This tutorial is a quick introduction to Harvest Control Rules (HCRs) and their use in fisheries managment. It uses the Introduction to HCRs app from the AMPED suite of apps for exploring harvest strategies.
An HCR is a pre-agreed decision rule that is used to set fishing opportunities in the future. An HCR should be designed so that the management objectives of the fishery have the greatest chance of being achieved. A good HCR is robust to different sorts of uncertainty (which we will discuss later on).
-In this tutorial we will use a simple HCR to set the catch limit of a fishery in every year. The HCR takes an estimate of the current value of SB/SBF=0 and uses it to set a catch limit in the next year. The catch limit is set according to the general rule shown in the figure below:
+In this tutorial we will use a simple HCR to set the catch limit of a fishery in every year. The HCR takes an estimate of the current value of SB/SBF=0 (the amount of adult biomass compared to the amount of adult biomass if there was no fishing) and uses it to set a catch limit in the next year. The catch limit is set according to the general rule shown in the figure below:
The current value of SB/SBF=0 is on x-axis along the bottom. The catch limit in the next year is on the y-axis on the side. The red line is the rule that sets the catch limit given the SB/SBF=0. The basic idea is that if SB/SBF=0 starts to fall, the catches are reduced. If the SB/SBF=0 starts to rise, the cathes are increased.
+The current value of SB/SBF=0 is on x-axis along the bottom. The catch limit in the next year is on the y-axis on the side. The red line is the rule that sets the catch limit given the SB/SBF=0. The basic idea is that if SB/SBF=0 starts to fall, the catches are reduced. If the SB/SBF=0 starts to rise, the catches are increased.
The shape of the HCR is determined by 4 parameters: Blim, Belbow, Cmin and Cmax. When the estimated SB/SBF=0 is greater than Belbow the catch limit is set at Cmax. When SB/SBF=0 is less than Blim the catch limit is set at Cmin. When SB/SBF=0 is betwen Blim and Belbow, the catch limit is set according to the slope.
Note that the fishery used in this tutorial is not based on a particular fishery or stock. It’s just a made up example.
@@ -338,7 +338,7 @@The purpose of the HCR is to set the catch limit each year. The HCR uses the current estimated value of SB/SBF=0 to set the catch limit. This rule will be applied every year in the future to set new catch limits each year.
+The purpose of the HCR is to set the catch limit each year. The HCR uses the current estimated value of SB/SBF=0 to set the catch limit. This rule will be applied every year in the future to set new catch limits in each year.
We start at the very beginning of 2019 and we want to use the HCR to set the catch limit for 2019. The SB/SBF=0 can be seen in the bottom-left plot.
The way the HCR operates can be seen by following the blue arrow from the SB/SBF=0 plot, at the bottom left, to the HCR plot, at the top right. The current estimated value of SB/SBF=0 is shown on the HCR plot as the blue dashed vertical line.
The catch limit in the following year is set by reading the corresponding catch limit from the HCR. This is shown by the blue dashed horizontal line on the HCR plot. The new catch limit is also shown on the catch plot at the top-left, as the blue dashed line. This represents what the catches will be in 2019.
@@ -355,15 +355,15 @@Press the Reset button in the left panel to clear the stock projection. Run through the projection again by pressing the Advance button. Make sure that you understand how the HCR uses the last value of SB/SbF=0 to set the catch limit.
-Keep pressing the Advance button until you get to the end of the projection. In the table below write down the final value of the catch and the final value of SB/SBF=0 that you see on the plots. Also, note down anything interesting (if anything) that you see. For example, is the short-term catch, different to the long-term catch?
+Press the Reset button in the left panel to clear the stock projection. Run through the projection again by pressing the Advance button. Make sure that you understand how the HCR uses the last value of SB/SBF=0 to set the catch limit.
+Keep pressing the Advance button until you get to the end of the projection. In the table below write down the final value of the catch and the final value of SB/SBF=0 that you see on the plots. Also, note down anything interesting (if anything) that you see. For example, is the catch at the start of the projection different to the catch near the end of the projection?
Different HCRs behave in different ways and some are better than others. The parameters of the HCR that we just tried are: Blim = 0.2, Belbow = 0.5, Cmin = 10 and Cmax = 140.
To see this, set up a different HCR by changing the HCR parameters in the left panel. Change Belbow to be 0.3 and Cmax to 130. Keep the other two parameters the same. You should see that the HCR plot has been updated to show the new shape of the HCR.
Notice that this HCR has a lower maximum catch than the previous HCR. However, the catches do not start to reduce until SB/SBF=0 is much lower (as Belbow is 0.3).
