Physical & climate risk • contested inputs
Water as a contested input
Related surfaces: global hazard and the ENSO index · gas supply and LNG · power-market structure
Water is not a resource on this page, it is a claim. The same cubic metre behind a dam wall can be electricity or it can be a rice crop, and somebody has to decide which. Thailand publishes the numbers that let you watch that decision being made: daily storage against capacity, for every large dam, split by the agency that operates it[1]. Read the national figure and the country is half full. Read it per agency and the power system's reservoirs are 62.8% full while the irrigation department's are 45.9%. That 16.9-point gap is not weather. It is two claimants on one input, and one of them keeps the lights on.
The rest follows from that. A developing El Niño points at both less hydro and more cooling load, gas is the substitution that absorbs the difference at 61% of Thai gas demand already going to power[5], and the gas build-out has no successor yet: the last wave finished in January 2025, EGAT's next gas blocks await the new power development plan, and what is being built now is solar at small-producer scale[4]. The last section asks the question the rest of the page depends on: who can actually see the water, and what happened when one of the watchers switched its front door off.
How to read this page: ● measured sourced data · ◐ inferred analyst reading, basis linked · ○ projected anchored to a real starting point. Bracketed citations link to the sources at the foot of the page.
The contest
Two claimants on the same water
Thailand's large dams are operated by two bodies with different jobs. EGAT, the state generator, holds the hydro reservoirs; the Royal Irrigation Department holds the agricultural ones. Bar length below is the size of the claim, capacity in million cubic metres, and the filled part is how much water is actually behind it. EGAT holds nearly four times as much capacity as the irrigation department and is holding it fuller[1].
National Hydroinformatics Data Center feed as at 2026-08-31, observed dates 2026-08-30 and 2026-08-31. The two agencies report a day apart and seven reservoirs are reported by both under different spellings; both are handled in the fetcher and documented in the snapshot's known_traps.
◐ inferred The gap is a scheduling decision, not a rainfall pattern. Hydro reservoirs and irrigation reservoirs sit in different basins and serve different demands, so they never read the same; but a power system that is short of cheap gas has a reason to hold water, and an irrigation system under the same dry season has a reason to release it. The basis is in the numbers above: same country, same month, 16.9 points apart, with the fuller set belonging to the claimant that can substitute its way out of a shortfall by burning something.
No published Thai national reservoir total was located to check these aggregates against. The percentages here are this site's arithmetic on the NHC feed, not a figure Thailand publishes.
Emptiest first
Every large dam, ordered by how full it is
The national percentage hides the thing that matters, which is that the distribution has a bottom. Eight reservoirs sit under a quarter full. A reservoir at a tenth of capacity is not a low reading, it is a reservoir that cannot be drawn on much further, and the colour bands say which of them are there[1].
The named lows: Mun Bon at 13%, Lam Nang Rong at 17.1%, Lam Phra Ploeng at 17.4%, Kra Siew at 18.5%, UBOLRATANA DAM at 19%. Ubolratana is the one to watch beyond its own basin: it is a large EGAT reservoir in the Northeast, so its storage is generation capacity that either exists or has to be replaced from somewhere else[1].
Circle size is declared capacity, colour is fullness on the same bands as the ladder. 41 dams, all with coordinates from the source feed[1].
The season that decides it
A developing El Niño points both ways
The Pacific is warming. The Oceanic Niño Index reads +1.39°C for MJJ 2026, with 2 consecutive seasons above the +0.5°C threshold. CPC requires five before it declares an episode, so this is a developing El Niño and not a declared one, and that distinction is the difference between a forecast and a fact[2,3]. The index itself, back to 1950, with the one claim on this site that depends on it, lives on physical risk.
What it would do to the price
○ projected A warm Pacific has historically come with milder Northeast Asian winters, and Japan and Korea are the buyers Thailand competes with for spot LNG. Softer demand there would ease the bill here. The mechanism, the hindcast behind it and its registered forward test sit with the fuel account, which is where that claim is scored, and it is one claim about one region rather than a rule.
What it would do to the volume
○ projected The same state means a hotter, drier Thailand: more cooling load, less inflow into the reservoirs above. Anchored to two measured starting points, the 59.3% national storage on this page and the 64% of Thai gas demand that went to power in 2026-06[1,5]. This site could not obtain published quantities for the Thai cooling-load and hydro effects from a source it could read, so the direction is stated and the size is not.
So the same climate state cuts the price of the substitute and raises the need for it. Which dominates is open, and a page that answered it would be pretending.
Last 24 hours of rain, 4,452 stations across 8 agencies
The reservoir numbers move with this. A median of 2 mm with 29.1% of stations dry is the shape of the day, not of the season[1].
