Abstract
Indonesia is the world's largest producer of palm oil (59 % of global supply) and the second-largest producer of coconut. However, the spatial dynamics and land use transition between these two strategic oil crops from the Arecaceae family remain underexplored at the provincial level over an extended time period. This study aims to analyze the spatio-temporal dynamics of oil palm and coconut across 38 Indonesian provinces from 2003 to 2024, identify spatial clusters and production hotspots using LISA and SKATER algorithms, examine the correlation and productivity differences between the two commodities, and assess regional productivity variation across Indonesian islands. Data were obtained from BPS-Statistics Indonesia's annual publication “Indonesian Permanent Estate Crops Statistics” (2003–2024). Spatial analysis was conducted using GeoDa software (version 1.22.0.21) with Queen contiguity weights, LISA, the SKATER algorithm, bivariate Moran's I, and differential Moran's I. Statistical analysis was performed using jamovi (version 2.6), including Pearson correlation, a paired samples
t
-test with Cohen's d effect size, and a one-way Welch's ANOVA. Compound Annual Growth Rates (CAGR) were calculated with R
2
from linear trend lines. The oil palm area expanded from 5.28 million ha in 2003 to 16.01 million ha in 2024 (CAGR = 5.42 %,
R
2
= 0.99), while the coconut area declined from 3.91 million ha to 3.29 million ha (CAGR = −0.82 %,
R
2
= 0.92). The area ratio increased from 1.35 to 4.87 (+260 %), and the production ratio increased from 3.21 to 16.32 (+408 %). LISA analysis identified three significant high-high clusters (hotspots) in Sumatra and Kalimantan, and no high-low outliers. The SKATER algorithm (
k
= 5) classified 32 provinces (84.2 %) as not clustered, indicating weak spatial autocorrelation. Pearson correlation between CPO and copra production was positive but weak (
r
= 0.327,
p
= 0.045). Paired samples
t
-test showed oil palm productivity (
M
= 2.87 ton/ha, SD = 0.883) was significantly higher than coconut productivity (
M
= 0.98 ton/ha, SD = 0.216), with a very large effect size (Cohen's d = 2.01, 95 % CI [1.33, 2.68],
t
(25) = 10.3,
p
r
= −0.130 (
p
= 0.526). One-way Welch's ANOVA (excluding Java due to
n
= 2, noting that West Java and Banten have small-scale production) revealed no significant productivity differences across islands (F(3, 3.84) = 1.51,
p
= 0.346,
η
2
= 0.24). The bivariate Moran's I showed no significant spatial relationship between CPO and surrounding copra production (
I
= 0.029,
p
= 0.314). Differential Moran's I for CPO changes was non-significant (
I
= 0.098,
p
= 0.101), while copra changes were significant (
I
= 0.394,
p
= 0.007), indicating spatial clustering of copra production changes. Indonesia has experienced a marked land use transition from coconut to oil palm over 21 years. The substantial productivity advantage of oil palm (Cohen's
d
= 2.01, 2.9:1 ratio) provides a strong economic incentive for this transition. The weak spatial clustering, though with detectable hotspots in Sumatra and Kalimantan, suggests that provincial-level factors remain highly influential. For coconut to remain competitive and for oil palm to develop sustainably, integrated policies supporting sustainable oil palm production while revitalizing the coconut sector are essential. Key interventions include developing high-yielding hybrid varieties (target > 3 tons/ha), replanting senile stands, intercropping systems, and value-added processing.