Title |
| |
Alternative Title |
| |
Title of Series | ||
Number of Parts | 295 | |
Author | ||
Contributors | ||
License | CC Attribution 3.0 Germany: You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor. | |
Identifiers | 10.5446/43561 (DOI) | |
Publisher | ||
Release Date | ||
Language |
Content Metadata
Subject Area | ||
Genre | ||
Abstract |
| |
Keywords |
FOSS4G Bucharest 201923 / 295
15
20
28
32
37
38
39
40
41
42
43
44
46
48
52
54
57
69
72
75
83
85
87
88
101
103
105
106
108
111
114
119
122
123
126
129
130
131
132
137
139
140
141
142
143
144
147
148
149
155
157
159
163
166
170
171
179
189
191
192
193
194
195
196
197
202
207
212
213
214
215
216
231
235
251
252
263
287
00:00
Open setGrass (card game)Call centrePower (physics)CASE <Informatik>Presentation of a groupAdaptive behaviorProjective planeBitMusical ensembleDecision theoryContext awarenessGrass (card game)MereologyDigital photographyTheory of everythingOnline help
00:43
Software bugConvex hullDecision theoryMereologyLecture/ConferenceComputer animation
00:59
Software bugInclusion mapMusical ensembleSoftware bugWater vaporFamilyRight angleInsertion lossMiniDiscDigital photographyComputer animation
01:48
Water vaporVideo gameInsertion lossComputer animationLecture/Conference
02:04
Maxima and minimaWater vaporMedical imagingProduct (business)InformationMultiplication signSoftware frameworkProjective planeSet (mathematics)Network topologyContext awarenessMereologyECosSinc functionTime seriesCall centreQuicksortRegular graph
03:28
Set (mathematics)Self-organizationWater vaporTelecommunicationLogicSoftware developerMultiplication signGroup actionFigurate numberMereologyProjective planeAngleBoss CorporationShared memoryData storage deviceSpacetimeMeeting/Interview
04:21
Call centreWater vaporInterface (computing)Field (computer science)Call centreConcentricSystem callDecision theorySoftware testingMultiplication signComputer animation
05:22
Programmable read-only memoryWebsiteTerm (mathematics)Computer clusterGame theoryCondition numberEmulationTime zoneProduct (business)BitPoint (geometry)AverageFunction (mathematics)Medical imagingTerm (mathematics)Query languageCall centreTranslation (relic)DatabaseAxiom of choiceInformationTime seriesNormal (geometry)Goodness of fitSpeciesSet (mathematics)MassData storage deviceComputer animation
07:08
WebsiteTime zoneBeta functionHydraulic jumpGoogolDialectAverageInstance (computer science)Software bugTerm (mathematics)Goodness of fitPressurePhysical systemTranslation (relic)Computer animation
07:35
TouchscreenCall centreComputing platformSystem callMilitary operationToken ringInformationOperator (mathematics)BitSystem callSmartphoneWater vaporMultiplication signInstance (computer science)Computer animation
08:26
Cartesian coordinate systemInstance (computer science)SmartphoneIP addressBit rateWeb 2.0Computing platformLecture/Conference
09:02
System callMilitary operationComputing platformCall centreImage resolutionFrequencyWater vaporSoftware bugFile archiverCuboidWater vaporTranslation (relic)Analytic continuationBitProduct (business)Software bugHand fanSet (mathematics)Term (mathematics)Computer animation
10:16
WebsiteLevel (video gaming)Condition numberWater vaporInstance (computer science)InformationTerm (mathematics)Cellular automatonField (computer science)AverageIRIS-TQuantileUltraviolet photoelectron spectroscopyMathematical singularityElement (mathematics)System identificationVideo trackingExtension (kinesiology)Computer clusterSlide ruleSlide ruleCASE <Informatik>Computer clusterTwitterProcess (computing)Call centreField (computer science)Service (economics)System identificationQuery languageState observerPhysical systemInterface (computing)GeometryMultiplication signProduct (business)Water vaporSpacetimeInformationSystem callAuditory maskingDivisorObservational studyVector spaceComputer animation
12:16
FrequencyInformationFeedbackSystem callLecture/Conference
12:33
Computer clusterInterface (computing)TwitterSlide ruleInformationWater vaporValidity (statistics)Product (business)Call centreFrequencyQuicksortCASE <Informatik>Projective planeOperator (mathematics)FeedbackSystem callInformationBusiness modelKey (cryptography)Entropie <Informationstheorie>Presentation of a groupCoefficient of determinationFlow separationMoment (mathematics)Streaming mediaComputer animation
13:37
Call centreComputing platformToken ringMilitary operationSystem callSystem callStreaming mediaThermal expansionOperator (mathematics)Moment (mathematics)PlanningProcess (computing)InformationWebsiteLecture/ConferenceComputer animation
13:59
System callToken ringMilitary operationComputing platformCall centreProcess (computing)Data conversionThermal expansionOperator (mathematics)Regular graphProduct (business)MereologySystem callInformationWeb portalInternet service providerTheory of everythingComputer animation
14:36
System callMilitary operationComputing platformCall centreArithmetic meanAreaBit rateCASE <Informatik>Presentation of a groupCall centreType theoryLecture/ConferenceComputer animation
