Exploring dry and wet conditions in Indonesia
Month by month from January 1950 to December 2025, for every province, kabupaten or kota, and kecamatan. Three standardised indices, computed on TerraClimate v1.1, show rainfall, the balance of rainfall and evaporative demand, and evaporative demand alone.
El Niño and La Niña
How to read this
The three indices
- SPI standardises rainfall. Negative is dry.
- SPEI standardises rainfall minus potential evapotranspiration. Negative is dry. Here it uses the SPI class table, which is a working choice, not a published standard for SPEI.
- EDDI standardises evaporative demand alone. Positive is high demand. Its cool side is low demand, which is not the same as wet. Its outermost categories cannot occur with a 30-year baseline and are left out of the legend.
SPEI shares rainfall with SPI and evaporative demand with EDDI, so two of them agreeing is not independent confirmation.
Accumulation period
A short period has little room to move where the season is normally almost dry. In Indramayu, August and September 1997 had 3 mm and 1 mm of rain in a strong El Nino year, yet the 3-month SPI stayed above -1: the 1991 to 2020 median for those two months is 10 mm and 6 mm, and nearly half the baseline years had under 5 mm. The 6-month SPI and evaporative demand both show that season as dry. The WMO guide makes the same point: 1-month and 3-month SPI "may be misleading in regions where it is normally dry". The chart of all four accumulation periods is there for this reason.
The period is the window of months being summed, not how long a dry or wet spell has lasted. The World Meteorological Organization's guide describes 3-month SPI as "a seasonal estimation of precipitation" and 12-month SPI as reflecting "long-term precipitation patterns" (WMO-No. 1090, 2012). That wording is for SPI; it is used here as a plain label for the other two indices.
Classes, share of area and events
- SPI classes follow WMO-No. 1090, Table 1: moderately dry from -1.0, severely dry from -1.5, extremely dry from -2.0, and the mirror image for wet. EDDI categories are the ones stored with the data.
- The share chart shows one bar per month, divided among all the classes and adding to 100%.
- Share of area counts every 1/24 degree grid cell (about 21 square kilometres) that a boundary touches, each at its full area. Cells on a boundary count on both sides. For a small kabupaten or kota, and for almost every kecamatan, this means a handful of cells, much of them shared with neighbours, so its percentages are coarse.
- Events follow WMO-No. 1090 section 6.1: an event starts in the first month the index reaches -1.0 and ends when it becomes positive. Here the rule is applied to the mean index over the province, kabupaten or kecamatan, and it is extended to wet events and to SPEI and EDDI, which the guide does not cover.
Less reliable periods
Two parts of the record are kept but marked.
- Before 1970, everywhere. At the cells that hold rain gauges, TerraClimate's annual rainfall, measured against each one's own 1991 to 2020 normal, is about 10% of normal lower than the gauges in the 1950s and 4% lower in the 1960s. The 1950s gauges are almost all in Java.
- Before 1984, in northern and central Sumatra and West Kalimantan. TerraClimate's rainfall and evaporative demand both shift at the start of 1984 there. Three gauges inside the region read near normal in 1970 to 1983 where TerraClimate reads about 20% low. The edge of the region is gradual, there is no gauge check for West Kalimantan, and the cause has not been established.
Counts and longest events are given both for the whole record and outside these periods.