The research, led by Wen Zhang, a doctoral student in the University of Arizona’s School of Natural Resources and the Environment, drew on more than 40 years of satellite observations to track changes in the vegetation leaf area index, a measure closely linked to vegetation activity and productivity. Leaf area index quantifies the amount of leaf surface per unit of ground area, making it one of the most direct remotely sensed indicators of how much photosynthetic machinery an ecosystem is deploying at any given time. By examining how this index fluctuated across four decades, the team could distinguish the long-term greening trend from the shorter-term swings superimposed on it. What they found was that the extremes are diverging: the upper peaks of vegetation activity during wet years and the lower troughs during dry years are moving farther and farther apart as time goes by.

“The upper and lower extremes are getting farther and farther apart as time goes by,” Zhang said. “Vegetation activity is increasing during wet years, but dry years are hitting plants harder. It’s a bit like the nursery rhyme about the little girl with the curl: When it’s good, it’s very good, but when it’s bad, it’s awful.” The metaphor captures a phenomenon that ecologists describe as a boom-and-bust dynamic, in which the amplitude of ecosystem variability grows even as the average trajectory appears healthy. In practical terms, a dryland that greening statistics suggest is thriving may in fact be swinging between states of lush productivity and stress with a frequency and intensity that earlier decades never showed.

The most likely driver of this pattern, Zhang explained, is the combination of rising atmospheric carbon dioxide with natural rainfall variability, although she cautioned that more data is needed to pin down the precise mechanisms. There is evidence that under elevated CO2 concentrations, plants can use water more efficiently, because higher CO2 levels allow them to photosynthesize while keeping their stomata, the microscopic pores on leaf surfaces, partially closed. This improved water-use efficiency reduces water loss and enables plants to grow more leaves, particularly in water-limited environments where moisture is the primary constraint on growth. The result is the well-documented CO2 fertilization effect that underlies the dryland greening trend observed from space.

But the same physiological advantage carries a hidden cost. “Larger vegetation requires more resources to maintain, so when a moderate drought hits the following year, these larger plant structures need more resources than are available, which leaves them far more sensitive and vulnerable,” Zhang said. In other words, the extra leaf area that CO2 fertilization produces during favorable years becomes a liability when water is scarce. Bigger canopies demand more transpiration to stay cool and more carbohydrates to maintain, and when a drought arrives, the oversized vegetation experiences proportionally greater stress than it would have in a lower-CO2 world. This mechanism can transform an ordinary dry year into a disproportionately severe bust, amplifying the natural oscillation of dryland ecosystems rather than damping it.

The consequences of this growing volatility extend well beyond ecology into the economics of agriculture and livestock production. In rain-fed farming regions such as the American Southwest, where crops depend directly on precipitation rather than irrigation, higher year-to-year variability may force a heavier reliance on artificial irrigation simply to maintain consistent productivity. Pasture and rangeland forage production, which follows the same boom-and-bust rhythm as natural vegetation, will likewise become harder to predict. For ranchers who must decide each season how many animals their land can support, that unpredictability is not an abstract concern but a direct threat to planning and livelihoods.

“Higher variability in forage production presents a significant challenge for rangeland managers,” said Bill Smith, senior author of the study and an associate professor specializing in land, water and climate change geospatial analysis in the School of Natural Resources and the Environment. “Ranchers depend on stable forage production so they can accurately plan out their land needs each growing season. Less predictable forage production can thus disrupt their plans with potential detrimental consequences to livelihoods.” In regions where stocking decisions must be made months in advance of the growing season, a single bust year that follows an unusually productive boom can leave managers with herds that their pastures cannot sustain, forcing costly destocking or supplemental feeding.

Beyond its immediate agricultural implications, the intensifying flicker in dryland productivity may be an early warning of deeper ecological change. David Moore, a study co-author and professor in the School of Natural Resources and the Environment who chairs the watershed management and ecohydrology program, pointed to a pattern observed across many ecological systems. “If you look at lots of different ecological systems, their productivity tends to flicker on and off right before a big change happened. It’s a sign that they’re under stress and losing their resilience. It’s possible that’s what’s happening with drylands,” he said. This idea, sometimes discussed in the scientific literature as a critical slowing down or flickering signal preceding regime shifts, suggests that the growing variance in dryland vegetation could foreshadow a transition to a fundamentally different ecosystem state, though predicting the ultimate outcome of such flickering remains difficult.

