Where does all that feed come from?

The environmental trade-off model, i am working with lately, is based on the idea that feed can move within the whole study area! A strong assumption?So in the last field trip to Burkina Faso, we investigated the matter.

In fattening and dairy systems in peri-urban Burkina Faso, feed is mostly brought to the animals. Most of the feed is crop-residue, from maize, sesame, sorghum and other local crop. Most of the farmers have their own, but also buy the crop residue from their neighbor, or the buy the right to graze on the field with residue.

maize residues prepared for sale to the neighbor

There is also some natural grasses that can be found on common land, but some farmers also get grass from privately owned land for which they got the usage right. Most farmers pay for these right, but we met one who got grass for free.

Natural grass dried and stored for the dry season

Next to grass and crop residue, farmers feed other local feed, from local tress but also a whole bunch of concentrates. Firstly, around Bobo-Dialassou is providing a full concentrate produced in Cote d'Ivoire at subsidized price.

The concentrate from Cote d'Ivoire

Secondly, because of the area produces cotton, the cotton seed are transformed to oil. The residue of that oil production is called cotton seed cake. Though it is produced locally, the high fat content makes cotton seed cake a high value feed with a regional market. The cotton seed cake from Bobo can be found up to Mali, where the price are better.

Sesame ready for seed extraction, afterwards the residue is fed to animals

The state company that produces oil and cotton seed cake at high quality has specific rules about cotton seed cake. Their oil is more expensive than the one of the competition. So the oil is always sold with the cotton seed cake, which can fetch a higher price. So only the one who buys the oil can get the cotton seed cake (that is of better quality than the one of the competition). So the big oil buyer will also develop a market for the cotton seed cake. Because prices are so much better in Mali, the oil buyers usually sell the cotton seed cake abroad.

Cotton seed cake from the state company (left)

Cotton seed cake from the local companies

In conclusion, feed moves around quite a bit within the study area but also move out of the study area for animals in drier locations such as Mali.

Moving feed out of the study area

urban livestock fattening : an adaptation strategy?

During my last field trip in Burkina Faso, we stopped in Bobo-Dialassou, the second biggest city in the country to talk to a livestock keepers association for fattening.

meeting the livestock keeper association

They are 70 male members who specialize into livestock fattening. After longer discussion we also discovered that some of them are also dairy farmers and other still have big herds out of town. Yet, our discussion focused mainly on the objective of the farmers' association, namely fattening that takes place in town.

urban livestock keeping in Bobo-Dialassou

Livestock is kept in one location and all the feed and fodder is purchased. It is a completely land less activity, that for some is a risk management strategy (for those who have still herds far away) and a fully sedentary activity for other (how gave up traditional livestock keeping and focusing on activities in town).

Feed for sale

The association main role was information sharing, they have to head of information who get airtime from the association and have responsibilities to call for meetings. During meetings the all new information about markets, be it for input or for sale of animal is exchanged. Presence at these meeting is obligatory, but one can get excused for a very important matter.

Also the association buys feed in bulk for its member to reduce cost. The major objective of the association is to expand into the periphery where fattening also happens.

Follow the cotton!

I was quite exited on my last field trip to Burkina Faso to see for the first time in my life cotton growing in the field. We were visiting during the harvesting season, so suddenly one sees white mountains of cotton appear in the landscape.

the mountains of cotton

I was pretty astonished to learn that it was all rainfed, that it is an integral part of crop rotation in the area. Whereas farmers in the West of Bobo-Dialassou are well organized and have an organized supply chain to the cotton companies, around Bama this is not the case. This means that only farmers that are big enough to attract a cotton company to come and pick will produce cotton.

A cotton flower before harvest

Cotton is grown for its white fiber that we use to produce cloth. Yet, the flower also has seed that can be used to produce oil. When the oil is refined it is good for human consumption. The cotton seed cake is the waste resulting from oil production. It has high fat and is therefore a very good feed for livestock.

An open cotton flower with the seed in the middle

We visited Madame Kagone, the wife of a doctor who produced oil is an artisanal way her whole life. But when we met her, we discovered that she was building a quite big cotton seed oil production plant. We never discovered how it was really funded, but we were told that the goverment was becoming very restrictive and would check that the oil production for human consumption comply all the refining steps that makes oil edible for humans. So there was a clear move from artisanal production towards middle sized plant.

the refinery part of the oil production plant

So we visited such a plant, which still was under construction. We discovered that the cotton seed were bough from the state company that processes the cotton fiber. Oil and cotton seed cake are two separate products and is sold to re-sellers, and so we lost the trace on whether the cotton seed cake remained local or not.