-As before, repeatedly press the Advance button and follow the evolution of the stock and the catches. Note how the behaviour of the catch and SB/SBF=0 are different to the initial example. Write down the final values and any observations in the table below.
+As before, repeatedly press the Advance button and follow the evolution of the stock and the catches. Note how the behaviour of the catch and SB/SBF=0 are different to the initial example. Write down the final values and any observations in the table at the end of this section.
As a final example set Belbow to 0.8 and Cmax to 150. Notice that this HCR has a higher maximum catch but starts reducing the catches at a higher level of SB/SBF=0. Again, repeatedly press the Advance button and follow the evolution of the stock and catch limits. Note the behaviour and final values in the final values.
+As a final example set Belbow to 0.8 and Cmax to 150 and keep the other parameters the same. Notice that this HCR has a higher maximum catch but starts reducing the catches at a higher level of SB/SBF=0. Again, repeatedly press the Advance button and follow the evolution of the stock and catch limits. Write down the behaviour and final values in the table.
In the real world, fisheries management is affected by different types of uncertainty. However, the projections we have run so far have not considered uncertainty. This means that if we rerun the projection, the outcome will always be the same (they are deterministic simulations).
-Because there is lot of uncertainty in fisheries, it is very important that a chosen HCR is robust to uncertainty. An HCR that performs well when there is no uncertainty does not necessarily perform well when there is uncertainty. We will look at this here.
+In the real world, fisheries management is affected by different types of uncertainty. However, the projections we have run so far have not considered uncertainty. This means that if we rerun the same projection, the outcome will always be the same (they are deterministic simulations).
+Because there is lot of uncertainty in fisheries, it is very important to choose an HCR that is robust to this uncertainty, otherwise the outcome may not be what you expected. For example, an HCR that performs well when future stock recruitment is stable may not perform well when stock recruitment varies a lot. We will look at this here.
We can include two sources of uncertainty: variability in the biological productivity and estimation error. Both of these sources of uncertainty can affect the dynamics of the fishery and the performance of the HCR.
Click on the Show variability option option in the panel on the left to show the uncertainty options.
Biological productivity variability reflects the natural variability in the stock dynamics, for example through variability in the recruitment, growth and natural mortality. Fisheries managers have no control over this source of uncertainty. As such it is very important than an adopted HCR is robust to this uncertainty.
+Biological productivity variability reflects the natural variability in the stock dynamics, for example through variability in the recruitment, growth and natural mortality. Fisheries managers have no control over this source of uncertainty. As such it is very important that an adopted HCR is robust to this uncertainty.
We saw in the previous examples without uncertainty that eventually the stock abundance settles down to a constant value. What happens when we include natural variability?
Set the HCR parameters back to their original values (Blim = 0.2, Belbow = 0.5, Cmin = 10, Cmax = 140). Increase the Biological productivity variability to 0.2 and project forward through time using the Advance button.
-INSERT FIGURE
You should see that the SB/SBF=0 now bounces around and is not perfectly flat. This is because the variability in the stock productivity is affecting the abundance. As the HCR uses the estimate of SB/SBF=0 to set the catch limit, it means that the catch limit set by the HCR also bounces around. This then goes onto affect the stock and so on.
@@ -410,7 +409,7 @@As mentioned above, fisheries management is affected by many types of uncertainty.
-Now turn on all the sources of uncertainty. Set Biological productivity variability to 0.2, Estimation error variability to 0.04 and the Estimation error bias to 0.1. Keep the HCR values the same as the initial values (Blim = 0.2, Belbow = 0.5, Cmin = 10 and Cmax = 140).
+Now turn on all the sources of uncertainty. Set Biological productivity variability to 0.2, Estimation error variability to 0.2 and the Estimation error bias to 0.1. Keep the HCR values the same as the initial values (Blim = 0.2, Belbow = 0.5, Cmin = 10 and Cmax = 140).
Project this forward. How are the results different to the first projection we ran which had the same HCR parameters but no uncertainty?
Use the same three HCRs that we used above. For each HCR, run about 5 full projections. Note down the final true SB/SBF=0 (the black line on the plot) and the final catch. Also note down any interesting behaviour.
-From this, is an HCR that you prefer? It’s probably quite hard to say at this stage. We shall look at this more closely in the next tutorial…
+From this, which HCR do you prefer? It’s probably quite hard to say at this stage. We shall look at this more closely in the next tutorial…
A HCR is a decision rule for setting future fishing opportunities. In this example the input to the rule is estimated stock abundance (SB/SBF=0) and the output is the catch limit in the following year.
