In the range of water contents studied (0.1–0.4 g H2O g dw−1), Coffea arabica seeds were less sensitive to desiccation than C. costatifructa, C. racemosa and C. sessiliflora seeds. At 0.20 g H2O g dw−1, 53% of C. arabica seeds germinated after direct immersion in LN (rapid cooling, 200°C min−1), but none of them developed into normal seedlings. By contrast, in C. costatifructa, C. racemosa and C. sessiliflora, when seeds were dehydrated to the optimal water content (0.19, 0.28 and 0.31 g H2O g dw−1, respectively), the percentages of seeds which developed into normal seedlings after LN exposure were 26, 78 and 31% of the desiccation control, respectively. Normal seedlings could be recovered from cryopreserved C. arabica seeds only if they were desiccated to 0.20 g H2O g dw−1 and precooled slowly to −50°C prior to immersion in LN. Precooling seeds at 2°C min−1 allowed 25% of seeds to develop into normal seedlings. The thawing rate had no effect on the survival of cryopreserved C. arabica seeds. In all cryopreservation experiments, the total germination did not reflect the percentage of seeds which developed into normal seedlings. Examination of excised embryos indicated a partial explanation of this difference since only the shoot apex was destroyed in abnormal embryos, whereas the hypocotyl and radicle were normal.

Seed desiccation sensitivity was studied in nine species of the genus Coffeaby measuring seed viability after equilibration over various saturated salt solutions. A quantal response model based on the logistic distribution was developed in order to describe the typical S-shaped patterns observed. The closeness of fit of the desiccation sensitivity model was shown, and the assumption that seed desiccation sensitivity follows a continuous distribution within species was verified. For each species, the water content at which 50% of initial viability was reached, WC50, and a specific parameter describing the intra-specific variability, β, were calculated using a non-linear regression. A simplified water sorption model was developed which allowed easy calculation of water activity and water potential corresponding to WC50 (aw50 and Ψ50) for relative humidities ranging between 10 and 100%. Distribution of WC50and Ψ50 (or aw50) in the genus Coffea was homogeneous within the following intervals: from 0.05 to 0.38 g H2O.g−1dw for WC50 and from −168 to −11 MPa for Ψ50. Different classifications of the coffee species studied as regards to their desiccation sensitivity were obtained depending on whether WC50 or Ψ50was used for classification. The continuum for desiccation sensitivity observed within the nine species studied confirmed that coffee is an appropriate material for studying desiccation sensitivity.

Large differences in seed desiccation sensitivity have been observed previously among ten coffee species (Coffea arabica, C. brevipes, C. canephora, C. eugenioides, C. humilis, C. liberica, C. pocsii, C. pseudo-zanguebariae, C. sessiliflora and C.stenophylla). Of these species, C. liberica and C. humilis were the most sensitive to desiccation and C. pseudozanguebariae the most tolerant. A study was carried out using the same seed lots to investigate if these differences in desiccation tolerance could be correlated with differences in soluble sugar content. Soluble sugars were extracted from dry seeds and analysed using high performance liquid chromatography. The seed monosaccharide (glucose and fructose) content was very low (1.5 to 2 mg g-1dry weight [dw]) in all species studied. The sucrose content ranged from 33 mg g-1dw in C. liberica seeds to 89 mg g-1dw in seeds of C. pocsii. Raffinose was detected in the seeds of only five species (C.arabica, C.brevipes, C.humilis, C.sessiliflora, C.stenophylla), among which only three species (C.arabica, C.sessiliflora and C.brevipes) also contained stachyose. Both raffinose and stachyose were present in very low quantities (0.3–1.4 mg g-1dw and 0.1–0.7 mg g-1dw, respectively). Verbascose was never detected. No significant relationship was found between seed desiccation sensitivity and: (i) the sugar content; (ii) the presence/absence of oligosaccharides; and (iii) the oligosaccharide:sucrose ratio.

The relationships among water content, relative humidity and temperature were documented in both seeds and excised embryos of Coffea spp. using water sorption isotherms. Isotherms were constructed at 5, 15 and 25°C and calculated for lower temperatures. There were no apparent differences in sorption characteristics among whole seeds of several cultivars of C. arabica and among different species of Coffea. Excised embryos of genetically diverse Coffea germplasm also exhibited similar sorption characteristics, though there were substantial differences observed between embryos and whole seeds. The shape of isotherms of coffee seed tissues was intermediate to the reverse sigmoidal shape observed for orthodox seeds and the monotonic shape observed for desiccation intolerant plant tissues. The heats of sorption calculated for RH ≤ 25% for whole seeds of Coffea spp. were similar to orthodox seeds. In contrast, the heats of sorption calculated in the same RH range for excised embryos were intermediate between those of orthodox and recalcitrant embryos. Our observations are consistent with earlier observations that desiccation sensitivity or poor longevity is linked with low levels of water sorption at relative humidities less than 25%. An explanation for this remains elusive.

A broad variability for seed desiccation sensitivity, as quantified by the water content and the water activity at which half of the initial viability is lost, has been previously observed within nine African coffee species. In order to investigate if these different degrees of desiccation sensitivity correspond to an adaptive trait, additional data, such as the duration of seed development and seed water content at maturity, were measured for these species, and the relationships between these parameters and some climatic characteristics of their specific native environments were investigated. Since flowering in all coffee species occurs only a few days after the main rainfall marking the end of the dry season, simulations could be made, based on the continuous sequences of rainfall data compiled in databases of nine climatic stations, chosen for their appropriate location in the collecting areas. The simulations revealed a highly significant correlation between the duration of seed development and that of the wet season. Consequently, mature seeds are shed at the beginning of the following dry season. Moreover, the mean number of dry months that seeds have to withstand after shedding was significantly correlated with the parameters used to quantify seed desiccation sensitivity. By contrast, seed moisture content at maturity was not correlated with the level of seed desiccation tolerance. All these results are discussed on the basis of more detailed descriptions of the natural habitats of the coffee species studied.