Madame Kagone on the right with her friend who is also building an oil processing plant

We left, impressed that these type of businesses are run by women...

Exploring Bama : when urban demand offers new opportunities

I am just back from a field trip in Burkina Faso, where we went to explore the context in one of our sites for the CLEANED project funded by the DFID SAIRLA program. This project aims at identifying trade-offs and synergies in intensifying livestock value chains.

We went to Bama, in the periphery of the second biggest city of Burkina Faso, Bobo-Dialassou.


Bama's landscape is mostly shaped by four land uses, irrigated rice production, rain-fed crop production (mainly maize, sorghum, sesame and cotton) with trees (mango, papaya, cashew nuts) and natural savanna and forest, as shown in the land cover map below.

 We discovered that decision makers differentiate among three livestock keeping systems :

1. fattening system 
2. dairy system
3. traditional system

the rice paddies in Bama

In the fattening system, cows are kept for producing milk at the homestead and fed in a cut and carry system. In the dairy system, cows are mainly at the homestead for milking and feeding. However these cow also go grazing in the nearby fields and grazing area. Finally, in the traditional system, cattle is not kept for a particular purpose (meat or milk) and goes on transhumance for part of year, i.e. goes beyond a 100 km away from their homestead.

All farmers also have sheep, that might join the transhumance or not. Also pigs are becoming more and more important.

Bama, due to its proximity to the big city has the opportunity to supply the growing demand for milk and meat. Many people choose to produce to fulfill this demand and the amount of animal that remain the whole year long is increasing, increasing the challenge for feed in the dry season and is changing the landscape.

Wondering what is driving these changes? how people see their future in the area? of simply want to discover a new place? Then follow the SAIRLA and Burkina Faso tag in the up-coming weeks to discover what we have learned in Bama!

R : enabling the big data revolution?

This week, i am blogging from the Biological Control and Spatial Ecology group at the Université libre de Bruxelles, where i am spending about 8 days learning how to optimize coding to automatize the creation of databases.

Sounds Chinese? Attentive followers will have noticed the many blog posts around big data lately. If big data means bringing many different unorthodox data sets together to explore correlation, then we need someone that can integrate the different data sources into one database for analysis...
So let us look at my challenge : i want to extract DHS (Demographic health survey) from about 50 countries, which have similar variables that have been coded slightly differently, which means if you simply exact the data you cannot put it together easily, because one will write Poorest, the other poorest and another one might spell it differently... So is my only choice to everything manually, i.e. repeat the procedure 50 times (which might be a source of error)? Or are there tricks and tips on how to automatize data extraction and integration in a big database?

After my first week, the conclusion is that there are things you can automatize and others you have no choice but digging into the data manually... but you can optimize the whole process a lot, and code in a way that next time you need similar data, it becomes much easier... The R software with its whole suits in R-studio offers a very flexible and efficient coding environment to implement this. I have started coding in R about two years ago, yet the coding smartness i am discovering here is amazing!

Wanna follow my R journey? Follow the R! page here on my blog. It will collect all the small and big tips to make use of R for big and small data!

Big Data to end poverty?

Lately, i have written a whole series about Big Data.  I came across the World Bank movie about the importance of good data and how to make the big data revolution work for development. I enjoyed watching in and i hope you will too!

The tyranny of choice : comparing land cover maps for Uganda

Last week, i posted about land cover for Burkina Faso, and i came across this post that remained as a draft on my blog for half a year at least. Time to get it out there! Joanne Morris, a colleague of mine from SEI York did quite some work to compare land cover maps for Uganda. I thought this comparison might be interesting to others and asked her to write it up into this blog! So here it is!

Joanne Morris, from SEI York and the author of this post

I am privileged to have been invited to post on Catherine's blog, which I particularly enjoy reading for frank insights that are much easier to absorb and relate to than many of the dry research reports I have read elsewhere. A bit about myself - I am Joanne Morris, working for the Stockholm Environment Institute at the University of York, UK. I am broadly interested in natural resource use and management in semi-arid and arid regions, for balancing the demands for food, energy, livelihood and above all looking after the natural resource base so that we can continue providing for those demands into the future. I am especially interested in the use of biomass for agriculture, livestock and energy (e.g. as mulch, manure, compost, feed, fuel for burning, or biogas) - what is the most effective use of biomass to serve all three? And which form of biomass is best to use?

Anyway, back to the actual question at hand: during the year, we had to choose a landuse map to use as the base for the calculations we are doing for the Uganda pig value chain in Hoima. Perhaps you and others will be interested in the comparison of current landcover products out there that led to our final choice to use the SERVIR-Uganda 2000 Scheme II map.

