22:null 23:["$","title",null,{"children":"Manuscript figures: formerly, intermittently and water-stressed irrigated US cropland, 1997–2025 | Karthik Ramaswamy | Observable"}] 28:T648,Maps for the manuscript *Intermittent and formerly irrigated lands of the United States* (Lark et al., in preparation), rebuilt on the 1997–2025 irrigation-history masks that every number in the paper uses. Companion notebooks: [SALUS county results](https://observablehq.com/@karthikramaswamy/salus-results) and [SALUS pixel-level maps](https://observablehq.com/@karthikramaswamy/salus-pixel-level-maps-rainfed-bioenergy-sorghum-vs-irrigated-corn-on-formerly-intermittently-and-water-stressed-irrigated-us-cropland). - **Figures 1 and 2** replace the drafts made with LANID 1997–2017 and the 1997–2017 rules (sections 2–3). - **Two new maps:** continuously irrigated land under high water stress, the third extent of Fig. 4 (section 4), and the total effective availability of formerly and intermittently irrigated land (section 5). - **Figure 4 drafts** (SALUS rainfed bioenergy sorghum) are repeated here so all manuscript maps sit in one place (section 6). - **Interactive map** of every layer at 1 km or 3 km (section 7), national totals (section 8) and a check of the masks against the rules (section 9). **Style.** Nonoverlapping 3 km × 3 km blocks (99 × 99 Landsat pixels, 2.97 km), Albers lower-48 projection (EPSG:5070), grey land with white state lines, a yellow-orange-brown ramp, and a histogram legend under each map. Built by notebooks `42_manuscript_figs_1_2` (Figs 1–2 and the new maps) and `41_salus_pixel_maps` (Fig. 4) of the project *Biomass production opportunities* (result set `yanhua`). *Draft for internal review (Karthik Ramaswamy, 7 Oct 2026). Not for circulation.*29:T840,## 1. Definitions Each 30 m pixel's irrigation history comes from the annual LANID irrigation maps, 1997–2025 (29 years). The rules are those of Xie et al. (2021, Table 3) for 1997–2017, with the windows moved to the end of the longer record and the frequency threshold scaled from 21 to 29 years. | Extent | Rule, 1997–2025 (29 years) | Draft rule, 1997–2017 (21 years) | Area | |---|---|---|---| | **Formerly irrigated** | irrigated in at least 3 years before 2023, **not** irrigated in 2023–2025 | at least 3 years before 2015, not irrigated in 2015–2017 | 4.89 Mha | | **Intermittently irrigated** | irrigated at least once in **both** 1997–1999 and 2023–2025, and in fewer than 26 of the 29 years | at least once in both 1997–1999 and 2015–2017, at most 18 of 21 years | 12.59 Mha | | **Continuously irrigated** | irrigated in at least 26 of the 29 years | more than 18 of 21 years | 11.02 Mha | | **… under high water stress** | continuously irrigated land where Aqueduct 4.0 baseline water stress is High (40–80 %) or Extremely High (> 80 %) | – | 7.50 Mha | **Masks.** Formerly and intermittently irrigated land come from Yanhua Xie's 1997–2025 masks; the formerly irrigated mask is held to the strict rule above. Continuously irrigated land comes from LANID directly. Where masks overlap, formerly wins over continuously, which wins over intermittently: 1.40 Mha of the intermittent mask with 26 or more irrigated years counts as continuously irrigated, and 0.11 Mha of the formerly irrigated mask that was irrigated again in 2023–2025 counts as intermittently irrigated. **Why the threshold is 26.** The draft's continuous threshold, 19 of 21 years (90 %), corresponds to 26 of 29 years (90 %) on the longer record. Moving the end window from 2015–2017 to 2023–2025 changes which land qualifies: 2.0 Mha (42 %) of today's formerly irrigated land was last irrigated in 2018–2022 and would not have counted under the draft rule, while some land counted then has been irrigated again since. The draft reported 4.0 Mha formerly and 13.5 Mha intermittently irrigated.2a:T6a8,### 3.1 Reading Figure 2 - **Location (top left).** Intermittently irrigated land is dense in the Mississippi Alluvial Plain, eastern Nebraska, the California Central Valley and the southern High Plains of Texas, where it often covers half or more of the landscape, with a lighter spread across southwest Georgia, Michigan and Indiana, and the Plains. Its colour scale (0–80 %, as in the draft) is shared with the water-stressed map of section 4. - **Frequency (top right).** On average the land is not irrigated in 11.1 of the 29 years (median block about 12). It is irrigated most often — paler colours, 6–10 years without irrigation — in Nebraska, the Central Valley, the Mississippi Delta and Georgia, the cores of irrigated agriculture where intermittency means occasional rain-fed years. It is irrigated least often — darker colours, 20 or more years without irrigation — in Kansas, Oklahoma, the Dakotas, Montana and along the Gulf Coast, where irrigation comes and goes. - **Effective availability (bottom).** Combining the two gives the annual average area not irrigated: 4.83 Mha nationally, 38 % of the intermittently irrigated area. It is concentrated where intermittent land is dense — eastern Nebraska, the Central Valley, the Mississippi Delta and the Texas High Plains — because there even a modest share of years without irrigation adds up to a large area. - **Compared with the draft.** The longer record changes the frequency scale (29 rather than 21 years) and moves land that stopped irrigating after 2017 into Fig. 1, so the two figures should be read together (section 5 adds them up). Section 3.2 explains why the top-right panel looks different from the draft's.2b:T72d,### 3.2 Why the top-right panel differs from the draft The draft's top-right panel has most blocks at 15–20 of 21 years not irrigated and colours much more land than its own top-left panel (for example across Pennsylvania, New York, Ohio and northern Montana). Ours has most blocks at 8–12 of 29 years and colours only intermittently irrigated land. Rebuilding the draft's panel from LANID 1997–2017 (notebook 42, Step 6) explains the difference: - **Not the period or the rule.