Formulas for the transformation functions

The formulas used for the transformation functions in the Rescale by Function tool will be explained in this section. The transformation function formulas are presented as Spatial Analyst Map algebra expressions.

The general formula for each transformation function is identified. By default, the general formula is altered to fit to the minimum (lowerThreshold) and the maximum (upperThreshold) of the input dataset. How the default parameters are determined is also presented below.

Using the general formula, any number of function curves can be defined, and any portion of the function curve can be captured. In the default version of the formula, there is only one curve for the function that will fit to the minimum and maximum of the input dataset. As a result, in addition to applying the linear scaling of the transformed values, data represented in different units, such as Celsius or Fahrenheit, will produce the same output evaluation values.

Apply a transformation function

Conceptually, the steps for applying a transformation function are as follows:

  1. Specify the input dataset and function parameters.
  2. Apply the transformation formula.
  3. Scale the transformed values linearly to the evaluation scale.

The linear transformation equation used to scale the transformed values to the evaluation scale in step 3 above is as follows.

ScaledY = fromScale + (Y - min(Y)) * r

where:

  • Y = F(In) is the transformed value with formula F, which is shown in the following general forms:
  • In is the input value
  • r is the result of (toScale - fromScale) / (max(Y) - min(Y))
  • min(Y) is the minimum value of Y
  • max(Y) is the maximum value of Y
  • fromScale is the minimum of the evaluation scale (for example, 1)
  • toScale is the maximum of the evaluation scale (for example, 10)

Exponential

The following are the formulas and default values for the Exponential transformation function.

General form

The general form of the formula is as follows.

Exp((In – inShift) * baseFactor)

where:

  • In is the input value
  • inShift is the value subtracted from the input values
  • baseFactor is a multiplier for the shifted input values, which controls how steeply the exponential function increases

Default values

The default values of inShift and baseFactor are calculated as follows.

inShift = Con(toScale > fromScale, A, B)
baseFactor = Con(toScale > fromScale, C, D)

where:

  • A = (minIn * Ln(toScale) - maxIn * Ln(fromScale)) / (Ln(toScale) - Ln(fromScale))
  • B = (minIn * Ln(fromScale) - maxIn * Ln(toScale)) / (Ln(fromScale) - Ln(toScale))
  • C = (Ln(toScale) - Ln(fromScale)) / (maxIn - minIn)
  • D = (Ln(fromScale) - Ln(toScale)) / (maxIn - minIn)
  • Ln is the logarithm of a value with base e
  • minIn is the minimum of the input
  • maxIn is the maximum of the input

Gaussian

The following are the formulas and default values for the Gaussian transformation function.

General form

The general form of the formula is as follows.

Exp(-Spread * Square(In - MidPoint))

where:

  • Spread is the spread parameter, which controls the steepness of the decay from the midpoint
  • In is the input value
  • MidPoint is the midpoint parameter, which defines the center of the Gaussian curve

Default values

The default values of Spread and MidPoint are calculated as follows.

Spread = Ln(10) * 4 / Pow(MidPoint - minIn, 2)
MidPoint = (maxIn + minIn) / 2

where:

  • Ln(10) is the logarithm of 10 with base e
  • Pow(MidPoint - minIn, 2) is the value of MidPoint - minIn to the power of 2
  • minIn is the minimum of input
  • maxIn is the maximum of input

Large

The following are the formulas and default values for the Large transformation function.

General form

The general form of the formula is as follows.

1.0 / (1.0 + Pow(In / MidPoint, - Spread))

where:

  • In is the input value
  • MidPoint is the midpoint parameter, which defines the transition point of the function
  • Spread is the spread parameter, which controls how quickly the preference increases and decreases

Default values

The default values of Spread and MidPoint are calculated as below.

Spread = 5
MidPoint = (maxIn + minIn) / 2

where:

  • minIn is the minimum of input
  • maxIn is the maximum of input

Linear

The following are the formulas and default values for the Linear transformation function.

General form

The general forms of the formula according to the type of slope follow:

Positive slope (Min < Max):

Con(In < Min, 0, Con(In > Max, 1, (In - Min) / Diff))

Negative slope (Min > Max):

Con(In > Min, 0, Con(In < Max, 1, (In - Min) / Diff))

where:

  • In is the input value
  • Min is the specified minimum, which establishes the first point the Linear function must pass through
  • Max is the specified maximum, which establishes the second point the Linear function must pass through
  • Diff equals Max - Min

Default values

The default values of Min and Max are calculated as below.

Min = minIn
Max = maxIn

where:

  • minIn is the minimum of input
  • maxIn is the maximum of input

Logarithm

The following are the formulas and default values for the Logarithm transformation function.

General form

The general form of the formula is as follows.

Ln((In – inShift) * baseFactor)

where:

  • In is the input value
  • InShift is the value subtracted from the input values
  • baseFactor is the multiplier of the shifted input values, which controls the rise of the logarithm function

Default values

The default values of inShift and baseFactor are calculated as below.

inShift = Con(toScale > fromScale, A, B)
baseFactor = Con(toScale > fromScale, C, D)

where:

  • A = (minIn * Exp(toScale) - maxIn * Exp(fromScale)) / (Exp(toScale) - Exp(fromScale))
  • B = (minIn * Exp(fromScale) - maxIn * Exp(toScale)) / (Exp(fromScale) - Exp(toScale))
  • C = (Exp(toScale) - Exp(fromScale)) / (maxIn - minIn)
  • D = (Exp(fromScale) - Exp(toScale)) / (maxIn - minIn)
  • Exp is the exponential of a value with base e
  • minIn is the minimum of input
  • maxIn is the maximum of input

LogisticDecay

The following are the formulas for the Logistic Decay transformation function.

