Page 1 of 1

CALIPSO isolated 80 km layer filter

Posted: Wed Jun 21, 2023 4:06 pm America/New_York
by ses2271
Hi there,

I have some questions about applying the Isolated 80 km layer filter to the L2 data. My understanding from Tackett et al., 2018 is that the L3 data product is derived from the L2 profile product, which in turn is derived from the L2 layer product. However, it appears that the Isolated 80 km layer filter (as well as the misclassified cirrus fringe filter) are applied to the layer product. Have these filters been applied to the L2 profile product?

If not, is there a way to apply them to the profile product to create values comparable to the L3 data? In particular, I'm uncertain whether the "80 km" and "80 km w/ subgrid feature detected at 1/3 km" flags in the atmospheric volume description would both be filtered in the case of an isolated layer.

Secondly, since the filter applies to an entire layer, which may span multiple height bins in the profile product, is there a computationally efficient way that these filters can be applied to the profile product?

As always, thank you for your response!


Re: CALIPSO isolated 80 km layer filter

Posted: Tue Jun 27, 2023 12:53 pm America/New_York
by ASDC -

Thank you for your interest in CALIPSO data. The author of the article that you mentioned has been notified and we are waiting on a response. We apologize for the wait.



Re: CALIPSO isolated 80 km layer filter

Posted: Wed Jun 28, 2023 9:04 am America/New_York
by ASDC -
Hi Sarah,

Here is the response from the author Jason Tackett of the CALIPSO Science team:

Hi Sarah,

The isolated 80 km layer filter and the misclassified cirrus fringe filters make use of information in the L2 layer product to filter L2 profile data during the construction of L3. These filters are not active in the L2 layer or profile products. Rather, the filters are only applied to create the L3 product. Applying these filters to the L2 profile product requires knowledge of which range bins contain unique layer detections. The L2 5-km layer products report the top and base altitudes of all layers detected and the horizontal resolution required for detection. Because the layer and profile products report information at 5-km resolution, layers detected at 20 km and 80 km resolution are replicated on the 5-km grid to span multiple profiles. Even though a 20 or 80 km resolution feature is reported at 5-km resolution, it is important for the entire 20 or 80 km feature to be treated as a unique feature detection. So the trick for implementing these particular L3 filters is to know which range bins and profiles are associated with unique layer detections. Doing so is a bit involved. Here are the basics of the procedure. The goal is to create an array indicating the location of unique layers that maps directly to the L2 aerosol profile product atmospheric volume description SDS. Then the combination of the atmospheric volume description and the new array, described below, can be used to filter the L2 aerosol extinction.

Create a new array having the same size as the atmospheric volume description SDS. Let’s call it layer_ID. The merged layer product can be used to ‘paint’ the layers that are detected upon this new array. Within the L2 5-km merged layer product, the following SDSs are needed: layer top altitude, layer base altitude, and horizontal resolution. As background, L2 processing operates on 80-km ‘chunks’ of data at a time (16 profiles at 5-km resolution). The first profile in a L2 5-km product marks the beginning of the first chunk. Features are detected at 5-km resolution first, then 20 km, then 80 km. For a specific resolution, there can only be one layer with a given top altitude (e.g., only one 5-km resolution layer will have a top altitude of 1.23 km; all other 5-km detections in the profile will have tops at different altitudes). To create this layer_ID array, we need to add the unique layers on the new array in reverse order, such that higher resolution layers are painted over lower resolution layers. As each unique layer is painted, all range bins containing the layer are given a unique value.

First, paint on the 80 km resolution features, giving each unique layer a unique ID in the array. To do this, loop through the file in 16 profile increments (i.e., evaluate each chunk). Search the 16 5-km resolution profiles within the chunk for layers detected at 80-km resolution. Because there can be only one 80-km resolution layer with a specific layer top altitude (let’s call it ztop1), all instances of ztop1 for 80-km resolution layers within that chunk are part of the same unique layer. Use the layer top and base altitudes for that layer to paint its unique ID on the layer_ID array for each profile where it exists.

Next, paint on the 20 km resolution features overtop of the layer_ID array by looping through the profile in 4 profile increments (i.e., evaluate each 20 km segment). As before, search the four 5-km resolution profiles for layers detected at 20-km resolution. Identify where unique 20-km resolution layers exist by evaluating their layer top altitudes. Profiles having the same top altitude at 20-km resolution belong to the same unique layer. Add their unique IDs to the layer_ID array in profiles where they exist.

Finally, paint on the 5 km resolution features. By definition, each 5 km layer is unique, so this involves looping through all profiles in the file and painting on unique IDs for 5 km resolution layers based on layer top and base altitudes.

Once the layer_ID array is created, there is a one-to-one mapping of unique layer IDs to the atmospheric volume description array. So, to apply the isolated 80-km resolution layer filter, first identify the location of 80 km resolution aerosol layers using the atmospheric volume description. Use the layer_ID array to determine the ID for these layers. For a given layer ID, search for any other layers in contact with range bins having that layer ID. If there are none, then the 80 km layer is isolated. For the cirrus fringe filter, each unique aerosol layer needs to be examined using their unique IDs to determine if it is adjacent to an ice cloud by comparison with information in the atmospheric volume description (and altitude and temperature criteria for this filter).

There could be other ways to accomplish this task. The CALIPSO project is considering adding a unique layer ID SDS in a future release that will report something similar to what is described here which would make this a lot easier.

Regarding the question about "80 km" and "80 km w/ subgrid feature detected at 1/3 km" flags, both are evaluated by the isolated 80-km layer filter.

Hope this helps!