Part of that difficulty lies in the limitations of the tools scientists use to project the future. The study evaluated 13 of the leading global vegetation models and found that none of them captured the observed increase in year-to-year variability. “The models assume drylands are still stable and that plants will respond to changes in atmospheric carbon dioxide and rainfall in predictable ways,” Zhang said. “They fail to account for how plant responses are fundamentally changing over time.” In effect, the models reproduce the greening trend but not the instability that accompanies it, presenting a smoothed and overly optimistic picture of dryland behavior. Because these models feed into the Earth system models used for climate projections, the blind spot propagates upward into forecasts of carbon storage, water resources and food production.

“If Earth system models are not correctly capturing the sensitivity of dryland plants to climate change, then all bets are off when making projections 50 years into the future,” Smith said. “We hope this paper inspires new research focused on a better understanding and representation of drylands in the Earth system.” The message of the study is ultimately one of recalibration: the greening of the world’s drylands, often cited as evidence that rising CO2 is boosting global vegetation, conceals a loss of stability that satellites can now measure and that models must learn to represent. For the two billion people who depend on these landscapes, the difference between a stable green trend and an escalating boom-and-bust cycle is the difference between predictable harvests and a future in which every growing season is a gamble.

Subject of Research: Rising CO2-driven boom-and-bust vegetation instability in global dryland ecosystems

Article Title: Satellite data exposes escalating 'boom-and-bust' dynamic in greening drylands

Article References: Satellite data exposes escalating 'boom-and-bust' dynamic in greening drylands. (n.d.). Original publication

Image Credits: AI Generated

DOI: Not provided

Keywords: drylands, satellite data, leaf area index, CO2 fertilization, vegetation models, climate change, ecosystem stability, rangeland management, rain-fed agriculture, Nature Climate Change, University of Arizona, boom-and-bust dynamics

The research team, led by scientists working at the National Precision Agriculture Research Center in Xiaotangshan, Beijing, ran field experiments across three consecutive growing seasons from 2017 to 2019. They varied potato cultivars, planting densities, nitrogen rates, and, in some years, irrigation and potassium treatments across randomized plots. Using a handheld spectrometer positioned one meter above the canopy under clear-sky conditions, they recorded hyperspectral reflectance from 400 to 1000 nanometers at five growth stages per season, capturing everything from sparse early vegetation to full canopy closure. In total they assembled 690 field spectra paired with equally many destructive LAI measurements, spanning values from 0.15 to 6.52 square meters of leaf per square meter of ground, a range broad enough to stress-test any estimation model.

The physical principle behind the approach is well understood. In the visible part of the spectrum, between 400 and 700 nanometers, chlorophyll and other leaf pigments absorb incoming sunlight to drive photosynthesis, producing characteristic dips in reflectance. In the near-infrared band, from roughly 760 to 1000 nanometers, the spongy internal structure of leaves scatters photons, creating a high reflectance plateau. As a canopy thickens and LAI rises, this contrast sharpens. The trouble is that classic vegetation indices such as NDVI compress this rich information into a handful of bands, and their correlation with LAI, which peaked at only about 0.54 in this study for the simple ratio index, saturates once LAI climbs above 2. Beyond that threshold, additional leaves are effectively invisible to the index, causing systematic underestimation precisely when vigorous growth matters most for yield forecasting and nitrogen management.

To build a smarter estimator, the researchers first turned to the PROSAIL radiative transfer model, which couples a leaf-level optical model called PROSPECT-5 with the SAIL canopy reflectance model. By sweeping combinations of chlorophyll content, leaf structure, leaf angle, solar geometry, and LAI itself, they generated an enormous synthetic dataset of 78,780 simulated spectral curves spanning 400 to 2500 nanometers. A comparison showed that the simulated spectral envelope fully enclosed the variability of the actual field measurements, confirming that the simulations captured the real-world heterogeneity induced by different cultivars, growth stages, and management practices. Crucially, the inputs were resampled to an 8-nanometer interval between 450 and 962 nanometers to strip redundancy and keep computation manageable.

The centerpiece of the study is a hybrid neural network that pairs a convolutional neural network, or CNN, with a long short-term memory network, or LSTM. The CNN side uses three stacked one-dimensional convolutional blocks with 16, 32, and 64 filters, each followed by batch normalization, ReLU activation, and max pooling, to pull out local morphological features in the spectrum, such as absorption valleys and reflectance peaks. The LSTM side, given the funnelled output of the CNN, treats the spectrum as a continuous sequence ordered by wavelength and uses its internal memory gates to model long-range dependencies, effectively connecting how pigment absorption in the visible region relates to scattering in the near-infrared plateau. The authors argue this architecture fits the physics better than attention mechanisms, which are data-hungry, or plain multilayer perceptrons, which treat each wavelength as an isolated node and discard spectral continuity.