The commitment test
What Thailand is actually building
Plans swing. The stage where money has demonstrably moved is construction, because concrete does not get poured on a press release, and GEM publishes lifecycle status for every tracked plant[4]. Read Thailand that way and GEM's files answer: 3,555 MW under construction, every megawatt of it gas. Checked against the operators, even that has finished: RATCH's Hin Kong units entered commercial operation on 1 March 2024 and 1 January 2025, Gulf's four Rayong units between March 2023 and October 2024, and GEM's own wiki lists the B.Grimm Ratchaburi station as operating, so the construction column is a snapshot lag. No new gas or coal plant is under construction; what continues is solar at small-producer scale, roughly 600 MW installed per year by Gulf alone, which the tracker also misses.
GEM oil-and-gas-plant, coal-plant, solar, wind and hydropower trackers, retrieved 2026-08-31, CC BY 4.0. Proposed groups announced, pre-construction and permitted. The solar and wind trackers carry utility-scale size thresholds, so zero solar under construction means zero utility-scale projects above GEM's threshold, not zero solar investment[4].
Against the plan
● measured The draft PDP 2024 targets 51% of Thailand's total electricity generation capacity from renewables by 2037, against a previous goal of 36%.[11]
It was a draft when the source this page could read described it, in November 2024. No dated confirmation of approval since then could be pinned to a readable source, so this page says draft and dates the claim rather than asserting a current status.
And against the dispatch
● measured In October 2025 the National Energy Policy Council suspended four power stations for oversupply: North Bangkok, South Bangkok and Nam Phong, which are gas and together about 4 GW, plus Mae Moh units 8 and 9, which are coal. A further 0.6 GW was deferred to 2029, and the suspended plants are to resume after that.[12,14,13]
Seven privately owned Thai gas plants, over 11 GW between them, ran below a 10% capacity factor in 2025 and have cost EGAT and ratepayers THB 159 billion since 2023, because a contracted plant that does not generate still collects availability payments.
◐ inferred Thailand is long gas capacity it does not always run, short of cheap gas to run it with, and building more of it anyway. 36,738 MW of gas operates, 2,072 MW is mothballed, 4 GW was suspended for oversupply, and 3,555 MW is under construction[4,12]. The reading is that the binding constraint is fuel cost rather than iron in the ground, which is also why a dry season shows up as an import bill.
● measured The coal risk here is dispatch, not construction, and the fleet numbers say so plainly: Thailand has 10,726 MW of coal cancelled or shelved, more than the 6,138 MW it operates, nothing under construction, and 600 MW permitted[4]. Under price pressure the cheap move is running the plants that already exist harder. Anyone writing that Thailand is building new coal has the wrong mechanism.
The pattern is not only Thai. Worldwide, hydropower has 765,124 MW proposed against 193,808 MW under construction, roughly four times as much water power claimed as built[4].
The measurement layer
Who can actually see the water
Everything above rests on one national feed, so the honest question is what could check it. Six routes, with what each one measures and whether the reservoirs this page names were actually found in it. All checked on the dates shown.
| Route | Status | Measures | Resolution | The reservoirs here |
|---|---|---|---|---|
| NHC / thaiwater.net[1] National Hydroinformatics Data Center, Thailand | live 2026-07-30 | storage volume against declared capacity | daily, per dam, per agency | 41 large dams with a declared capacity, both operating agencies |
| NASA GSFC Global Water Monitor[7] NASA Goddard Space Flight Center | live 2026-07-30 | water surface level (altimetry), plus lake extent | 10-day near-real-time, and monthly, back to 1992 | Bhumibol, Sirikit, Ubolratana and Pa Sak Chonlasit in the monthly set; Lam Pao in the 10-day set; Bhumibol, Sirikit and Lam Pao also in the extent set |
| USDA FAS IPAD (G-REALM)[6,7] USDA Foreign Agricultural Service | retired 2026-07-30 | reservoir level (altimetry) | was 10-day and monthly, 1992 onward | Thailand was a named country of interest |
| DAHITI[8] DGFI, Technical University of Munich | unchecked 2026-07-30 | level, surface area, volume variation | varies by target | not checked |
| Theia Hydroweb[9] LEGOS / CNES, distributed by Theia (Data Terra) | live 2026-07-30 | lake and river water level and storage | continuous long-duration series | not checked |
| Copernicus water bodies[10] Copernicus Land Monitoring Service | live 2026-07-30 | lake water level and surface temperature | 10-day, since 2016 | not checked |
Altimetry measures the height of a water surface. It does not measure storage against capacity, and it cannot see which agency holds the water or what it is being held for. It is the independent check on a national feed, not a replacement for one.
The retirement that was not a loss
● measured The USDA Foreign Agricultural Service's reservoir page, the front end for a global satellite-altimetry record of reservoir levels running back to 1992, now answers HTTP 503 and serves a page titled "IPAD retired"[6]. That looks like a thirty-year public record going dark. It is not. The retired page's own script redirects to NASA Goddard, where the product is live: 518 targets in the 10-day set dated two days before this page was built, and the archive published as a single bulk download whose Lam Pao series runs from 30 September 1992 to 19 July 2026 across six satellite missions[7].
It matters for this page specifically, because the reservoirs it names are in there: Bhumibol, Sirikit, Ubolratana and Pa Sak Chonlasit in the monthly set, Lam Pao in the 10-day set[7]. Thailand's own feed can be checked against a satellite for the exact dams the argument rests on.