15:16
MereologyAreaSource codeCall centreMultiplication signData conversionBitGeometryNumberInstance (computer science)Time zoneLevel (video gaming)System callMusical ensembleCuboidOpen setProjective planeLecture/Conference
16:46
Call centreWater vaporCovering spaceProjective planeAreaCuboidFeedbackOpen sourcePresentation of a groupComputer animation
17:14
Interface (computing)Set (mathematics)Lecture/Conference
18:07
Projective planeCall centreLimit (category theory)Context awarenessInstance (computer science)MassMereologyFile viewerFocus (optics)DemosceneTime zoneComputer animation
20:01
Call centreAreaSystem callRevision controlMaxima and minimaMultiplication signInstance (computer science)Lecture/Conference
20:33
Revision controlMultiplicationSign (mathematics)WebsiteComputer animation
20:59
Perfect groupSign (mathematics)Data structureComputer clusterLecture/ConferenceComputer animation
Transcript: English(auto-generated)
00:07
Thanks for coming, my name is Alex Orenstein and I'll be talking to you today about a project called Sustainable Technology Adaptation for Malian Pastoralists. Can everyone hear me alright? Great.
00:24
Is this another microphone? Is this better? Whoa. Too much. So yes, in case you're disappointed, the grass that we're talking about refers to rangelands, not marijuana. I'm sure there's another presentation that you can go to.
00:42
But we're going to be talking about specifically how we use EO data, Earth Observation Data, remote sensing data, to help pastoralists, these are herders, make decisions. So to give you a bit of context, we're talking about herders in the Sahelian part of West Africa. This is the band that stretches from Senegal to Chad.
01:03
It's very dry and it's very big. You can see on the right side here, precipitation anomalies. This is showing that rainfall is getting more and more erratic. So for people like this, this is a picture of a Fulani family in northern Senegal.
01:23
Finding water and pasture can be particularly difficult. Here we see a family filling up an inner tube with water to bring back to their herds. So the appearance of naturally occurring water and pasture is of critical importance to herders. They have to move with the seasons and they have to migrate their cattle and all of their livestock to find pasture and water.
01:48
If they can't find pasture and water, this leads to starvation of animals, this leads to loss of livelihoods. But I want to impress upon you how difficult it can be to find pasture and water and how critical being able to find it is.
02:05
Here you see the same lake. This is in the centre of Mali in Mopti. On the left, this is the wet season. This is September. On the right, that's May. You can see the three trees show you it's the same lake. This is not a drought year. This is not a particularly difficult year.
02:22
This was a regular year from September to May. So if you don't have information ahead of time about where the pasture is and the dry season comes around, as a herder, you could be in serious trouble. Another image, this is in the north of Senegal and this is one of my favourite places to look at in time series because it's a lake.
02:46
And we can see here this is the dry season. It's empty and then boom, the rains come. And there's a tonne and it quickly shrinks as the dry season sets on. And now we're in March, April, May, June and it's empty. And this is quite a large lake.
03:01
So again, just to provide you context about why it's important to find this information. So this was the framework in which this project was launched. This was launched in 2015 and I've been working on this project since then as part of the GIS team.
03:21
And the idea was to create a call centre that herders could call into and figure out using EO data whether or not there was pasture or water in a place. And then you could apply all sorts of other data. And the idea was to take these data sets which have existed for a very long time. NDVI and rainfall and small water bodies data sets are not a new development.
03:46
They've been around for decades. But to put it in a medium that could be accessible. So the Netherlands Space Organisation provided the initial funding. And there was a consortium with SNV, Hofsluit Spatial Solutions, a Dutch GIS company.
04:04
Orange, the telecom company. A pastoralist organisation called Tassart. And then a French NGO as well was collaborating Action Contre Laffan. And here's the basic logic of how it works. The herder here in Gao, this is the northern part of Mali where the project is taking place.
04:24
Decides he wants to bring his animals to Tessit. It's a village south of the river Niger in Mali. And he wants to bring his cattle and needs to know how the pasture is. Now normally he could try and find out if he has cousins or friends or someone he knows over there. Or send someone on a motorcycle which can be very costly and time consuming.