+A HCR is a decision rule for setting future fishing opportunities. In this example the input to the rule is the estimated stock abundance (SB/SBF=0) and the output is the catch limit in the following year.
We have seen that different HCR parameterisations give different performances.
Uncertainty is a big concern in fisheries management. Here we looked at biological and estimation uncertainty. We have seen that they can change the performance of the fishery. It is very important that an HCR is robust to uncertainty. A HCR that performs well in the absence of uncertainty may not perform as well when uncertainty is present.
How do we know which HCR to use? How do we consider uncertainty? See the next tutorial!
diff --git a/tutorials/introHCR.pdf b/tutorials/introHCR.pdf index 88beccf..e8b01d8 100644 Binary files a/tutorials/introHCR.pdf and b/tutorials/introHCR.pdf differ diff --git a/tutorials/introUncertainty.Rmd b/tutorials/introUncertainty.Rmd index 92e2200..249b47a 100644 --- a/tutorials/introUncertainty.Rmd +++ b/tutorials/introUncertainty.Rmd @@ -112,7 +112,7 @@ The black line shows the current iteration (projection), the grey line the past This is the same on the HCR plot - the grey dots show the past iterations. Click **Run projection** again and again. More lines will appear. -The bars at the end of the time series plots are histograms of the values in the final year of the projection. +The bars at the end of the time series plots are histograms of the values in the final year of each projection. These histograms will fill in the more projections you run. Keep clicking **Run projection** until you get 30 or more iterations. @@ -206,7 +206,7 @@ You should consider not only the average value, but also the range of values. In the exercise above, we compared the values of catch, SB/SBF=0 and CPUE in the final year to choose a preferred HCR. We used these metrics as *performance indicators* (PIs) for comparing the performance. -Lots of different PIs are available that measure different things, for example, catch levels, changes in effort, probability of SB/SBF=0 being above the LRP etc. +Lots of different PIs are available that measure different things, for example, catch levels, changes in effort, probability of SB/SBF=0 being above the limit reference point (LRP) etc. When comparing HCRs, the chosen PIs should relate to the management objectives for the fishery. This allows you to measure how well the fishery is performing in relation to those objectives. Some HCRs will perform well for some PIs and poorly for others. This is where the ideas of prioritising PIs and evaluating trade-offs come in. @@ -236,6 +236,7 @@ Three of them we have already used above: *SB/SBF=0*, *Catch* and *CPUE*. The two new ones are: *Prob. SB > LRP* (the probability of SB/SBF=0 being above the LRP) and *Catch variability* (the amound by which catch changes over time). *Catch variability* is a bit different to the other PIs in that a *low* value is preferred (because you probably want your catches to be stable, rather than changing frequently). +A high value for *Prob. SB > LRP* is preferred because means you want to be safely away from the LRP. Click the **Run projection** button again so that we have two iterations. Notice that the values in the table have changed. We now have two extra values for each PI in (). diff --git a/tutorials/introUncertainty.html b/tutorials/introUncertainty.html index cd0253c..28f1bb2 100644 --- a/tutorials/introUncertainty.html +++ b/tutorials/introUncertainty.html @@ -361,7 +361,7 @@Click Run projection. You should see that now your projection is bumpy. Click Run projection again. You should get another time series but it is different to the previous one (the previous one is now in grey, the new one in black).
This second projection has the same stock and the same HCR as the first one but the outcome is different. The difference is a result of the uncertainty in the biological productivity and the estimation error. The different projections are known as iterations.
The black line shows the current iteration (projection), the grey line the past iterations. This is the same on the HCR plot - the grey dots show the past iterations.
-Click Run projection again and again. More lines will appear. The bars at the end of the time series plots are histograms of the values in the final year of the projection. These histograms will fill in the more projections you run.
+Click Run projection again and again. More lines will appear. The bars at the end of the time series plots are histograms of the values in the final year of each projection. These histograms will fill in the more projections you run.
Keep clicking Run projection until you get 30 or more iterations. You can see the distribution of the final values of the projection start to settle down. This means that we are starting to undertand how uncertainty affects the performance of the HCR. This is very important when it comes to selecting a HCR for a fishery. An HCR must be robust to uncertainty, otherwise it will not perform as well as expected.
In the exercise above, we compared the values of catch, SB/SBF=0 and CPUE in the final year to choose a preferred HCR. We used these metrics as performance indicators (PIs) for comparing the performance.
