One of the major weaknesses of existing national mechanisms according to Hiernaux et al. (2016) is the lack of accounting for seasonality.

In West Africa, permanent pastures and rangelands represent two-thirds of agricultural land, offering a mix of grass, legumes, woody species, planted woody fodder species and crop residues (Boudet, 1972; Hiernaux et al., 2009). The available forage mass in the Sahel is the main driver of the voluntary ingestion per livestock species (Dicko et al., 2006). The DM declines during the dry season, as well as fodder digestibility, and the proportion of unpalatable species recorded at the peak vegetative stage ranges from 32 to 61% (Hiernaux et al., 1997; Ayantunde et al., 1999). The late dry season is therefore the most critical period for ruminant nutrition in the Sahel (Le Houérou and Hoste, 1977).

Adapted to the regular seasonality of precipitation, solar radiation, temperature and air humidity, the herbaceous vegetation in the Sahel is largely dominated by annual plants with a short cycle, associated with more or less scattered woody plants, among which deciduous hardwoods dominate (Hiernaux and Le Houerou, 2006). There are relatively few reliable year-round monitoring studies describing the extensive use of vegetation (grasses, trees and shrubs) by livestock in arid and semi-arid climates in sub-Saharan Africa (Ayantunde et al., 1999; Fernández-Rivera et al., 2005; Achard and Chanono, 2006; Schlecht et al., 2006; Hiernaux et al., 2009). Studies monitoring the weight and reproductive performance of cattle such as Wilson (1986), Wilson (1989), Colin de Verdière (1994), Lesnoff (1999), Ezanno et al. (2003); Chirat et al. (2014) and Assouma et al. (2018) all converge to highlight the strong seasonality of animal production and cattle in particular, which reflects the availability of resources and adapted animal feeding practices.

In addition, the use of the standard average value of 25 g DM intake per kg LWdoes not take into account the seasonal change in behavior of animals with regard to selective grazing to adapt to changes in availability and quality of the fodder, which goes from a highly herbaceous green fodder digestible during the wet season to poorly digestible straw and litter during the dry season (Chirat et al., 2014). Assouma et al. (2018) showed that using this method leads to overestimating animal consumption by 20 to 300% depending on location and season.

Another limitation is the non-linearity of metabolizable energy requirements with the weight of the animal (Zemmelink, 1980). In particular, Assouma et al. (2018) showed that this resulted in underestimating the consumption of small ruminants and overestimating the consumption of cattle.

Finally, this approach does not consider that an animal can mobilize its body reserves, can lose weight and eat less than is necessary to cover all metabolic needs. Herders know how to take advantage of the high quality of the fodder of rangelands selectively grazed during the wet season (Ayantunde et al., 1999) to develop and build up fat reserves that the animal will gradually burn over the long dry season (Ezanno et al., 2003). In addition, dietary restrictions influence the metabolism of ruminants and the low quality of feed affects their digestive performance (Goopy et al., 2020). This can also lead to an overestimation of daily forage consumption and, in the case of severe below-maintenance intake, to higher enteric methane emissions per unit of fodder ingested (Goopy et al., 2020).

Not accounting for seasonality can lead to errors in calculations related to the stocks of feed and fodder needed for distribution in times of drought or the impact of climate change on pastoral and agro-pastoral systems, as illustrated in Table 1 , from FAO (2021). Animal ingestion and enteric methane emissions were estimated according to average energy requirements on different bases: annual (no seasonality, most common approach), two seasons (wet season and dry season) and three seasons (wet season, and cold and hot dry seasons). They were compared to observations of ingestion made in the field (hand picking method and total collection of faeces). The results in Table 1 show that annual feed rations calculated according to average energy requirements overestimate ingested quantities by 26 to 71% for this system. Modelling over three seasons reduces the overestimation (13%in the wet season), but it remains significant, especially in the hot dry season (55%). Enteric methane emissions calculated with these feed intakes using the Global Livestock Environmental Assessment Model – GLEAM (Gerber et al., 2013) based on Intergovernmental Panel on the climate change (IPCC) were also overestimated by 32 to 76% compared to measurements from the field based on total organic matter digestibility. Such results reveal that current estimates of enteric methane emission intensities for sub-Saharan livestock are inaccurately higher than actual observations, which may partly be due to the persistent neglect of seasonality, a key parameter of feeding systems is SSA.

In response to the highly seasonal availability of biomass, pastoral farming is based on livestock mobility (Turner et al., 2014). In the Sahel, biomass growth is concentrated over only a few months and fodder storage capacity is limited. Mobility is a strategy for adapting to the seasonality of fodder availability (Brottem et al., 2014), but it also constitutes a constraint for productivity and can be a bottleneck in the system when it is reduced for political, economic, conflict or health reasons (Moritz, 2006; Turner et al., 2011; Marega et al., 2018). The expansion of cultivated land also contributes to limiting the mobility of herds.