04:45
So instead he calls the Stamp Call Centre. They get this nice fancy imagery. Some field data on prices, herd concentration. And then they load up the interface.
05:01
And the agent receiving the call has a look and he sees this big red splotch over Tessit. And so he tells the pastoralist, it's not so great. Maybe don't go to Tessit. So this is the interface, what it looks like. And I think instead of doing this actually I'm just going to pull it up.
05:22
It's accessible online. It's stampmali.org. And basically we will do a bit of a role play. I want to bring my cattle to Tin Hamma, which is a village in Mali.
05:42
Tin Hamma was a poor choice. We are going to do instead Doro. So what happens is the NDVI is available here.
06:02
So the call centre agent who is manipulating this can open the legend. And unfortunately it's in French. But I can give you a translation. Bad? Good. It's basically what you need to know. So they can look around and this is the last Sentinel image.
06:22
And there is a textual output. It performs a query based on the point. And it figures out where is the closest water body that was detected from Sentinel to imagery. How does the biomass production this year, how does the dry matter productivity compare from this year to the long term average or to last year?
06:44
And then we can get a nice little time series. Here. Normally we have prices of cattle also loaded in. So there is a post GIS database that stores all of this really cool price information by species.
07:00
But unfortunately right now we actually don't have it. But we will soon. And then they can switch around. They can look at for instance the biomass anomaly. So how much pasture was produced this year compared to the long term average. Again, basic translation, bad, good.
07:25
That's roughly the gross overview of how the system works. So to go back. To give you a little bit of information about the center.
07:42
In the first year of operation, 2017-2018, there were 55,000 USSD requests. And 1,700 calls. I was actually unable to get call data from the last year. But we were actually pretty happy.
08:00
We were quite happy with how the pilot went. And the calls are taken in Sonrai, Arabic, Tamashek and Pul or Fulani. What's really interesting about this is that it provides a lot of information that technicians have had access to for quite some time.
08:21
Pasture, water, price data. And it makes it in a very, very accessible medium. There is a lot of work being done for instance in developing applications. But the smartphones can access to provide meteorological data to farmers and herders. You see a lot of that happening across the continent.
08:42
But in a place like Northern Mali where you don't have widespread 3G, this solution oftentimes isn't really applicable. Likewise if you're dealing with a population that has low literacy rates compared to the rest of the country.
09:01
And then as for the web platform, what I just showed you, it's had 27,000 NIC IPs and 774,000 hits since 2017. So it is being used, which is very nice. And then to give you an explanation about the data, we use Sentinel-2 for NDVI and Proba-V for dry matter productivity biomass.
09:25
Proba-V is a continuation of the Spot vegetation satellite, which some of you might have used. It's a very special place in my heart. I'm a very big fan of this dataset because it has a continuous archive since 1998. So what's really nice is you can develop long-term, you can look at anomalies over 20 years.
09:46
And in a place like the Sahel in Northern Mali, this is particularly important. It's quite rare that two years will look the same when it comes to vegetation growth. It's very much a roller coaster. So being able to have a long-term archive is very handy.
10:04
And then for small water bodies, these little blue bits here, that is also Sentinel-2. This is a translation of that text box that you saw coming up. And it has basically every paragraph and then where it comes from.
10:25
So the first paragraph, water bodies, Sentinel-2. Then it performs a query on dry matter productivity for Proba-V. And then here, the closest farmland is 18 kilometers to the northeast. And that's vectorized field data that was drawn in.
10:42
And then the quality of the pasture is no data. Again, really unfortunate timing on that. But that's a post-GIS dataset. It's also quite nice that we can combine this field data and Earth observation data. I'm quite happy about that. And then the next steps, where we're going through here,
11:02
is we will be developing a semi-automated system for cropland identification. So this is quite important. Right now we have vectorized field data that we've collected, but GAO is enormous. And if you think you're going to get on-the-ground field data for the entirety of GAO,
11:23
it's very ambitious, but it's just not going to happen. So we are currently developing some cropland masks, because in that interface, it is very important to be able to distinguish between cropland and pasture. Not all of this green space is going to be accessible. And then we're in the process of tracking dust storms as well,
11:43
and extending the service generally across Mali. So in case you want more, you can contact me, at orin underscore SA on Twitter. These slides are available at tinyurl slash cow call center. And the interface is open.
12:03
And the biomass data is available at geo.info, where you can play with it, download it. And yeah, I suppose opening up if we have time for questions. So thank you.
12:28
So after the farmer calls the call center, when you arrive on the place you want to go, can you give feedback to the call center? Can you call back and say your information was correct or not?