The deep features that emerged were demonstrably more sensitive to LAI than anything the raw spectrum could offer. When the researchers correlated activation values from each network’s fully connected layer against measured LAI, LSTM-only features topped out at a correlation of 0.49, CNN features reached 0.78, but CNN-LSTM features repeatedly exceeded 0.80, peaking at 0.82. Even more striking, gradient-based activation maps revealed that the hybrid network concentrated its attention precisely on the red-edge transition region between 700 and 750 nanometers and on segments of the near-infrared plateau, the exact wavelength zones that plant physiology predicts should carry the strongest canopy-structure signal. In other words, the network autonomously rediscovered the biophysics it was never explicitly taught.

Raw deep learning, however, still stumbled on a chronic agricultural problem: too few ground-truth samples, unevenly distributed toward medium LAI values. Deep networks trained from scratch on field data beat traditional machine learning benchmarks, with the CNN-LSTM reaching a validation coefficient of determination of 0.74 and a root mean square error of 0.55 square meters per square meter, but random forest and XGBoost models tuned to spectral features collapsed from training accuracies near 0.88 to validation values of just 0.38 and 0.44, a textbook overfitting pattern. The decisive move was transfer learning. The researchers first pre-trained the CNN-LSTM network on the 78,780 PROSAIL simulations, where it achieved near-perfect internalization of radiative transfer physics, with validation R-squared of 0.995. Then they froze the convolutional feature-extraction layers and fine-tuned only the fully connected layers using the limited field measurements, transplanting physical knowledge from simulation into the messy reality of the field.

The payoff was substantial. Applied directly to field data without fine-tuning, the pre-trained model managed only a validation R-squared of 0.53 and badly underestimated dense canopies. With transfer learning, validation accuracy rose to 0.80 with RMSE of 0.50, a 41 percent accuracy improvement over the non-transfer model in the high-LAI range, where mean absolute error fell to 0.53 compared with 0.90 for the direct-transfer version and 0.79 for the traditional partial least squares regression baseline. The systematic bias that plagued simpler methods, overestimating sparse plots and underestimating closed canopies, largely vanished, with predictions converging tightly around the one-to-one line. Across individual years the framework held R-squared values of 0.77 in 2017 and 0.81 in 2018, dipping to 0.56 in 2019, a decline the authors traced to extreme nitrogen treatments that pushed canopies beyond the simulated parameter space, such as chlorophyll accumulation exceeding the pre-training limits.

The study is candid about its constraints. The PROSAIL parameter space was fixed by local measurements, so applying the framework elsewhere demands recalibration and at least a small set of in-situ samples for fine-tuning. Training on a single NVIDIA RTX 5070 GPU with 64 gigabytes of memory took about 65 minutes per pre-training session, a manageable cost but not a trivial one, and validation remains limited to a single site over three years. Still, the implications reach beyond potatoes. The work demonstrates a practical recipe for fusing physical simulation with deep learning: let a radiative transfer model teach a hybrid CNN-LSTM the general laws of how light interacts with leaves, then let a few dozen field measurements adapt those laws to one farm, one season, one cultivar. For agronomists, that points toward rapid, non-destructive canopy maps that can guide variable-rate fertilization and irrigation with a fraction of the fieldwork, and for the broader field of quantitative remote sensing, it offers a template for bringing physically grounded AI to any crop whose light signature tells the story of its growth.

Subject of Research: Estimating potato leaf area index from canopy hyperspectral data using a hybrid CNN-LSTM deep learning model with simulation-to-reality transfer learning

Article Title: Estimation of potato leaf area index using a hybrid CNN-LSTM architecture with transfer learning

Article References: Liu, Y., Yue, J., Fan, Y., Feng, Z., Zhang, H., Guo, W., Qiao, H., Yang, F., Liu, H., Zhou, W., & Feng, H. (2026). Estimation of potato leaf area index using a hybrid CNN-LSTM architecture with transfer learning. Artificial Intelligence in Agriculture. https://doi.org/10.1016/j.aiia.2026.09.003

Image Credits: AI Generated

DOI: 10.1016/j.aiia.2026.09.003

Keywords: potato, leaf area index, hyperspectral remote sensing, CNN-LSTM, transfer learning, PROSAIL, radiative transfer model, precision agriculture, deep learning, vegetation indices, crop monitoring, spectral saturation

The research team, led by Kisman Bhattarai and Nithya Rajan of Texas A&M’s Department of Soil and Crop Sciences, conducted field experiments at the university’s research farm near College Station in 2021 and 2022. The site sits on Weswood silty clay loam in a humid subtropical climate, and the region’s rivers ultimately drain into the Gulf of Mexico, where nutrient-driven hypoxia has become a serious ecological concern. Maize in central Texas is grown primarily as a rainfed crop, making the interaction between fertilizer timing and rainfall a matter of both economic and environmental consequence.

