Food Assessment by Satellite Technology (FAST)

In the following a brief introduction is given on the FAST crop yield forecasting technology. 

Results are presented in the projects section on food security data.

EARS Crop Growth Model (ECGM)

This model simulates crop biomass development during the growing season. The ECGM is fed with the satellite derived global radiation and relative evapotranspiration data. It simulates crop biomass development on a daily basis. The daily increase of biomass B is calculated with:

B_t = B_t-1 + P - R

where P is the dry matter production and R the maintenance respiration. The calculation of the dry matter production starts from Monteith (1977), who demonstrated that crop biomass production in linearly related with light interception, independent of crop type. This implies that the daily dry matter production may be expressed in terms of the daily average global radiation (I_g) with

DB = a.C.I_g

where 'a' is a conversion constant, C the fraction of soil covered by the crop, and I_g the daily global radiation. Monteith's results apply to a relative limited geographic area. Elements of light use efficiency and water limitation therefore have to be included.

Biomass growth calculation scheme

Water limitation

CO2 uptake and consequently dry matter production are reduced by stomatal closure as a result of limited water availability to the plant. Since water vapour and CO₂ share the same diffusion path, there is strong relation between evapotranspiration and CO₂ assimilation. This relation has been developed by Stewart (1977) and was verified and documented on the basis of experimental data from all over the world by Doorenbos and Kassam (1979):

(1-RY) = k_y.(1-RE)

Here RY is the relative yield (RY=Y/Y_P) and RE the relative evapotranspiration(RE=LE/LE_P). The yield response factor (k_y) is crop specific and accounts for its drought sensitivity.

Light use efficiency

Only 50% of the absorbed solar radiation is photosynthetic active radiation (PAR). PAR excites electrons which provide the electric energy for photosynthesis. However not all excitations lead to photosynthesis. A considerable and variable part is dissipated into heat. At very low light the PAR used for photosynthesis can be 80%, but at higher light the efficiency decreases considerably. The light use efficiency as a function of light level has been investigated by Rosema et al. (1998). It may be described with:

f = 1 / (1.25 + 2.r.I_g)

where r is an electron transport resistance, which is plant specific and represents its ability to assimilate CO₂.

Dry matter production

Based on the previous discussion we may generalize the dry matter production P as:

P = a.C.RY.f.I_g

From the daily dry matter production we have to subtract the carbohydrates that are used for maintenance respiration.

Respiration

This term is proportional to the total biomass. The maintenance respiration increases with the temperature and is usually expressed as a function of the air temperature (Supit et al.1994).

Crop yield forecasting

Based on the previous equations the dry matter production is simulated during the growing season as a function of relative evapotranspiration and global radiation. The ECGM is run pixel-by-pixel through the growing season. An automatic or pre-defined start of season may be chosen. The model is run one time with the EWBMS actual evapotranspiration data fields and one time with the potential evapotranspiration data fields as input, resulting in an actual biomass estimate (B) and a potential biomass estimate (B_p). The ratio of the two is the relative biomass, which is taken to be equal to the relative economical yield (RY)

RY = B/B_p = Y/Y_p

Meteosat and MSG data are received since 1993. 5 years previous to the current year are processed to determine the mean relative yield (RY_m). Then the Difference Yield (DY) is defined and calculated as follows

DY = (RY-RY_m)/RY_m = (Y-Y_m)/Y_m

DY represents the deviation from the average economic yield in the previous 5 years. Reversely the current economic yield (Y) may be estimated from the MSG derived difference yield (DY) and the average economic yield during the past 5 years (Ym). The latter may be obtained from e.g. the FAOSTAT database. The relative yield (RY) stabilizes after 60-80 growing days. Therefore, it is possible to make a crop yield forecast from about halfway the growing season on the basis of the MSG derived difference yield. This is illustrated in the figures below.

Sorghum/millet Difference Yield forecast for the Horn of Africa, starting from the first dekad of July until the last dekad of September 2007. The forecast early July and early August do predict the end of season result (Sept D3) quite well.