From Google, a USGS list of Land Cover Data Links for Africa and a good review paper of global land cover maps (Gong et al 2013), I gathered the following list of possible candidates with some of their key features. We were interested in the resolution first and foremost - within even 1km in Uganda we saw very different landcovers, with fields mixed in with natural forest, unlike in Europe or the US where you can have fields of wheat of maize stretching for thousands of hectares. Therefore, ideally we needed a product with a higher resolution than 1km. Second was the date of the data - landuse changes quickly in Africa - as we found on our field trip, where large swathes of commercial forest have been planted in the last 10-15 years, so ideally something more recent that the last best global landcover layer which was the Global Land Cover 2000 layer from the JRC. Finally we were interested in how many landcover types were differentiated for Hoima - could we get cropland, bushland and wetland from it, as well as forest?

Landcover map	Resolution and satellite source	Year and projection of data	Coverage – Available for Uganda?	Landcover classes in Hoima
SERVIR – GHG Landcover Maps for Eastern and Southern Africa	0.000271dd (~30m) Landsat 30m – supervised classification using maximum likelihood algorithm on LandSat thematic mapper (LandSat 5, 7 and 8 – Uganda)	2000 & 2014 (Uganda) 2000, 2005, 2010 – other countries   WGS 1984	Uganda Land Cover 2014 Scheme I and Scheme II same for 9 (ESA) countries namely: Ethiopia, Botswana, Lesotho, Malawi, Namibia, Rwanda, Tanzania, Uganda, and Zambia.	Scheme II – 14 classes Scheme I – the 6 IPCC classes
FROM-GLC - (Finer Resolution Observation and Monitoring of Global Land Cover)	30m Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+)	mainly 2009-2010, filled with 2007, few 1998	Global	Level 1 –11 classes Level 2 –25 classes to combine with L1 predominantly forest, doesn’t differentiate in Hoima well enough
GLCN Africover (FAO)	na (shapefile) LANDSAT TM images (Bands 4,3,2)	mainly 2000 (Uganda) - shapefile, not raster WGS 1984	Africover Eastern Africa module, 10 countries: Burundi, Democratic Republic of Congo, Egypt, Eritrea, Kenya, Rwanda, Somalia, Sudan, Tanzania and Uganda	LCCS classification, 8 major landcover classes, can download all or selections e.g. agriculture
GLCN Globcover-SHARE (FAO)	30 arc second/ 0.00833dd (~1km) combination – see comments	varies by country – 1990 – 2012 WGS 1984	Global	Thematic layers: 10 LCCS classification
GLCN Globcover by country (ESA) - Uganda	~300m/ 0.00278dd, ENVISAT, 300m MERIS sensor	2005 and 2009	Africa	46 LCCS classes Cropland by lake
GLCF MODIS (NASA/ USGS)	raw tiles: 0.0041667dd (~500m) OR 5’ (~10km) or 0.5dd (~60km) – aggregated	2001 – 2012 (every year)	global – by tile: Uganda PN3536	16 classes IGBP classification
GLC 2000 (JRC)	0.00893dd (~1km) SPOT 4, VEGETATION instrument	2000 WGS 1984	World	22 categories - LCCS classification no wetland, cropland mainly by lake
IFPRI World Land use – Land cover	30 arc seconds, 0.00833 (~1km) based on a reinterpretation of the Global Land Cover Characteristics Database (GLCC ver. 2.0), EROS Data Centre (EDC, 2000)	2000  GCS_Clarke_1866	World	17 classes IGBP classification agriculture focus no wetland shown
IGBP-DISCOVER	1km, resampled to 0.25, 0.5 and 1.0 degree grids AVHRR	1992-1993  geographic	Global	17 classes, IGBP classification
University of Maryland (UMD)	1 degree, 8 kilometer and 1 kilometer pixel resolutions AVHRR	1981-1994	Global	14 classes
CORINE-land cover	Scale of 1:100 000	1990 and 2000	Europe	44 classes
AVHRR – Global Land Cover Characterisation	AVHRR – 1km	1992 – 1993	Global	7 classes
GOFC/ GOLD Land Cover Project office	coordinating agency

    *dd = decimal degrees; GLCF – Global Landcover Facility; GLCN – Global Landcover Network; JRC – Joint Research Centre; AVHRR - Advanced Very High Resolution Radiometer; FAO – Food and Agricultural Organisation; ESA – European Space Agency; NASA - National Aeronautics and Space Administration; USGS – US Geological Survey; IFPRI – International Food Policy Research Institute; IGBP – International Geosphere-Biosphere Programme; LCCS - ; GOFC-GOLD – Global Observations of Forest and Land Cover Dynamics

