I developed my first wind drawing process in 2005 using analog tools. I attached pieces of plastic to a pen and suspended it above a sheet of paper. Left to draw for a day, the resultant drawing told the story of that day's wind conditions. With the top of the drawing oriented toward north, I could tell which direction the wind came from and the amount and quality of marks indicated how long and fast the wind was blowing.

Curiosity about the forces that made each drawing led me to develop digital wind drawing processes. Using weather instruments to measure wind conditions, I connect the instruments to a compute and collect data about wind speed, direction, time and date. The data become a raw material to work with as I write software to generate new works.

Sketching with the software, I try different ways of organizing formal elements in a two-dimensional space, playing with the appearance of and relationship between formal elements. When all of the parameters work well together, I end up with a digital drawing process that shows how the wind changes throughout a day.

Each process is rooted in a question of how to use wind speed, direction, time of day, and date to generate form. Some processes visualize a single day's wind conditions in a single image. Others show how winds on one day compare to winds on other days.

I have developed processes that show how wind conditions change in real time on a screen and lately, I have been developing processes to visualize wind conditions as three-dimensional forms.

In the last year, I've been exploring ways to show wind conditions through time, to simultaneously see the similarities and differences of how winds change throughout a day and/or a year. The work generated with Process.2021.01 above is a three-dimensional form that shows average hourly wind conditions at one location on different days. The piece 3D printed in white above shows wind conditions on two consecutive days in Boone, NC. One day's winds were stronger than the other, creating a lofted surface that was taller and wider than another. Process.2022.01 generates a screen-based visualization of real-time and historical wind conditions at the time it is viewed with a system of curves that move across the screen as if they were pushed by the wind. Its colors are mapped to wind direction and the curve cluster sizes relate to the year they represent. Clusters with more curves are more recent.

Plan A

My original concept for the artwork Transformations was to use etched plexiglass to show marks for each hour of a day and each day of a year when the wind was blowing. By layering the marks of one year over others, hours and days line up from layer to layer. The x-axis is the hour of day— midnight to 11pm—and the y-axis is day—January 1 to December 31.

Using weather data found online, I used wrote software in Processing to generate a composition that visualizes Boone’s wind conditions for each hour and day in a year. Each hour’s average wind speed and direction are represented by a set of arcs. The number of arcs represent miles per hour and the arcs’ degree of rotation is keyed to wind direction. Arc rotations were limited to one of sixteen possible angles. As I developed the software, I added text elements that showed information about each group of arcs to ensure that they were placed and oriented correctly.

When the software was ready, I used it to generate a PDF filled with curves for each year of data. As you can see, the density of arcs is higher at the top and bottom meaning Boone's winds are stronger in fall, winter and spring and often blow throughout the day. Summer months are calmer and winds mostly appear during daylight hours.

With the design ready, the next step was to develop the fabrication process. I knew I wanted to etch into a transparent material. A few summers ago I had conducted some experiments with using a laser cutter to etch into glass. While I liked the look of the frosted glass, the lines for each arc were too thick. With a number of arcs ranging from 39,000 to 51,000 a year, I needed an etching with much thinner lines at a large scale.

My App DigiFab colleagues recommended that I try using a computer numerical control (CNC) router to carve into plexiglass and achieve the scale I wanted. To control the router, a toolpath is needed to describe the path through space the cutting tool will follow. Toolpaths are written in G-code, a programming language for in computer-aided design and manufacturing (CAD/CAM). The 3D design application Rhino has a plug-in called RhinoCAM that creates toolpaths and simulates a CAM process.

I wanted to make some tests of a small section of each layer so imported the PDFs into Rhino, selected a small section of arcs and used RhinoCAM to create toolpaths for each layer. Using a scrap piece of plexiglass, I made some tests with an engraving bit. They looked better than the laser cutter, but the depth of the cut was so small that the bit sometimes wouldn't cut into the plexiglass if there was any variation in its surface. My colleagues suggested that I try a diamond drag bit that is spring loaded and will maintain contact with the cutting surface. I bought one and was thrilled with the results.

After making 6x6 inch tests for each year's layer, the next step was to attempt a full scale etching on a 48x48 inch sheet of plexiglass. To do so, the RhinoCam had to generate toolpaths to engrave tens of thousands of arcs. This took a long, long, long time. So long, in fact, that I wouldn't have enough time to complete the piece in time. I had to scale back the scope of what I wanted to do.

Plan B

With no end in sight for generating Plan A's toolpaths, it was time for Plan B. Rather than show 17,520 (24 hours x 365 days) hours per layer, I decided to show twenty-four consecutive days instead with only 576 hours per layer. Limiting the number of days to the same number of hours per day opened up some possibilities for interesting structural relationships between intersecting and overlapping sets of arcs. In many ways, I preferred the new design to the first one.

After changing the code in Processing to fit the new parameters, I generated new PDFs for each year's layer and brought them into Rhino. RhinoCam was able to generate toolpaths for the new design fairly quickly.

Fabrication

Now ready to go, I mounted a 48x48 inch piece of plexiglass to the ShopBot's bed, securing its position with double-stick carpet tape and clamps. After removing the film from the plexiglass' top, I set X, Y and Z-axes to zero and started the engraving process.

Once the engraving was complete, I replaced the engraving bit with a 1/8-inch O flute and cut four mounting holes and trimmed the outer edges with a profile cut. Doing these cuts immediately after the engraving while the plexiglass was still on the bed ensured that the position of the arcs related to each hour and day would line up from layer to layer. It took about four days to engrave and trim all three layers.

I'm delighted with how this piece turned out. I never knew how it would look until it was installed and I love how the shadows cast from each layer become one.

I plan to take more photographs of the piece and add/replace to these in the future.

Moving forward, I hope to make more works using this process. As I do, I'll stay open to changing the design's parameters. I enjoy how elements from one layer to another relate to each other and I wonder if I can develop a design that allows for more emergent formal elements when layers overlap.

I'm still interested in visualizing several years of wind date for one location, but I need to either develop a new process to make that achievable or work at a different scale.

I've always been interested in how my wind drawings relate to time and place and as I refine this process, I'll look for interesting stories the wind can tell about specific places through time.

Finally, as I develop more generative processes I plan to explore fabrication with different materials.