CLIMATE VARIABILITY, AGRICULTURE AND REMOTE SENSING - Food Security, Food Prices and Climate Variability - page 55

Precipitation has very large spatial and temporal variability, and therefore is very difficult
to model. Rainfall measurements need to be very accurate over a wide range of spatial and
temporal variability (Zeng, 1999). Capturing this day-to-day and year-to-year variability of
rainfall is very difficult in a model. Rain gauge observations are often biased due to the effect
of wind and other factors (Sevruk, 1982), but in most areas this bias is relatively small com-
pared with satellite precipitation estimates based on cloud identification or rain rates that
either systematically overestimate or underestimate the amount of actual rain falling (Xie and
Arkin, 1995). Gauge data are the basis for all methods to estimate precipitation, and therefore
the frequency of observation, density of the network and accuracy of each measurement is
critical for the quality of the rainfall models, regardless of the other inputs. Rain gauge data
are not available over most oceanic regions and sparsely populated regions. Averaging point
values on a sparse, irregular grid into surface means introduces sampling errors that can be
significant and non-random. Places with high variability of rainfall and low sampling with
gauges will have systematically incorrect rainfall estimates (Nicholson, 1986; Grist and Nichol-
son, 2001).
Agriculturally relevant drought conditions both regionally and globally are much more
clearly assessed than long-term trends because most droughts are measured over a period of
months, a relatively short period of time that is well suited to satellite rainfall data records.
New online tools such as the Global Drought Monitor are now available to provide timely
and quantitative information about regions experiencing drought. Drought is defined as an
extended period of time when a region notes a deficiency in its water supply, whether surface
or groundwater. This generally occurs when there is consistently below average precipitation.
Because drought is often defined in terms of ecosystems and economic damages, a global
drought map is difficult to interpret and determine if it is significant for food security or food
production in a particular area. Nevertheless, these products will help the food security
community to better understand rainfall variability and its change over time (Bolten et al .,
2010; Rojas et al ., 2011).
Despite the technical difficulty of creating global precipitation datasets, satellite-derived
rainfall datasets have greatly improved in recent years and form the basis for much knowledge
of drought and drought impacts (Huffman et al ., 1995). They do have some drawbacks,
however, for early warning of food security crises, since the accuracy of the products depend
directly on the density and fidelity of ground observation networks. Countries such as Kenya,
for example, have hundreds of meteorological stations, but most are not reported to the
World Meteorological Organization's Global Telecommunications Network in near real
time, or made accessible to remote sensing scientists seeking to create global, gridded
precipitation products. Kenya reports fewer than ten of these stations to the international
networks daily, reducing significantly the ability of satellite estimations to capture the actual
variability of rainfall within Kenya. Reliance on local provision of observations makes the
global dataset vulnerable to significant increases in error when political and economic problems
result in a cessation of reporting of rainfall data to global networks. Thus the error structure
of satellite-derived rainfall datasets is more related to the political and economic circumstances
of the region than it is to any biophysical variable (Brown, 2008).
Despite new satellites that capture precipitating clouds and humidity such as the Tropical
Rainfall Measuring Mission (TRMM), satellite-based gridded rainfall data often fail to capture
adequately extreme events that may bring much of the rainfall in a season. The TRMM
Multi-satellite Precipitation Analysis (TMPA) data product provides rainfall at 0.25 degree