Here are examples of how a selection of the most recent layers look for Hoima. SERVIR came out quite clearly as the most useful because of its high resolution (30m) and the landcover classes resonated with what we saw on a round route drive through the district - that much of the area around Hoima town is mixed smallscale crop and bush/ forest land, while towards the south it gets drier and more bushland. Along the lake shore was very dry with no fields. The only landcover missing is the large commercial pine and eucalyptus forests that we saw on our drive. It would be ideal if it was Africa wide so that we could very easily transfer our calculations to another country without worrying about the change in landcover maps, perhaps by the time we move away from the ESA (East and Southern Africa) countries, SERVIR will have expanded :)

SERVIR - Scheme 2, 2000

The SERVIR Land Cover maps were developed for Green Houses gases Inventories to provide baseline data for Land use, land-use change and forestry (LULUCF) sector for each country, and were developed with country representatives, informed by country specific interest, definitions, descriptions, mapping goals and policy statements and documents with guidance from IPCC Good Practice guidelines.
SERVIR Scheme I - 6 IPCC classes:  Forestland, Grassland, Wetland, Cropland, Settlement and Other land; Scheme II - 14 classes


SERVIR - Scheme 2, 2014
 

GlobCover - 2009, Uganda

GlobCover is an ESA initiative which began in 2005 in partnership with JRC, EEA, FAO, UNEP, GOFC-GOLD and IGBP, and incorporates input from the global community. The aim of the project was to develop a service capable of delivering global composites and land cover maps using as input observations from the 300m MERIS sensor on board the ENVISAT satellite mission. ESA makes available the land cover maps, which cover 2 periods: December 2004 - June 2006 and January - December 2009.
We decided not to use it because it is not as detailed within Hoima as SERVIR.


GlobCover SHARE
 The GlobCover SHARE dataset was created by FAO-Land and Water Division, in partnership with various institutions, and makes use of several sources (GLC-Share report ):

• Global Landcover Datasets: Globcover 2009 (MERIS 300m), MODIS VCF (Vegetation Continuous Fields), CROPLAND Hybrid database (mixed resolution), GLC2000 (SPOT Vegetation 1km), Mangroves (Landsat 30m, FAO Global Database of Mangroves)
• By country – variety of satellite sources, including: Landsat 30m, MODIS 250m, FAO LCCS, SPOT 10m, AirPhotos 1m, Ikonos 4m – majority is Landsat 30m 

11 classes: 1: Artificial Surfaces; 2: Cropland; 3: Grassland; 4: Tree covered Area; 5: Shrubs Covered Area; 6: Herbaceous vegetation; 7: Mangroves; 8: Sparse vegetation; 9: BareSoil; 10: Snow and glaciers; 11: Water bodies
 We decided not to use it because of the coarse resolution (1km).
 

GLCN - Africover
Cultivated-agriculture layer highlighted above the general landcover layer .
The Africover map is highly detailed, however, the legend is quite confusing to decipher, and therefore SERVIR was chosen for ease of handling.


IFPRI 
We chose not to use this map as the data is a bit old (2000) and the resolution too coarse (1km). 


MODIS
Although very good for time-series, because it is available for every year at quite fine resolution (500m), is it still not as detailed as SERVIR, so we decided not to use it. 


GLC-2000

GLC 2000 makes use of the VEGA 2000 dataset: a dataset of 14 months of pre-processed daily global data acquired by the VEGETATION instrument on board the SPOT 4 satellite, made available through a sponsorship from members of the VEGETATION programme, including JRC.
As the data is from 2000, it is too old for our purposes. 


FROM-GLC
Although great for a forest/bushland study, we are more interested in the cropland and wetland landcovers, so we decided not to use this map.


I was pleasantly surprised and impressed at the range of data out there - openly accessible - with varying characteristics to suit different purposes. At national level there will no doubt be many more! The GlobCover SHARE project will be one to watch - the 2014 beta is out, although I did not yet find the data. And here's to hoping SERVIR continue expanding their maps to more countries! Merry Christmas all.
 
On a side note, I would have done a follow-up comparison of precipitation maps out there, but NCAR got there first, with a comprehensive set of pages about each of the main datasets available here.

I would like to thank Joanne for having shared her comparison here. The project we have been working on settled to work with the SEVIR map 2000 scheme II. The decision was based on a trade-off between high resolution (30m for SEVIR) and accuracy tested with our validation point from the field.