I think it highly depends if you intend to do statistical analysis or not. In general, your sampling locations should be representative for the entire field. Probably the guidelines of the FAO offer useful help: http://www.fao.org/soils-portal/soil-survey/sampling-and-laboratory-techniques/field-guides/en/.

Flo

Here is an alternative link, as the FAO link seems not to work...

http://www.nrcs.usda.gov/Internet/FSE_DOCUMENTS/nrcs142p2_052523.pdf

Dear Jhoanna Santiago

Actually samples should be representative of a site/field/spot. Increasing the number of sample cores makes the sample more representative. Sampling should be done considering the slope of land, fertility gradient, flow direction of irrigation water etc, However, attached PDF may help you. Please consult it.

Jhoanna , very good question , we most of the times overlook  such issues , which are  so important  in the context of soil fertility evaluation in terraced fields.  Slope of the terraced field is most important .   Usually , soil samples should be  taken at toe slope, foot slope, back slope, shoulder slope and crest positions of the field at 0-15 cm and  0-30 cm soil depth to evaluate the status of soil  fertility ...Many times,  soil samples are also  collected from the lower (deposition), middle and upper (loss) zones of the terraces at  depths of  0-15 cm and 15-30cm.

Dear Jhoanna. S,

Soil samples may be collected from depths of 0-30cm (deposition layer),  40-60 cm and the rest at 70-80 cm basing  on your research and soils classification in the field.

Here is another very good piece of work , the abstract is highlighted below: 

Abstract :  In the Highlands of Ethiopia, soil erosion is a pressing challenge causing deterioration of soil quality including soil fertility. To overcome this problem, the government has been taking various sustainable land management (SLM) measures. This study was conducted in 2011 to investigate the long-term impacts of soil conservation on soil qualities and crop performance at Anjeni watershed in the central highlands of Ethiopia. Soil and crop samples were collected from the lower (deposition), middle and upper (loss) zones of the terraces at a depth of 30cm.The test crops were maize (Zea mays) and bread wheat (Triticum aestivum). Soil samples were also taken at toe slope, foot slope, back slope, shoulder slope and crest positions of the watershed from 0-30 cm soil depth to evaluate the status of soil qualities along the catena. Results of the study showed that soil pH, exchangeable cations, available phosphorus, sum of exchangeable bases and percent base saturation showed non-significant difference between the loss zone and deposition zones, whereas higher mean value of organic carbon, and total nitrogen were obtained at the deposition zone than the loss zone. For both testing crops, higher mean yields were found at deposition zones followed by the middle zones while the lowest value was obtained from the loss zones. Soil pH, exchangeable cations, available phosphorus, sum of exchangeable bases, percent base saturation, organic carbon and total nitrogen showed significant variation due to slope position differences. Toe slope position followed by crest slope position showed higher mean value of the parameters. The shoulder slope position had the lowest mean value for all parameters. From the results of the study, it was possible to conclude that soil conservation measures implemented at Anjeni watershed reduced soils erosion, improved soil qualities and increased crop yield. It is, therefore, possible to recommend the need for scaling up of results obtained from learning watersheds on soil conservation activities to the highlands of Ethiopia to improve the soil quality and livelihoods’ of the society.( Source : Amare et al . 2013, Journal of Agricultural Science; Vol. 5, No. 2; 2013 ).

@Jhoanna you have already got experts opinion here but one little thing I shall like to add to it is that analyzing in proper multiplications  is as important as proper sampling