** The draft's rule applied to LANID 1997–2017 gives 13.5 Mha of intermittently irrigated land — the draft's figure — and a top-left panel like the draft's, but the years-not-irrigated histogram of that land peaks at 6–7 of 21 years, with only 11 % of blocks at 15 or more (panel A below), like ours. - **The draft averaged over all land irrigated at some time.** Averaging over every pixel irrigated at least once in 1997–2017 (C), or every pixel irrigated in 1–18 of the 21 years (D), reproduces the draft's panel: the same coverage, the same pale cores (Nebraska, the Central Valley, the Mississippi Delta) and a histogram rising to 18–19 years, with 45–47 % of blocks at 15 or more. The extra land is mostly sporadically irrigated land outside the intermittent mask. - **Which to use.** Our panel describes the same intermittently irrigated land as the top-left and bottom panels, so the three panels are consistent: block by block, the bottom panel equals the top-left share × (years not irrigated / 29). The draft's panel describes a broader set of land; if the draft's bottom panel was computed from it, it would overstate the availability of intermittently irrigated land. We recommend keeping the rebuilt panel, and adding an all-irrigated-land version only if that is what the panel is meant to show (to confirm with Tyler).2c:T453,## 10. Notes - **Denominator.** Every share is of the whole block (99 × 99 pixels = 882 ha at 3 km), as in the drafts, so blocks over water or outside the lower 48 count their full area. - **Colour caps.** Availability maps use 0–25 % (as in the drafts) and share maps 0–80 % (also as in the drafts), so maps on the same scale compare directly; about 1 % of blocks exceed 25 % in Fig. 1. The histogram bar heights are proportional to the square root of the block count so the long, thin tail stays visible; the pooled last bar holds the blocks above the cap. - **Not limited to SALUS.** These maps count the full masks. The pixel notebook's "% of cell in the extent" layer counts only the land SALUS simulated (94–98 % of each extent), so its values are slightly lower. - **Files.** PNG (300 dpi) and PDF in `outputs/yanhua/figures/` (`fig1_formerly_3km`, `fig2_intermittently_3km`, `fig_ws_continuously_stressed_3km`, `fig_total_available_3km`); 1 km and 3 km multiband GeoTIFFs in `outputs/yanhua/figs12/grids/` for ArcGIS Pro; draft captions in `outputs/yanhua/figures/figs1_2_captions_draft.md`.2d:T662,## 11. References - Xie, Y. & Lark, T. J. (2021). Mapping annual irrigation from Landsat imagery and environmental variables across the conterminous United States. *Remote Sensing of Environment*, 260, 112445. (LANID; Table 3 gives the 1997–2017 rules.) - Xie, Y., Lark, T. J., Brown, J. F. & Gibbs, H. K. (2019). Mapping irrigated cropland extent across the conterminous United States at 30 m resolution using a semi-automatic training approach on Google Earth Engine. *ISPRS Journal of Photogrammetry and Remote Sensing*, 155, 136–149. - LANID 2018–2025 extension: source and citation to be confirmed. - Y. Xie, formerly and intermittently irrigated masks 1997–2025 (unpublished). - Kuzma, S., Bierkens, M. F. P., Lakshman, S., Luo, T., Saccoccia, L., Sutanudjaja, E. H. & Van Beek, R. (2023). *Aqueduct 4.0: Updated decision-relevant global water risk indicators.* Technical Note. World Resources Institute. https://doi.org/10.46830/writn.23.00061 - Boryan, C., Yang, Z., Mueller, R. & Craig, M. (2011). Monitoring US agriculture: the US Department of Agriculture, National Agricultural Statistics Service, Cropland Data Layer Program. *Geocarto International*, 26(5), 341–358. (30 m grid of all rasters.) - US Census Bureau (2018). Cartographic boundary files: states and nation. - Price, L., Tadiello, T. & Basso, B. (2026). SALUS simulations of rainfed bioenergy sorghum and irrigated corn (Fig. 4). Unpublished data, Michigan State University. - Lark, T. J. et al. (in preparation). Intermittent and formerly irrigated lands of the United States. *Check every entry against the published version before submission.*2e:T58a,const FILES = { "f12_1km_continuously_stressed_share_pct.png": FileAttachment("f12_1km_continuously_stressed_share_pct.png"), "f12_1km_formerly_share_pct.png": FileAttachment("f12_1km_formerly_share_pct.png"), "f12_1km_intermittently_effective_pct.png": FileAttachment("f12_1km_intermittently_effective_pct.png"), "f12_1km_intermittently_share_pct.png": FileAttachment("f12_1km_intermittently_share_pct.png"), "f12_1km_intermittently_years_not_irrigated.png": FileAttachment("f12_1km_intermittently_years_not_irrigated.png"), "f12_1km_total_effective_pct.png": FileAttachment("f12_1km_total_effective_pct.png"), "f12_3km_continuously_stressed_share_pct.png": FileAttachment("f12_3km_continuously_stressed_share_pct.png"), "f12_3km_formerly_share_pct.png": FileAttachment("f12_3km_formerly_share_pct.png"), "f12_3km_intermittently_effective_pct.png": FileAttachment("f12_3km_intermittently_effective_pct.png"), "f12_3km_intermittently_share_pct.png": FileAttachment("f12_3km_intermittently_share_pct.png"), "f12_3km_intermittently_years_not_irrigated.png": FileAttachment("f12_3km_intermittently_years_not_irrigated.png"), "f12_3km_total_effective_pct.png": FileAttachment("f12_3km_total_effective_pct.png") }; display(html`

Files. 12 PNG grids attached to this notebook (FileAttachment needs literal names, so they are listed here).

