Continuous thunderstorm monitoring: Retrieval of precipitation parameters from lightning observations

I. A long-range VLF receiver network to continuously monitor thunderstorms: Experimental location error calibration and validation. Lightning emits radio noise (sferics) over a broad region of the electromagnetic spectrum. In the Very Low Frequency domain sferics propagates thousands of kilometers in the earth-ionosphere waveguide. An experimental network of five ground-based radio receivers (STARNET), situated along the US east-coast and Puerto Rico, measured sferics between July 1997 and February 1998. Those measurements are compared with measurements from the National Lightning Detection Network and Lightning Imaging Sensor. Analysis showed relative location errors below 20-km and up to 100-km over the East and West Coast of US, respectively. Simulations based on expected error sources could explain most of those errors. A location error correction scheme, which combines theoretical/experimental models and satellite infrared observations, reduced location errors to less than 60-km in the western part of US. ^ II. A long-range VLF receiver network to continuously monitor thunderstorms: Cloud-to-ground and intra-cloud detection efficiency. The study computes STARNET\u27s cloud-to-ground (CG) and intra-cloud (IC) detection efficiency (DE) and bulk IC:CG ratios over the US using established measurements from NLDN and the Lightning Detection and Ranging (LDAR) systems. CG-DE shows exponential range dependence with 100% within the network and ∼20% at 4,000 km range. Day and night DE shows similar decay. IC and CG DE values evaluated using WAR data were in the range of 57%–70% and 7%–19%. The bulk IC:CG ratios over the continental US showed a range of 0–4, while in a finer scale over Florida they followed the storm variability detected by LDAR. ^ III. Rainfall retrieval from lightning and satellite infrared observations adjusted with TRMM precipitation radar. A new surface rainfall estimation algorithm based on lightning and satellite infrared observations is developed. The algorithm uses as reference the precipitation fields from the TRMM Precipitation Radar (PR). The algorithm used a total of 631 TRMM orbits from December 1997 to January 1998 to determine the rain rate and rain area relationships. The algorithm underestimates rain area for both cloud types by ∼20%, while for the rain volume it overestimates by 19.54% (lightning clouds) and 12.52% (non-lightning clouds) with respect to PR. The estimated mean diurnal cycle of rainfall and convective/stratiform rain distributions agree with PR. Hourly rain gauge comparisons showed an overestimation of ∼6%, while at monthly scale the bias is 2.4% (0.27%) for 1-degree (2-degree) grids. Lightning augmentation reduces the rain area and volume bias by 28% and 30%, respectively. Rain gauge comparisons show bias reduction ranging from 10% to 38% for monthly estimates, and up to 87% for hourly 0.1-degree estimates.