The project team surveyed farmers and assessed their gardens to evaluate if soil fertility and associated production of the most important staple, sweet potato, was indeed declining as a consequence of increasing land pressure, shortening fallow periods and a number of other factors including pests and diseases. This survey included 95 farmers in four Highland provinces with an average of three villages in each province. Farmer's perceptions on soil fertility decline and management options were obtained. They were then asked to show the team one garden which had recently been brought back into production after fallow, the 'new' fertile garden, and another garden which was about to go into fallow, the 'old' run-down garden. Plant and soil samples were collected from these two gardens to quantify changes as a result of cropping over time.

Sweet potato yields
Sweet potatoes are harvested sequentially, making it notoriously difficult to obtain reliable yield data. The team used single harvest yields to assess relative production from this 'one-off' farmer survey. In both gardens, old and new, single harvest yields of up to 30 t/ha were observed with a median yield of 6 t/ha. Due to the sequential harvesting these yields were considerably lower than the average total yields of 13-15t/ha for lowland and highland systems reported by Bourke and Vlassak (2004). The large variation in the measured yields and cultivars prevented the use of covariates such as number of previous harvests and age of sweet potato plant to estimate total yields. However the team's observations were similar to others (2006) from an EU funded project on early maturing sweet potato varieties in the lowlands executed by NARI.

Likely causes for the low yields were nutrient deficiencies, pests and diseases and presence of viruses in the sweet potato planting material (traditional and improved). Nutrient deficiencies were evident even though the predominantly volcanic ash soils in the Highlands are regarded as very fertile soils. In a 1997 study more than 50% of sweet potato tissue samples collected were deficient in N, P, K, S and B. Unclean planting material is also a major contributing factor to low yield, but it is unclear if pathogen-tested varieties would suffer nutrient deficiencies comparable to traditional varieties.

Yield trends in old and new gardens
Even though yields were low, there were subtle differences between the new and the old garden. The median yield in the old garden was 4 t/ha and 8 t/ha in the new garden. Given the very large variation in yields it is more realistic to assess 'expected' yield difference between the gardens. Within a 95% probability the expected yield difference ranged from 2 to 8 t/ha. This large difference clearly showed substantial changes in production potential as gardens are continuously being used for sweet potato production. If the factors driving yield in new gardens are known and can be managed, substantial yield reductions as gardens age can be avoided, even within traditional sweet potato varieties.

Pests and diseases: The cause for low yields in old gardens?
There was no difference in the degree of insect infestation of tubers between the old and new gardens (p=13%). There was also no significant difference in the expected percentage of cracked tubers between the old and new gardens (p-value > 7%). But the average yield of vines with cracked tubers was higher in the new gardens, indicating that new gardens produce higher yields even though tubers may have cracks.

Despite higher yields, the new gardens tended to have larger counts of nematodes (p=5%), suggesting that the nematodes could be either non-parasitic or infestation is not a critical factor driving yield. The distribution of weevils was significantly different between the old and new gardens. The chances of the presence of weevils or larvae are significantly higher in the old garden (p-value < 1%). Weevil infestation was also associated and increased with the number of times old gardens were harvested. However, the differences between new and old gardens still persisted even after adjusting the yields for weevil infestation.

Location and soil properties
Separating altitudes into low (<2200m) and high (>2200m) showed that high altitudes have considerably higher total N and soil organic carbon (SOC) than the lower. Despite these subtle differences in apparent soil fertility they had no impact on sweet potato yield. This is of particular interest because SOC is often regarded as an important indicator for soil health and fertility.

Farmers perception
The farmer's response to the survey was a strong willingness to cooperate. Farmers related yield reduction to soil fertility decline and were eager to collaborate as partners and requested feedback from our work. Their observation was that yields in older gardens are lower than in newer gardens and they blamed poor soil conditions for this decline. We asked farmers whether they could grow corn or common bean as a control question to assess their ability to judge soil fertility. The responses to both questions were closely related and farmers thought that the new gardens were more suitable for growing corn or common bean than the old gardens (corn p=4%, bean p=8%). However, the farmer's soil fertility assessment was not related to the sweet potato yields we observed, but it supported the argument that soil fertility differences exist between old and new garden.

Soil fertility management and planting systems
Yields in the new gardens were not affected by planting systems, but yields in the old gardens were highest if farmers used large mounds. In both, old and new gardens, mounds more than 70 cm high were associated with fertility management such as composting and short fallow (in old gardens). The high mounds system was mostly associated with high altitude. However, the difference in planting systems does not explain all the effect of altitude (p < 1%). Rainfall had an additional effect on the expected yield. The rainfall, planting system in old and new gardens and altitude effects together explain 25% of the observed variability in yield.

Conclusion
The results from this scoping study support the call for urgent in-depth field based research on soil fertility and nutrient cycling on the Andisols of PNG with particular reference to soil water. Current methodology to assess the production potential of these soils also requires revision.