graphic and does not result in a mathematical equation.
If the data are reasonably good, the method can be
precise, and with practice different persons will achieve
similar results. The procedure is amenable both to
agronomic and economic analyses. Data from a rice trial
in La Virginia, Riseralda, Colombia, provide an example.


Table VI-5.


Partial results from an N-P-K experiment
with IR-22 rice in La Virginia, Risaralda,
Colombia, conducted by The National Rice
Program, Colombian Agricultural Institute
(ICA).


 Treatments Replications
N P K I II Mean



 kg/ha

 75 20 20,60 5435 5170 5302
 75 60 20,60 6275 4535 5405
 225 20 20,60 5715 6330 6022
 225 60 20,60 6805 6130 6468
 150 40 0,40,80 6540 5907 6223
 0 40 40 4440 4500 4470
 300 40 40 6950 6800 6875
 150 0 40 5400 6450 5925
 150 80 40 5670 5280 5475

 Source: Hildebrand (1972)

 The experimental design of the original trial was a
partial or incomplete factorial arrangement in a modified
central composite design (see Chapter V) with five levels
each of N, P, and K. Response to N and P were stronger
than response to K, so for purposes of this example
differences in K levels will be ignored. Partial data are
shown in Table VI-5.
 The first step is to construct a graph with the axes
representing quantities of P205 and quantities of N (Fig.
VI-9). The mean yield levels for each known combination of
N and P are then plotted on this graph. These values
represent different levels on the response surface. Then,
with these values as guides, topographic contour lines are