-Lots of different PIs are available that measure different things, for example, catch levels, changes in effort, probability of SB/SBF=0 being above the LRP etc.
+Lots of different PIs are available that measure different things, for example, catch levels, changes in effort, probability of SB/SBF=0 being above the limit reference point (LRP) etc.
When comparing HCRs, the chosen PIs should relate to the management objectives for the fishery. This allows you to measure how well the fishery is performing in relation to those objectives. Some HCRs will perform well for some PIs and poorly for others. This is where the ideas of prioritising PIs and evaluating trade-offs come in.
In the previous exercise we only looked at the values of catch, SB/SBF=0 and CPUE in the final year of the projection. We have not considered what happens during the course of the projection, only what happens at the end, i.e in the long term. When comparing HCRs we should compare what happens in the short- and medium-term as well as the long-term.
Click on the Show performance indicators button. Nothing happens - yet. Now, press the Run projection button.
You should see that a table has appeared with various PIs in it. The PIs are in the rows in the table. The PIs are measure over different time periods. The different time periods (short-, medium- and long-term) are the columns. The value in each cell of the table is the average value of that PI in that time periods.
There are 5 PIs i the table. Three of them we have already used above: SB/SBF=0, Catch and CPUE. The two new ones are: Prob. SB > LRP (the probability of SB/SBF=0 being above the LRP) and Catch variability (the amound by which catch changes over time).
-Catch variability is a bit different to the other PIs in that a low value is preferred (because you probably want your catches to be stable, rather than changing frequently).
+Catch variability is a bit different to the other PIs in that a low value is preferred (because you probably want your catches to be stable, rather than changing frequently). A high value for Prob. SB > LRP is preferred because means you want to be safely away from the LRP.
Click the Run projection button again so that we have two iterations. Notice that the values in the table have changed. We now have two extra values for each PI in (). The single value is the median (average) of the two iterations. The values in the () are the 20 and 80 percentile. The 20 and 80 percentile represent low and high bounds of the PI with the majority of values of the PI falling between them.
Note that the Prob. SB > LRP PI does not have percentiles. This is because it is a probability.
It doesn’t make much sense to calculate the median or the percentiles this for just 2 iterations. Keep clicking Run projection until you have about 50 iterations. You should see that the numbers in the table start to settle down (in the same way that the histogram settles down).
diff --git a/tutorials/introUncertainty.pdf b/tutorials/introUncertainty.pdf index 5de4252..6667e67 100644 Binary files a/tutorials/introUncertainty.pdf and b/tutorials/introUncertainty.pdf differ diff --git a/tutorials/measuringPerformance.Rmd b/tutorials/measuringPerformance.Rmd index 3afa64b..b92782b 100644 --- a/tutorials/measuringPerformance.Rmd +++ b/tutorials/measuringPerformance.Rmd @@ -20,7 +20,7 @@ fontsize: 12pt In the previous tutorial (*Introduction to Uncertainty and Performance Indicators*) we looked at the performance of HCRs by running a large number of individual projections. We included two sources of uncertainty (*biological productivity variability* and *estimation error*) and began to look at how to compare the performance of different HCRs using Performance Indicators (PIs). -The more iterations (number of projections) we had, the better the estimate of uncertainty in the PIs. +The more iterations (number of projections) we had, the more stable the histograms were and the better the estimate of uncertainty in the PIs. In this tutorial we build on this by assembling a basket of candidate HCRs, calculating a range of PIs and comparing their performance in a number of ways. @@ -63,9 +63,9 @@ This runs a projection with 100 iterations. In the previous tutoral we ran one i There are 8 PIs in the table. *SB/SBF=0* and *Catch* are fairly self explanatory. *Effort (rel. 2018)* and *CPUE (rel. 2018)* are the fishing effort and CPUE relative to their values in 2018 respectively. *Prob. SB > LRP* is the probability of SB/SBF=0 being above the LRP. *Catch variability*, *Effort variability* and *CPUE variability* measure the variability in the catch, relative effort and relative CPUE respectively. They measure how much the catch etc. change over the time (the bumpiness in the plots). The higher the value, the more the value changes over time. It should be noted that we don't necessarily want high values for all of the PIs. -It is generally thought that the higher the value of *Prob. SB > LRP*, *Catch* and *CPUE (rel. 1999)* the better the HCR is performing. +It is generally thought that the higher the value of *Prob. SB > LRP*, *Catch* and *CPUE (rel. 2018)* the better the HCR is performing. -However, for *Effort (rel. 1999)* and the three *variability* PIs, lower values are preferred. High effort implies high costs (something we want to avoid). +However, for *Effort (rel. 2018)* and the three *variability* PIs, lower values are preferred. High effort implies high costs (something we want to avoid). Stable catches and effort are preferred to catches and effort that varying strongly between years. @@ -98,7 +98,7 @@ Each panel shows the median (average) value of a PI for each of the HCRs in the We are looking at the median values of 8 PIs for 3 HCRs in 3 different time periods. This is a lot of information! We want to be able to choose which HCR best fits our objectives but it can be difficult when there is so much to look at. diff --git a/tutorials/measuringPerformance.html b/tutorials/measuringPerformance.html index 94aaa47..63b411d 100644 --- a/tutorials/measuringPerformance.html +++ b/tutorials/measuringPerformance.html @@ -317,7 +317,7 @@In the previous tutorial (Introduction to Uncertainty and Performance Indicators) we looked at the performance of HCRs by running a large number of individual projections. We included two sources of uncertainty (biological productivity variability and estimation error) and began to look at how to compare the performance of different HCRs using Performance Indicators (PIs). The more iterations (number of projections) we had, the better the estimate of uncertainty in the PIs.
