1710 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 21 Development of an Inexpensive Raindrop Size Spectrometer WILLIAM HENSON,* GEOFF AUSTIN, AND HARRY OUDENHOVEN Atmospheric Physics Group, Department of Physics, University of Auckland, Auckland, New Zealand (Manuscript received 24 June 2003, in ®nal form 6 May 2004) ABSTRACT The deployment of weather radar, notably in mountainous terrain with many microclimates, requires the use of several or even many drop size spectrometers to provide con®dence in the quantitative relation between radar re¯ectivity and rainfall. While there are several different commercial disdrometers available they are all ex- pensive, large, or fragile, which militates against multiple deployment in the ®eld. The design brief was for a reasonably accurate and sensitive, low-cost and rugged disdrometer to support ®eld work. A design based on piezoceramic disks normally used in hydrophones is described. Calibration and typical ®eld results are presented. 1. Introduction Joss (2000) provides a brief overview of disdrometer technology. The measurement of rainfall and its prediction have The development of a reliable disdrometer system is been crucial throughout the ages. Mankind has always nontrivial, as there can be many and various sources of depended on water, not only to drink but also to grow error (largely electrical and environmental) that can crops and to cook. Countries have even gone to war cause results to be compromised. The main aim of de- over access to water (Ziegler 1987). The desire to pre- veloping a low-cost disdrometer system was, ®rst, to dict rainfall has fuelled developments in the study of complement the current equipment already used by the rainfall processes. However, until the advent of weather Atmospheric Group at the University of Auckland (New radar, the study of raindrops and raindrop size spectra Zealand). This includes a very high space±time-reso- was not seen as being greatly important, except for soil lution radar system, a dense network of high-time-res- erosion processes (Laws and Parsons 1943) where by olution drop-counting rain gauges (Hosking et al. 1986) large drops have a much greater impact than the vol- and meteorological towers. A secondary aim was to then umetric equivalent number of small drops. Radar cross use the disdrometer(s) to discover some basic relations sections depend on the sixth moment of the raindrop in the raindrop spectra not only for individual rain size distribution, whereas rainfall rate is proportional to events but also for different synoptic conditions. It is the diameter to the third power times the fall speed of hoped that with the construction of up to half a dozen the raindrops. Therefore, knowledge of the distribution disdrometers, used together in conjunction with a dense at the time of measurement is important if an accurate rain gauge network and a high-resolution X-band radar, estimate is to be made. In recent years the Joss±Wald- the study of raindrop spectra in relation to dual Z±R vogel disdrometer (JWD; Joss and Waldvogel 1967) has measurements can be advanced along paths that up until been the most common disdrometer system in wide- now have not been tried. The construction of the new spread use to measure raindrop sizes and, therefore, disdrometer system was based around piezoceramic raindrop spectra. As the cost of a single JWD unit can transducers that allowed us to achieve the goals we had be prohibitive (at least in New Zealand), the study of in mind. raindrop spectra has not advanced as far as it could haveÐeven though many types of automatic disdro- 2. Design and construction meter systems have been developed. LoÈef¯er-Mang and In order to ful®ll the design criteria of the project (i.e., low cost and accuracy), a dif®cult set of opposing con- * Current af®liation: Atmospheric/Oceanic Sciences Department, straints needed to be satis®ed. There had been a previous McGill University, Montreal, Quebec, Canada. development of a disdrometer system at Auckland Uni- versity (Camilleri 2000), and it was decided that pursuing Corresponding author address: Dr. William Henson, McGill Uni- a similar line but basing the design around piezoceramic versity, Montreal, QC H3A 2K6, Canada. crystals rather than one-off marine hydrophones would E-mail: [email protected] be a good idea. Piezoceramic crystals have the advantage q 2004 American Meteorological Society Unauthenticated | Downloaded 09/27/21 08:37 AM UTC NOVEMBER 2004 HENSON ET AL. 1711 of being cost effective, readily available in a variety of materials, and accurate over a large range of pressures. Being ceramic, the disks by themselves are quite fragile; however, by effectively encasing them (as seen in Fig. 1), they can be given suf®cient protection against not only the elements