One of the highest sources of uncertainty in a feed balance is the number of animals. In the Sahel, animal numbers for the most common species -cattle, sheep, goats, horses, and donkey- are difficult to quantify each year for various reasons, mobility being an important one. Systematic censuses of animals are rare and often outdated. Animal numbers reported in national and international statistics are estimated on the basis of an annual growth rate by species and applied on a national scale which conceals external events that strongly influence the dynamics and production of livestock (droughts, epizooties, etc.). This growth rate is generally not reviewed or reassessed regularly.

Feed balance are generally established for administrative units such as countries, regions or provinces, or even districts. Although this has advantages in terms of data collection and planning, this approach masks the seasonal movements of animals, which are often cross-border (Hiernaux et al., 2016). This can lead to overestimates or underestimates in the feed balance. To solve this problem, De Haan et al. (2016) defined specific cross-border geographical units, autonomous in terms of animal and food mobility, called “grazing sheds”. Although this approach has proven to be relevant for large-scale forage balances, it will be difficult to apply to sub-national balances, which would require detailed information on animal movements.

In a study conducted in the Sahelian zone of Burkina Faso, Sanon (2007) showed that cattle, sheep and goats spend respectively 38, 59 and 57% of their time feeding during the dry season and 72, 73 and 65% during the rainy season. During the dry season, the search for straw and leaves from trees and shrubs, fallen fruit or pods can reach 4.5% of the time for cattle, 28% for sheep and 51% for goats. Foliage supplementation (distribution of collected foliage) is also known to improve productivity and animal growth rates (Sanon et al., 2008). Browsing of woody biomass has a distinct advantage over tropical grasses in terms of its superior nutritional value (both in energy and protein content) during the dry season and constitutes an important element of the feed balance (Ouédraogo-Koné et al., 2008).

The main contributions of woody species to the subsistence of human and animal populations in the Sahelian zone have been described by Sanon et al. (2008) for West Africa, by, Ravhuhali et al. (2022) for Southern Africa, and by Gowda et al. (2019) for East Africa,. The main characteristics of browsed woody plants are their high CP and mineral content (Franzel et al., 2014). However, the high lignin content and the presence of tannins, which are antinutritional factors, potentially toxic for ruminants, may limit its consumption (Van, 2006).

However, knowledge on the seasonality and nutritional value of these species is still limited and not systematically investigated (Ouédraogo-Koné et al., 2008; Zampaligré et al., 2013). A recent study conducted in Senegal by Assouma (2016) based on field observations and excreta collection, as well as a reference document drawn up by the FAO with the support of CIRAD (FAO, unpublished) help to fill this gap.

Assouma (2016) showed that in northern Senegal tree foliage and fruit accounted for an average of 13% of livestock feed on a DM basis, but this share varied from nearly 0% in the rainy season to 40% in the hot dry season. The same author measured that the ingestion of different tree species ranged from 50 g DM per kg of LW^0.75/day to more than 80 g for leaves and fruits/pods/flowers/seeds during the cold dry and the hot dry seasons. The digestibility of the leaves ranged from 40% to over 80%, depending on the species. These values ​​are similar to the ones obtained for fruits/pods/flowers/seeds by (Guérin, 1994).

Feed balances generally do not distinguish between the different categories of feed resources available in pastures (herbaceous and leguminous plants, tree leaves, fruits, pods, flowers and seeds) and do not take into account substitutions animals make between these different resources from one season to another. The consequence of these limitations can be a significant overestimation of intake, as shown by Assouma et al. (2018).

The same authors estimated that the tree leaf balance, calculated as the ratio between what is consumed and what is available as foliage mass, varied from 0% to 10.3% depending on the month of the year. This means that even when tree leaf browsing is at its maximum, the actual foliage intake by the animals remains modest.

The biomass produced during the rainy season estimated from the NDVI index can be separated in crops and rangelands, as well as herbaceous and woody plants. The actual amount of usable and accessible forage depends on assumptions made about other uses of biomass, losses, distance to water and a number of other factors ( Figure 1 ). The lack of information to estimate the coefficients for the calibration of the available biomass into actual use constitutes a major obstacle to the establishment of accurate feed balances. De Haan et al. (2016) showed that applying the classic maximum livestock utilization rate of 30% of natural biomass from pasture in sub-Saharan drylands resulted in an average annual balance of 75% over the period 1998-2011.

This means that a systematic 25% surplus in natural vegetation should have existed during this period, when deficits did in fact occur. On the other hand, estimates made at the local scale by Diawara et al. (2018) give lower rates. These results show the need to refine these calibration factors on the basis of local observations. FAO (2018) also suggests using local factors.

Similarly, there is also a need to refine the coefficients applied to crop residues, which are often long-established default coefficients, and which have never been reviewed. These coefficients can vary at subnational level, for example farmers may return different share of crop residues to the crop field in different regions. Some crop residues may be used for construction while others may be used for fuel or animal bedding.

Availability of natural biomass in rangelands can be estimated from NDVI interpretation. NDVI is a relatively simple and easily available indicator for estimating biomass during the rainy season, but it is not relevant in the dry season, because the signal from dry vegetation and litter, also consumed by animals, is null. A recent review of existing indices for estimating DM during dry seasons from remote sensing information provides useful recommendations (Lo et al., 2022). This will make it possible to better monitor in real time the evolution of fodder availability during the dry season (straw, litter, stubble in the fields, burned areas).