+In the previous tutorial (Introduction to Uncertainty and Performance Indicators) we looked at the performance of HCRs by running a large number of individual projections. We included two sources of uncertainty (biological productivity variability and estimation error) and began to look at how to compare the performance of different HCRs using Performance Indicators (PIs). The more iterations (number of projections) we had, the more stable the histograms were and the better the estimate of uncertainty in the PIs.
In this tutorial we build on this by assembling a basket of candidate HCRs, calculating a range of PIs and comparing their performance in a number of ways.
The initial values of the HCR parameters should be: Blim = 0.2, Belbow = 0.5, Cmin = 10 and Cmax = 140. If not, set these parameters. Press the Project HCR button to run the projection. This runs a projection with 100 iterations. In the previous tutoral we ran one iteration at a time. Now we are running a 100 at a time. The results can be seen in the time series plots and the table of PIs (hopefully, this looks familiar to the previous tutorial).
There are 8 PIs in the table. SB/SBF=0 and Catch are fairly self explanatory. Effort (rel. 2018) and CPUE (rel. 2018) are the fishing effort and CPUE relative to their values in 2018 respectively. Prob. SB > LRP is the probability of SB/SBF=0 being above the LRP. Catch variability, Effort variability and CPUE variability measure the variability in the catch, relative effort and relative CPUE respectively. They measure how much the catch etc. change over the time (the bumpiness in the plots). The higher the value, the more the value changes over time.
-It should be noted that we don’t necessarily want high values for all of the PIs. It is generally thought that the higher the value of Prob. SB > LRP, Catch and CPUE (rel. 1999) the better the HCR is performing.
-However, for Effort (rel. 1999) and the three variability PIs, lower values are preferred. High effort implies high costs (something we want to avoid). Stable catches and effort are preferred to catches and effort that varying strongly between years.
+It should be noted that we don’t necessarily want high values for all of the PIs. It is generally thought that the higher the value of Prob. SB > LRP, Catch and CPUE (rel. 2018) the better the HCR is performing.
+However, for Effort (rel. 2018) and the three variability PIs, lower values are preferred. High effort implies high costs (something we want to avoid). Stable catches and effort are preferred to catches and effort that varying strongly between years.
SB/SBF=0 is slightly more complicated. We don’t want this value to be too low (we want it away from the LRP). However, high values of SB/SBF=0 imply that there is some forgone catches. Ideally, SB/SBF=0 should be measured against the TRP.
Care must therefore be taken when using PIs to compare performance of HCRs. Not all of them should be high.
Looking at the summary plots and the table of PIs, we think that this HCR is worth considering in more detail. Click on the Add HCR to basket button to add the HCR to the basket of candidate HCRs. You should see that the counter Number of HCRs in basket increases by 1.
@@ -366,7 +366,7 @@
We are looking at the median values of 8 PIs for 3 HCRs in 3 different time periods. This is a lot of information! We want to be able to choose which HCR best fits our objectives but it can be difficult when there is so much to look at.
To make things easier, we can drop PIs that we think are unimportant (perhaps they do not measure anything related to your management objectives) by deselecting them from the list in the left panel. Similarly, HCRs can be deselected if they are thought to be of no interest.
diff --git a/tutorials/measuringPerformance.pdf b/tutorials/measuringPerformance.pdf index a0c0157..6ed63a6 100644 Binary files a/tutorials/measuringPerformance.pdf and b/tutorials/measuringPerformance.pdf differ