Developing a Historical Precipitation Record - Satellite-based Applications on Climate Change - page 93

7.1 Introduction

Global precipitation can be monitored using a number of satellite-based estimates
combined with gauge estimates over land. As discussed in Ferraro et al.'s paper on
global precipitation in this topic, satellite-based precipitation estimates began in
1979 using infrared instruments. Over time more satellite estimates from both
infrared and microwave instruments were developed. The Global Precipitation
Climatology Project (GPCP, Adler et al. 2003 ) adjusted these different estimates
to minimize satellite-to-satellite bias and merged them with gauge data to produce a
monthly product for climate analyses beginning 1979.
Climate variations have been documented over multi-decadal time scales (e.g., see
Trenberth et al. 2007 ). The available gauge-based analyses can describe large-scale
precipitation variations over most land regions beginning about 1900. However, most
of the Earth's surface is covered by oceans, and evaluation of that component of the
global hydrologic cycle has required satellite sampling. For most of the twentieth
century, there is a large gap in the record for ocean regions. Filling the gap would
allow the historical response to climate change to be better understood and would
provide historical perspective to precipitation monitoring products. In addition,
climate models used to forecast precipitation changes with changing global
temperatures could be better validated and improved using oceanic precipitation
over the twentieth century. Therefore, attempts have been made to reconstruct
historical oceanic precipitation. This chapter briefly reviews those precipitation
reconstructions and discusses how well historical oceanic variations can be evaluated.

7.2 Historical Reconstructions

A reconstruction uses the available historical data and globally complete statistics
describing those data to perform an analysis. For most global climate fields, satellite-
based data are critical for developing the reconstruction statistics. Since the mean-
annual cycle is defined by these modern base data, reconstructions are performed
on anomalies from the annual cycle. The annual cycle can be added back onto
the reconstruction anomaly later if desired. An example is the historical sea-surface
temperature (SST) anomaly reconstruction (Smith et al. 1996 ), based on covariance
maps produced using a satellite and in situ SST analysis. The covariance maps
were produced using empirical orthogonal function (EOF) analysis (e.g., see Davis
1976 for a detailed definition of EOF analysis). An EOF analysis (also called principal
component analysis) decomposes a time series of maps into a set of spatial fields or
modes, E i ( x ), and associated time series, a i ( t ), for a set of modes i ¼

, N .The
modes are only a function of space, x , while the time-series weights for themodes are a
function of only time, t . Each mode represents an orthogonal component of the full
field's variance, and an approximation of the true field, F ( x , t ), can be reconstructed
from the weighted sum of the modes,

1, 2,

...