Recorded because the near-miss is the reusable part: a 503 and a retirement notice looked like evidence of absence, and the data was one redirect away. A dead front door is not a dead dataset.
◐ inferred The asymmetry worth noticing: the country most exposed here publishes better water data than most rich ones. Daily storage against capacity, per dam, per operating agency, free and without a key. The satellites can see a level; only the agency can tell you whose water it is.
Sources and method (14)
- [1] National Hydroinformatics Data Center (NHC), Thailand: thaiwater.net dam and 24-hour rainfall APIs. Daily storage against capacity for every large dam, per operating agency, public and keyless. Thai government open data.
- [2] NOAA Climate Prediction Center, Oceanic Nino Index (ONI), the 3-month running mean of ERSST.v5 sea-surface-temperature anomalies in the Nino 3.4 region, from DJF 1950. US Government work, public domain.
- [3] NOAA CPC, ONI definition and episode rule: an episode requires the index to hold past ±0.5°C for a minimum of five consecutive overlapping seasons.
- [4] Global Energy Monitor, five asset-level power trackers (oil and gas plant, coal plant, solar, wind, hydropower), retrieved 30 July 2026 via the files behind GEM's own public tracker maps, CC BY 4.0. Per-tracker URLs, row counts, status vocabularies and unit checks are recorded in src/data/power/gem_power_pipeline.json.
- [5] EPPO, Energy Policy and Planning Office, Ministry of Energy, Thailand: table T03_02_02, Consumption of Natural Gas by Sector, monthly, in mmscfd.
- [6] USDA Foreign Agricultural Service, International Production Assessment Division (IPAD): the front end that hosted the G-REALM reservoir product is retired. Checked 30 July 2026, /cropexplorer/global_reservoir/ answers HTTP 503 and serves a page titled 'IPAD retired — USDA FAS production site', whose script redirects to NASA GSFC.
- [7] NASA Goddard Space Flight Center, Global Water Monitor, lake and reservoir water level from satellite radar altimetry (Topex/Poseidon, Jason-1/2/3, Sentinel-3/6, ENVISAT, SARAL, SWOT). Checked 30 July 2026: 518 targets in the 10-day near-real-time set dated 2026-07-28, 450 in the monthly set and 439 in the lake-extent set dated 2026-07-29. The whole 10-day archive is published as one 324 MB bulk download of 1,040 files; the Lam Pao series in it runs from 30 September 1992 to 19 July 2026 across six missions.
- [8] DAHITI, Database for Hydrological Time Series of Inland Waters, DGFI-TUM: altimetric water level, surface area and volume variation, free after registration. Reachable 30 July 2026; Thai reservoir coverage not checked, because the catalogue search is JavaScript-driven and returns the same shell for any query.
- [9] Theia / Data Terra, Hydroweb operational altimetry lake water level and storage: continuous long-duration series for lakes over 100 km², reservoirs and the largest rivers, produced by LEGOS and distributed through the Theia catalogue.
- [10] Copernicus Land Monitoring Service, water bodies portfolio: lake water level and surface temperature from Sentinel-3, global, since 2016.
- [11] Watson Farley & Williams, 'Thailand Powers Up: New Renewable Energy Incentives and Opportunities in 2024', 5 November 2024: the draft PDP 2024 targets 51% of total electricity generation capacity from renewables by 2037, against a previous goal of 36%, and is described as a draft.
- [12] Transformer Magazine, 'Thailand pauses 4 GW of gas capacity', 27 February 2026: the National Energy Policy Council suspended four power stations in October, three of them gas-fired and together 4 GW, and deferred a further 0.6 GW plant to 2029, with the suspended plants to resume after that date.
- [13] IEEFA, Christopher Doleman and Sam Reynolds, 'Thailand's gas conundrum: Overbuilt, underutilized, and increasingly expensive', 26 February 2026: seven privately owned Thai gas plants ran below a 10% capacity factor in 2025, and those seven, over 11 GW, have cost EGAT and ratepayers THB 159 billion since 2023.
- [14] Bangkok Post, 'Weak demand results in suspension of 4 power plants', 29 October 2025: the four suspended plants named as North Bangkok, South Bangkok and Nam Phong (gas) plus Mae Moh units 8 and 9 (coal). Paywalled, cited not reproduced.
Reservoir and rainfall figures are the NHC feed as at 2026-08-31, accumulated one observation per run by scripts/fetch_thaiwater.js; the feed is a snapshot with no history, so the series starts at the first run and cannot be backfilled. Capacity by lifecycle status is a 2026-07-30 pull of five GEM trackers by scripts/fetch_gem_power.js, which records each tracker's own URL, row counts, status vocabulary and unit checks. Gas demand shares are computed from EPPO's monthly sector table, latest month 2026-06. ENSO figures are NOAA CPC's published index with episodes recomputed from CPC's own five-season rule on every refresh. Monitoring-route statuses were each checked on the date shown in the table.