but also the knocks and rough treatment that may occur during shipment. One unforeseen problem that did occur with the piezoceramics was that typically the disks oscillate in a radial mode (perpendicular to the travel of the raindrop) as well as in the planar mode (perpendicular to the face of the piezoceramic). This caused a high-frequency signal to be introduced onto the main pulse signal. This high-frequency signal was suc- cessfully ®ltered using a combination of electronic and mechanical ®ltering. Another problem that impacts dis- drometers is sensitivity to being shaken by the wind and to aerodynamic lifting caused by wind ¯owing over the cap. To solve this the disdrometers were typically de- ployed in wind baf¯es similar to those seen in Fig. 2. The effect of the wind on the disdrometer is for the disdrometer to incorrectly sample an increased number of small raindrops. This could obviously give a wildly incorrect raindrop size distribution measurement; how- ever, due to the rain rate and radar re¯ectivity being the third and sixth moments of the raindrop size distribution, even if a wind baf¯e is not used, the wind will have little FIG. 1. Magni®ed view of the disdrometer. effect on a Z±R relationship measurement. Tests con- ducted showed that if the wind speed is anticipated to peak over 5 m s21, then for the best results the disdro- meter should be operated inside a wind baf¯e. The decision on how large to make the disdrometer was made using a combination of two different criteria. FIG. 2. A pair of disdrometers inside a wind baf¯e. Unauthenticated | Downloaded 09/27/21 08:37 AM UTC 1712 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME 21 FIG. 3. Picture of the circuit board mounted inside the disdrometer body. First, the disdrometer should be large enough so that conductors. This microprocessor allowed us to utilize the chance of a raindrop striking the edge of the cap is several channels of its analog-to-digital converter and minimized, and second, the disdrometer should be small thereby increase the effective number of bits. Future enough so that the chance of overlapping impacts is work will include using all of the available analog-to- minimized. This was done assuming a Marshall±Palmer digital channels with different gains so that a linear raindrop size distribution and a rain rate of 10 mm h21 approximation to a logarithmic ampli®er can be pro- and that at least 90% of the raindrops fully impact on duced. One advantage of the MC68HC912 is that it is the cap. The dead time was 30 ms, and the number of designed to be both forward and backward compatible raindrops sampled (i.e., not subject to being obscured with Motorola microprocessors, so upgrading the chip by an overlapping impact) was at least 90%. The two in the future will be straightforward. With the choice criteria resulted in two curves that intersected at a dis- of microprocessor and ®lter (for the removal of the high- drometer radius of approximately 35 mm; this was the frequency signal from the piezoceramic), the power con- radius we chose. sumption meant that the disdrometer could operate for The ``brain'' of the disdrometer had to be a battery- in excess of 5 days unattended and at 60-s temporal powered microprocessorÐa computer as a datalogger resolution for approximately 13 days before the memory was simply unacceptable in terms of portability and would have to be downloaded and cleared. A picture of noise contamination through the power supply. It was the circuit board mounted inside the disdrometer can be decided to use a MC68HC912 from Motorola Semi- seen in Fig. 3. During the initial tests of the disdrometer, it was ap- parent that the initial voltage spike generated by the piezoceramic after a raindrop impact had approximately 250 ms to the peak for all raindrop sizes tested, and the longest ringing time (often referred to as the dead time) expected was 30 ms. A typical response from the dis- drometer to a water drop impact can be seen in Fig. 4. Using the speed of the microprocessor it was possible to take successive A/D samples and, using basic logic, determine when a peak was found (i.e., the middle of three recordings being higher than the outside two re- cordings and above a threshold). The microprocessor FIG. 4. Typical response from the disdrometer to would then use a parabola-®tting algorithm to estimate a water drop impact. the peak voltage, store the value, and then wait for 30 Unauthenticated | Downloaded 09/27/21 08:37 AM UTC NOVEMBER 2004 HENSON ET AL. 1713 (i.e., a power-law relationship) has been seen before with the Joss±Waldvogel disdrometer, but differences are expected, as the Joss±Waldvogel disdrometer relies on the raindrop to physically move the Styrofoam cap and measures the voltage induced as a magnet passes a set of coils.