The implementation of tools was initiated by a training/expertise action on each of the project’s tools (LASER, NIRS and NDVI/GIS) carried out by international experts. The databases (knowledge) of partners have been identified and made available to the project. A common database was developed and shared via the web. This made it possible to characterize the main production systems in each country.


Portable NIRS for the evaluation of the chemical composition of fodder and organic fertilizers.

NIRS is an analytical technique based on the concept of electromagnetic radiation absorption by organic matter.

The rapid return of results to farmers and the low cost of analysis make these portable devices particularly interesting tools for decision-making to optimize the rational management of grasslands or forage stocks and to improve the quality of ruminants’ diet within the objective of controlling production costs.

The use of infrared spectrometry permits the evaluation of mineralogy, texture, contents of carbon, nitrogen and phosphorus, content of exchangeable cations, and the form and availability of phosphorus in the soil.

The practical implementation of a portable NIRS tool requires a preliminary calibration phase and the validation of results. The calibration of the tool requires the collection of a large number of spectral measurements in the field (forage and animal organic matter), coupled with chemical analysis carried out with traditional laboratory methods.

The portable NIRS device (LabSpec 4 STD 350-2500) and its accessories, as well as the chemometric software, have been purchased and used since September 2013 to acquire spectra of forage resources and animal organic matter. Its calibration was carried out with the help of two experts in spectrometry, Pierre Dardenne (CRA-W, Gembloux) and Serge Nabeneza (CIRAD – Reunion Island).


The medium infrared spectrometry technique (MIRS) is a very suitable tool for qualitative and quantitative analysis of soils in Madagascar. Like NIRS, MIRS it is a rapid, non-destructive and inexpensive method of measurement.

The characterization of mineralogy and texture will make it possible to identify the mineral composition of the soil, which controls to a large extent the fixation of the mineral elements and therefore the availability of nutrients. The determination of physicochemical properties (C, N, K, Ca, Mg content, cation exchange capacity, total and available P content) will give us an index of soil fertility.

The calibration of the MIRS tool was done in Madagascar through the collection of soil sample spectra from Madagascar and Mozambique. The biochemical analyses (laboratory) were continued in order to improve the calibration models of the infrared tools.


For the dynamic monitoring of forage biomass and its nutritive value.

The ability to produce forage biomass is closely linked to soil fertility and to environmental conditions. There are currently several models capable of predicting biomass growth from soil composition data (content of nitrogen, carbon, etc.) and climate data (solar radiation, temperature, precipitation, potential evapotranspiration, etc.). Initially, this action objective is to set the parameters and validate one or several soil-climate-dependent models, adapted to each country targeted by the action.

The satellite images allow the measurement of changes in vegetation activity during the different seasons and the data acquired by satellite in the visible and near-infrared zone allow the calculation of the Normalized Difference Vegetation Index (NDVI), which is highly correlated with the vegetation cover density and with the plant capacity to absorb sunlight and convert it into biomass.

In a second phase, the project aims to couple/correlate the NDVI indices with the geo-referenced measurements on the quantity and quality of forage biomass and the soil fertility evaluated on the plots and paths studied (activity 3). The multi spectral satellite images resolution of ten meters (SPOT 4 and 5) for Madagascar and Mozambique will come from the satellite receiving and processing station located on the site of the University of Technology of Saint-Pierre in Reunion Island.

The tool for tracking biomass will relate the growth models of soil-climate dependent grass and the geo-referenced information (NDVI index, meteorological data and field measurements) for mapping (Geographic Information System) the availability and the nutritional value of forage resources in an area. The tool will provide information in real time, which will enable farmers to better manage the number of animals and/or their herd mobility according to the expected availability of forage resources (itinerary, stocking). Once tested and validated, the tool can be used by support organizations and livestock counselling.


To improve knowledge of herds and ruminant livestock systems in areas where technical supervision is insufficient or non-existent, it is necessary to implement tools for collecting livestock zootechnical and economic information. The absence of sufficiently complete and reliable sources of information on animal production and health leads the researcher or the engineer to set up his own network of herd observation point. These tools make it possible to establish a diagnosis of the technical constraints of livestock systems and to provide technical and economic reference on the performance and economic efficiency of herds.

The LASER Software (Support Software for Monitoring Ruminant Production) is a relational management system developed in 1999 by CIRAD. Its aim is to facilitate the management and analysis of demographic, zoo-technical and epidemiological data, identified at the single animal scale within ruminant herds (demography, growth, milk production, health). This tool makes it possible to make an inventory of the productive performance of livestock systems and to evaluate the margins of progress. It contains various additional modules for carrying out adequate statistical analyses and producing different types of indicators which can be used in models of simulation of impacts of environmental changes for ruminant herds (climatic hazards, undernourishment, management changes, etc.). The chosen indicators require repeated observations on the same animals and in the same herds. This results in a specific organization of surveys, database, data entry forms and data extraction requests.

Here is the software’s website link, where the software is described and freely available to download:


3C-BIOVIS tool (Calculator of stocking capacity and plant biomass by satellite imagery) provides information on the availability of forage resources in real-time based on the acquisition of satellite images. This information makes it possible to advise on the mobility of the herd and on the management of its feeding according to the availability of fodder at the farm level and/or at the area level.

(Click here to access on the tool)

This tool was designed primarily to be used by researchers, engineers and technicians involved in rural development. However, other participants (producers, pastoral cooperatives, NGOs, public services, etc.) can use it, as long as they have IT tools and access to the processed satellite images (including the NDVI calculation).

3C-BIOVIS tool is software that contains three integrated models: (1) calculation of the quantity of biomass available, (2) information on the dry matter content of fodder and (3) calculation of animal stocking.

Field and remote sensing data were programmed and modelled in a computer language to calculate the availability of forage resources. On the one hand, we distinguish the input variables (NDVI, plot size, date, etc.) and the output variables (dry matter content, stocking capacity, etc.). On the other hand, we distinguish the state variables that constitute the engine of the system by mobilizing the different models and internal calculations of the software. The calculations are done on all the pixels of the forage plot and thus provide an estimate of the average productivity.

The validation of the tool was to compare the values measured in the field and the values predicted by the cross-validation method “leave-one-out cross-validation (Holden et al., 1996). The mean squared error was calculated to give the performance score of the model on the test sample. This operation was repeated by selecting each validation sample that was not used for the construction of the model. The average of the quadratic errors was calculated to estimate the prediction error. On average, 70% of the variability of direct measurements in the field was explained by the models integrated into the software.