12:43
During the pilot period, that definitely happened. There was a period where there was a sort of initial validation that was being taken before the product was put into commercial operation. And during that period there was a lot of back and forth between the herders and the call center employees.
13:07
Thanks for the nice presentation. So I think this was a G4AW project, is that right? Yes. So the business case should also be very developed then. So how do you sustain this?
13:21
So there are these calls, they have a price, and there are the U.S.S.Ds, I guess they are for free. How do you close the business model here that is sustainable? Because your project is already finished, G4AW is still continuing? G4AW is finished. Right now there is another funding stream from the Dutch Ministry of Foreign Affairs,
13:43
which is funding the expansion. But as for the moment, the calls are 75 SEFA a minute. So that's about 11 cents. And the U.S.S.D. calls are free for the moment, but I am unsure if they are going to stay free.
14:00
I don't have that information. So that is basically the plan, is that with an expansion it will be sustainable. The operating costs, at least on the side for Orange, are quite low. And a lot of these processes are being automated, so removing a lot of the human resources costs.
14:21
And it is reducing the cost, so the hope is to find a conversion as more calls come in. And the products of FITO, FITO was part of the partnership, or did you just use their regular portals for this, for ProbaV? This is just the public Copernicus data that FITO produces.
14:42
Anything else? Cool. Well, thank you very much. So there's that one village that didn't work.
15:01
How do you deal with the gazetteer issue in this type of area? How to find that village? I mean, it's possible the person at the call center doesn't even know where that village is, right? So, yeah, how do you address that? In this case, it's reducing the possibilities. So you figure out, okay, well, what's the nearest village that we know?
15:24
In here, you know, because if it's, for instance, not an open street map, if it's not geonames, if it's not from any of the number of sources where we've pulled the village names from, it's going to be tough. Luckily, the call center agents, for the most part, have a pretty solid understanding.
15:40
They're all from herding backgrounds themselves, and they're all from the area. And typically, like when you're working with a population who are mobile, and have geography built into their livelihoods, usually being able to find the place you're talking about happens after a matter of conversation.
16:05
So it does require a little bit of extra time to be like, okay, well, where is that near? What are we looking at?
16:22
Where are the call centers located? So the call center is in Bamako. While it covers Gao and everyone working in the call center is from Gao, it's in Bamako.
16:41
It's a big stretch that we showed on the map. Oh, that was just to show the Sahelian band, like the zone, the agroecological zone. The coverage area currently is the purple box. That's where the project covers. So it's only Mali. Yeah, it's only Mali. It's only Mali for the time being.
17:03
Hi. Sorry I'm late. Great presentation, though. So what's your greatest ask or request of the open source GIS community? I mean, we would definitely love feedback. We would love it if people would be able to play with the datasets, play with the interface, let us know.
17:27
And then obviously, as well, there's definitely a need for new datasets. You know, there's a whole bunch of much smarter people than me currently at this conference who can come up with really great ideas,
17:41
I think, for other data that could be integrated relatively painlessly. So I think, honestly, yeah, just being able to look at the interface, look at the data, tell us what you think.
18:02
I think I saw some satellite edges. So is it something for improvement? Well, you mean like the edge of the scene? Yes. Well, so here, it's limited based on the project scope. So this isn't a limitation here caused by the satellite, it's the project scope.
18:24
And these have already been pre-processed before they're put into the viewer. So this is, for instance, a WMS that's generated through GeoServer before it's put on here.
19:12
Thank you.
19:26
Or also agricultural croplands in, I don't know if in this project, but maybe in other contexts? Well, for this project, it definitely has a focus on herders.
19:40
And there is definitely a lot of work being done on cropland, but really as how best to avoid croplands. Like how we can make a mask so that if we see, for instance, a zone that has a nice NDVI peak, it's not someone's farm. And one of the important parts of this project is that it has to have a do-no-harm aspect.
20:04
So the call center agent, for instance, does not want to encourage pastoralists, for instance, to go into an agricultural area. And also, at the same time, the call center agent also cannot tell the pastoralist where to go. The pastoralist can't call in and say, okay, where's the best pasture?
20:23
And then the call center agent says, that can't happen because it has to minimize risk. That's great because putting herders going through agricultural lands or worked croplands, it's always a conflict there.
20:40
And this is also good that they know where to go and fix the... Yeah, I mean, that's definitely... This is a good side effect. Yeah, definitely. And the conflict right now happening in the center of Mali and Mopti between herding and farming communities is a pretty clear sign of why it's really, really, really important to make that distinction.
21:07
Perfect. Thank you. Thank you. I am very sorry. I had some little cow statues that I was going to give out, but I packed my bags very, very late.
21:22
So you're just going to have to come to Dakar and visit me to pick up all of your statues.