Calculate average slope in a quadratic Let's say I have crop yield on y-axis and a measure of heat-stress on x-axis

According to this graph, as heat increases, the yield also increases and after reaching a maximum (at heat = 0.67), the yield starts decreasing. I want to quantify the decrease in yield after crossing the maximum. A sentence like
"After crossing the optimum heat-stress (x=0.67), the yield decreases, on average, by X%"
I want to calculate the X%. At the moment, what I have done is:
1) calculate the mean yield of the maximum till the minimum which will be mean of yield from Heat-stress=0.67 till
   Heat-stress=4.6 (the highest x-value). Let's call this value mean.y'
2) express yield from Heat-stress = 0.67 till Heat-stress = 4.6 as a percentage of mean.y
    Yield.p = (Yield - Mean.y/Mean.y) * 100

3) Take the mean of Yield.p which will give  me a percentage value which is in my case is 13%
So I can write it as:
"After crossing the optimum heat-stress (x=0.67), the yield decreases, on average, by 13%"
Is this correct or could I express in some other way?
 A: I am not sure why you want to express an "average decrease" from the optimal yield as this leaves out important information about the relationship between yield and heat stress. I will provide my thoughts on three aspects of this question: the data, the relationship and the rate of change.
Understanding the Data
To start, what are your data points? I am assuming that your individual data points represent growing seasons for a given crop with the heat stress experienced during the season and the final resulting yield. If this is correct questions that I have include:

*

*How are the growing seasons distributed with regards to heat stress?

*How much variance is there in the yield of growing seasons with a specific heat stress? and

*How well does this quadratic relationship represent the data?

If this is not correct, then you need to be clearer about the data you are presenting.
Understanding the Relationship
At the moment you are showing that there is a clear quadratic relationship between heat stress (shown as the independent variable in your graph) and yield (the dependent variable). Summarizing all of the data points where heat stress is greater than 0.67 into a single number loses that structure. Saying that "After crossing the optimum heat-stress (x=0.67), the yield decreases, on average, by 13%" is not as meaningful as saying "Yield, $Y$, is related to heat stress, $H$, by the quadratic relationship $Y = 2500 - 300(H-0.67)^2$."
NB: I have just quickly eyeballed this formula from your graph.
You can also identify points of interest from your data, like you have by identifying the heat stress value for which yield is maximized. Other points of interest might include the minimum and maximum heat stress values observe, the range of heat stress values for which the yield is 90% of the maximum, etc.
Understanding Rate of Change
Additionally, it is not really meaningful to talk about an "average slope" for a quadratic formula. Moreover, you again have more structure in the data you are presenting. In identifying the relationship between heat stress and yield, you have also identified the relationship between heat stress and rate of change of the yield: this is simply the derivative of the formula given above.
Instead of a single value for "average decrease" over a given interval, it is more illuminating to state that the rate of decrease in yield depends on the heat stress (the derivative of a quadratic is a linear function). The further the heat stress is from 0.67, the faster the yield decreases. Conversely, the closer the heat stress is to 0.67, the slower the yield decreases. This is especially noteworthy if the heat stress value is located in a narrow range around 0.67 for most growing seasons.