`);2f:T17ee,const LAYERS = [ {key: "formerly_share_pct", label: "Fig. 1 — formerly irrigated land (% of block, all available)", short: "Formerly irrigated", cap: 25, unit: "%"}, {key: "intermittently_share_pct", label: "Fig. 2 top left — intermittently irrigated land (% of block)", short: "Intermittently irrigated", cap: 80, unit: "%"}, {key: "intermittently_years_not_irrigated", label: "Fig. 2 top right — mean years not irrigated, 1997–2025 (of 29)", short: "Years not irrigated", cap: 29, unit: "years", shareOf: "intermittently_share_pct"}, {key: "intermittently_effective_pct", label: "Fig. 2 bottom — intermittent effective availability (% of block)", short: "Available (intermittent)", cap: 25, unit: "%", shareOf: "intermittently_share_pct"}, {key: "continuously_stressed_share_pct", label: "Water-stressed continuously irrigated land (% of block)", short: "Continuously irrigated, water-stressed", cap: 80, unit: "%"}, {key: "total_effective_pct", label: "Total effective availability, formerly + intermittently (% of block)", short: "Available (total)", cap: 25, unit: "%"} ]; const RAMP = (t) => d3.interpolateYlOrBr(0.06 + 0.94 * t); const colorFor = (L) => d3.scaleSequential([0, L.cap], RAMP).clamp(true); const decodeValue = (key, k) => { const m = f12.variables[key]; return m.lo + (k - 1) / 254 * (m.hi - m.lo); }; const fmtValue = (L, x) => (L.unit === "years" ? `${x.toFixed(1)} years` : `${x.toFixed(1)} %`); const legendFor = (L) => Plot.legend({color: {type: "linear", domain: [0, L.cap], interpolate: RAMP}, label: `${L.short} (${L.unit === "years" ? "years" : "% of block"}; ≥ ${L.cap} in the end colour)`, width: 360}); const gridCache = new Map(); function loadGrid(name) { if (!gridCache.has(name)) gridCache.set(name, (async () => { const img = await FILES[name].image(); const c = document.createElement("canvas"); c.width = img.naturalWidth; c.height = img.naturalHeight; const ctx = c.getContext("2d", {willReadFrequently: true}); ctx.drawImage(img, 0, 0); const rgba = ctx.getImageData(0, 0, c.width, c.height).data; const v = new Uint8Array(c.width * c.height); for (let i = 0; i < v.length; i++) v[i] = rgba[4 * i]; return {w: c.width, h: c.height, v}; })()); return gridCache.get(name); } function colorize(g, colorOf) { const lut = new Uint8ClampedArray(256 * 4); for (let i = 1; i < 256; i++) { const c = d3.rgb(colorOf(i)); lut.set([c.r, c.g, c.b, 255], 4 * i); } const c = document.createElement("canvas"); c.width = g.w; c.height = g.h; const ctx = c.getContext("2d"); const im = ctx.createImageData(g.w, g.h), d = im.data; for (let i = 0; i < g.v.length; i++) { const k = g.v[i]; if (k) { const o = 4 * i, l = 4 * k; d[o] = lut[l]; d[o + 1] = lut[l + 1]; d[o + 2] = lut[l + 2]; d[o + 3] = 255; } } ctx.putImageData(im, 0, 0); return c; } async function rasterMap({file, shareFile = null, geo, colorOf, valueText, legend, title, fitTo, aspect = 0.62, width: w}) { const g = await loadGrid(file); const s = shareFile ? await loadGrid(shareFile) : null; const img = colorize(g, colorOf); const h = Math.round(w * aspect); const dpr = window.devicePixelRatio || 1; const canvas = html`