Input information for natural biomass in the FBS correspond to total areas in ha and DM yield in kg/ha in various seasons (or degradation rates between seasons) for grass and for tree fodder. The availability is expressed in DM in kg.

Estimating the seasonal availability of crop residues and by-products can be done directly from agricultural production, which is generally quantified and monitored at sub-national scale. Input information correspond to cultivated areas for each crop, yields as well as percentage of DM and of each by-product compared to grains (e.g. bran) [see for example FAO (2018)].

Existing factors are used to calibrate available biomass as a usable and accessible resource and to estimate the share of crop residues and by-products used as animal feed, and performance parameters are used to estimate animal requirements (Hiernaux et al., 2016; FAO, 2018). National average factors that have not been revised for many years should be avoided. Instead, regular consultations with pastoral organizations and extension services, as well as case studies and consultations with stakeholders in the processing industry can help revise these parameters or establish new ones.

Parameters should also take into account the seasonality of the use of crop by-products and residues that may be stored after harvest and distributed later.

The nutritional value of feed resources, including grass, tree fodder, crop residues and by-products, is characterized in the FBS by their CP content and their energy content (ME expressed in Mcal/kgDM). However, the content of Digestible Proteins in the Intestine (PDI in French) proposed by the National Institute of Agronomic Research in France (INRA, 2018) is a better estimate of the amount of protein actually used by the animal and protein content is therefore expressed in kg of PDI/kg DM in the tool. Once the inventory of all types of animal feed has been established, this information can be obtained from existing international or regional reference tables such as Feedipedia (https://www.feedipedia.org/), or the Sub-Saharan Africa Feed Composition Database of ILRI (https://feedsdatabase.ilri.org/).

While a standard characterization in PDI and ME is suitable for animal feeds such as those distributed in supplementation and in stalls, the same is not true for rangeland grass and legumes, the quality of which is very variable over time and space, and further complicated by specific palatability. Specific references to the Sahel are not easily identifiable in international databases. It is therefore recommended to use the CP (or PDI) and ME content produced by national or regional institutions and agricultural research centers.

It is recommended to carry out regular livestock surveys in order to revise/validate the official animal numbers. A first complete and systematic census should also be carried out in order to have a reliable reference situation that can then be updated according to known and already applied methodologies, with a growth rate validated each year and taking into account the possible impacts of droughts, animal diseases etc. Livestock census data should be disaggregated between the main production systems (e.g. pastoral, agro-pastoral and peri-urban). Working with herders to collect data on the number and composition of the herd on a local and seasonal basis is also key to better estimating numbers.

Mobility is well described in various regional reports (see for example Touré et al., 2012). However, it is difficult to include it in feed balances because it depends on the scale at which it is established and it is variable from year to year. Information regarding animal mobility is mostly dispersed and available at different scales. This information must be updated to identify routes of mobility and their regular changes, including where animals are located at different times of the year.

Given the considerable impact that mobility can have on the feed balance, it is recommended to strengthen the collection of data relating to transhumance at the points of entry into the country and, more generally, of data concerning the mobility of livestock at national level. This must be done in collaboration with pastoralist organizations and can be facilitated by the use of GPS or drone technologies and spatial modeling of areas of seasonal animal concentration.

In tropical regions, animal feed requirements are influenced by a number of factors that may differ compared to temperate climates (INRA, 2018). First, animals produce more body heat due to their physical activities of grazing and ingesting coarse feed resources. They also provide intense chewing work to enhance digestibility of poor-quality fodder. Animal needs must be estimated in ME, a common unit in animal nutrition, but also in protein (PDI) in addition to the intake capacity of DM. To do this, the numbers by animal category (age, sex and reproductive status of adult females) must be collected or estimated from surveys or modelling. The equations for estimating the needs of each category can be obtained by local studies or by meta-references such as those proposed by INRA (2018). They depend on parameters that should therefore be collected and input in the tool for each existing production system. First of all, the energy and protein requirements depend on the metabolic weight of the animals (live weight raised to the power 0.75 or LW^0.75). Requirements for growth or weight gain should be included only for the season in which animals are gaining weight and can be calculated from a daily weight gain. Pregnancy requirements depend on the stage and birth weight of the animal. They are only significant at the end of gestation (the last 3 months for cattle, for example). Lactation requirements depend on daily milk productivity. Finally, requirements for the activity/movements depend on the distances covered.

Parameters necessary to estimate the requirements consider the structure of the herd (for example, the weight of the different types of animals such as cows, heifers, young cattle, calves, ewes, lambs etc.). The other zootechnical parameters to be collected are the seasonal weight variations, the protein and fat content of the milk, the average daily milk production, the average annual weight gain for growing animals and the average daily distances walked for each season. These parameters must be collected on a regular basis (e.g. farm performance monitoring) and by type of production system.

The next sub-sections provide equations that are included in the FBS tool and that can be used in the absence of a specific reference in the country considered (INRA, 2018), based on several meta-analyses in tropical regions, reference pages and equation numbers are indicated in brackets).