General form

The general form of the formula is as follows.

C / (1 + A * Exp((In – Min) * B))

The definitions of the parameters in the formula are the same as in LogisticGrowth.

Note:

In the LogisticGrowth equation, the Exponential calculations are determined from the negative of the value for calculating exponential: Exp( - (In - Min) * B).

LogisticGrowth

The following are the formulas and default values for the Logistic Growth transformation function.

General form

The general form of the formula is as follows.

C / (1 + A * Exp( - (In – Min) * B))

where:

  • Min is the specified minimum, which controls the starting point of the logistic growth
  • C is the capacity or upper horizontal asymptote
  • A determines the intercept when In = Min
  • B determines the base of exponential function Exp

Default values

The default values of C, A, and B are determined as follows.

C = 100
A = C / yInterceptPercent - 1
B = Ln(A) / (0.5 * (Max + Min) - Min)

where:

  • yInterceptPercent is the percent of y intercept divided by C, which is set to 1 for LogisticGrowth and 99 for LogisticDecay by default
  • Min is the specified minimum, which is set to the minimum value of the input by default
  • Max is the specified maximum, which is set to the maximum value of the input by default
  • Ln(A) is the logarithm of A with base e
  • 0.5 * (Max + Min) is the x coordinate of the inflection point y = C / 2, making the inflection point in the middle of Min and Max

MSLarge

The following are the formulas and default values for the MSLarge transformation function.

General form

The general form of the formula is as follows.

Con(In > nMeans, 1 - (nStdv / (In - nMeans + nStdv)), 0)

where:

  • In is the input value
  • nMeans equals meanMultiplier * mean
  • nStdv equals STDMultiplier * std, in which std refers to the standard deviation among the entire population

Default values

Both meanMultiplier and STDMultiplier are set to 1 by default.

MSSmall

The following are the formulas and default values for the MSSmall transformation function.

General form

The general form of the formula is as follows.

Con(In > nMeans, nStdv / (In - nMeans + nStdv), 1)

where:

  • In is the input value
  • nMeans equals meanMultiplier * mean
  • nStdv equals STDMultiplier * std, in which std refers to the standard deviation among the entire population

Default values

Both meanMultiplier and STDMultiplier are set to 1 by default.

Near

The following are the formulas and default values for the Near transformation function.

General form

The general form of the formula is as follows.

1.0 / (1.0 + Spread * Pow(In - MidPoint, 2))

where:

  • In is the input value
  • Spread is the spread parameter, which controls the steepness of the decay from the midpoint
  • MidPoint is the midpoint parameter, which defines the center of the curve

Default values

The default values of Spread and MidPoint are calculated as follows.

Spread = 36 / Pow(MidPoint - minIn, 2)
MidPoint = (maxIn + minIn) / 2

where:

  • Pow(MidPoint - minIn, 2) is the value of MidPoint - minIn to the power of 2
  • minIn is the minimum of input
  • maxIn is the maximum of input

Power

The following are the formulas and default values for the Power transformation function.

General form

The general form of the formula is as follows.

Pow(In – inShift, Exponent)

where:

  • In is the input value
  • inShift is the value subtracted from the input values
  • Exponent is the exponent to raise the function to

Default values

The default values of inShift and Exponent are calculated as follows.

inShift = Con(toScale > fromScale, A, B)

where:

  • A = Con(fromScale == 1, minIn - 1, minIn)
  • B = Con(toScale == 1, minIn - 1, minIn)
  • minIn is the minimum of input
  • maxIn is the maximum of input

Exponent = Con(toScale > fromScale, C, D)

where:

  • C = Con(fromScale == 0, C1, Con(fromScale == 1, C2, 2))
  • C1 = Con(toScale <= 1, 1, Ln(toScale) / (maxIn - inShift))
  • C2 = Ln(toScale) / Ln(maxIn - inShift)
  • D = Con(toScale == 0, D1, Con(toScale == 1, D2, 2))
  • D1 = Con(fromScale <= 1, 1, Ln(fromScale) / (maxIn - inShift))
  • D2 = Ln(fromScale) / Ln(maxIn - inShift)
  • Ln is the logarithm of a value with base e
  • minIn is the minimum of input
  • maxIn is the maximum of input

Small

The following are the formulas and default values for the Small transformation function.

General form

The general form of the formula is as follows.

1.0 / (1.0 + Pow(In / MidPoint, Spread))

where:

  • In is the input value
  • MidPoint is the midpoint parameter, which defines the transition point of the function
  • Spread is the spread parameter, which controls how quickly the preference decreases and increases

Default values

The default values of Spread and MidPoint are calculated as follows.

Spread = 5
MidPoint = (maxIn + minIn) / 2

where:

  • minIn is the minimum of input
  • maxIn is the maximum of input

SymmetricLinear

The following are the formulas and default values for the Symmetric Linear transformation function.

General form

The general forms of the formula for different types of slopes are as follows.

For positive slopes (Min < Max):

Con(In < Min , 0, Con(In < MidP, (In - Min) / HDiff, Con(In > Max, 0, (Max - In) / HDiff)))

For negative slopes (Min > Max):

Con(In < Max, 1, Con(In < MidP, (In - MidP) / HDiff, Con(In > Min, 1, (MidP - In) / HDiff)))

where:

  • In is the input value
  • Min is the specified minimum, which establishes one point the SymmetricLinear function must pass through
  • Max is the specified maximum, which establishes the other point the SymmetricLinear function must pass through
  • Diff equals Max - Min
  • HDiff equals 0.5 * Diff
  • MidP equals Min + HDiff

Default values

The default values of Min and Max are calculated as follows.

Min = minIn
Max = maxIn

where:

  • minIn is the minimum of input
  • maxIn is the maximum of input

Related topics