The experimental design was thorough. The researchers tested eight different split nitrogen applications in 2021, ranging from applying all nitrogen at planting (a 100:0 split) to dividing it evenly between planting and the R1 silking stage (a 50:50 split). The full fertilizer rate was 240 kilograms of nitrogen per hectare, based on the highest recommended rate for corn in the region. In 2022, they added two reduced-rate treatments—50 percent and 25 percent of the full rate—to see whether cutting nitrogen inputs could maintain yields under stress. All treatments were replicated four times in a randomized complete block design, and the team measured everything from plant height and leaf area index to chlorophyll content, biomass, grain yield, plant nitrogen uptake, and residual soil nitrate at multiple depths down to 60 centimeters.

The two growing seasons could not have been more different, and that contrast turned out to be the study’s greatest asset. In 2021, growing-season precipitation totaled 572 millimeters, above the 20-year average of 508 millimeters, and timely rains coincided with the reproductive stages. In 2022, by stark contrast, only 100 millimeters of rain fell during the growing season—just 20 percent of the long-term average—accompanied by air temperatures one to three degrees Celsius higher than the previous year. The 2022 season was classified as drought-prone, and the results reflected it: average grain yield in fertilized plots collapsed from 11.67 megagrams per hectare in 2021 to 4.09 megagrams per hectare in 2022, a 63 percent decline.

Under the favorable conditions of 2021, the timing of nitrogen application had surprisingly little effect on final outcomes. Split applications did not change aboveground biomass or grain yield, although delaying 50 percent of the nitrogen until the R1 stage temporarily reduced leaf area index and chlorophyll content. Crucially, the plants recovered. Once the second fertilizer dose was applied, all split treatments showed rapid nitrogen uptake during early reproductive growth, and by the R5 stage there were no significant differences in plant nitrogen uptake among the split treatments. This recovery demonstrates that maize can compensate for early-season nitrogen deficits when water is available, absorbing substantial nitrogen during grain filling when demand peaks.

The drought year told a different and arguably more important story. Under severe water stress, reducing the nitrogen rate to 50 percent of the full application—120 kilograms per hectare instead of 240—did not reduce biomass or grain yield compared to full rates. Even more striking, the reduced rate improved nitrogen uptake efficiency and agronomic efficiency while slashing residual soil nitrate. The 50 percent treatment left 83 percent less residual nitrogen in the soil than the full-rate treatments, and the 25 percent treatment left 58 percent less. In 2022, between 40 and 60 percent of applied nitrogen remained as residual soil nitrate after harvest in the full-rate plots, with the highest treatment accumulating nearly 160 kilograms of residual nitrogen per hectare—fertilizer that had been paid for but never used, sitting in the soil as a future source of water contamination.

The physiological measurements revealed why drought changed the calculus. When soil moisture is scarce, plants cannot take up nitrogen regardless of how much is available, so extra fertilizer simply accumulates. Chlorophyll readings, taken with a handheld SPAD meter, declined from the V6 stage onward in all plots in 2022, even as plant nitrogen uptake continued to rise slowly. In 2021, by contrast, plant nitrogen uptake explained 86 percent of the variability in chlorophyll content at the silking stage, but in 2022 that relationship broke down entirely when both growth stages were combined. The study also found that cutting nitrogen to 25 percent of the full rate significantly reduced plant height, leaf area, chlorophyll content, and grain yield—establishing that there is a floor below which nitrogen reduction becomes genuinely yield-limiting, even under stress.

The efficiency metrics told a nuanced story. In the drought year, nitrogen uptake efficiency and agronomic efficiency were highest in the reduced-rate treatments, meaning each kilogram of fertilizer produced more grain and was recovered more completely by the crop. However, nitrogen utilization efficiency—grain yield per unit of nitrogen in the plant—declined at the 25 percent rate, showing that the crop could not convert such scarce nitrogen into grain effectively. Delaying 50 percent of nitrogen until R1 produced the highest nitrogen utilization efficiency in 2022, hinting that well-timed late applications help the crop convert absorbed nitrogen into yield even under stress.

The practical implications are significant for dryland maize systems facing increasingly erratic weather. The findings suggest a strategy of applying 50 percent of nitrogen at planting and holding the remainder in reserve, then deciding based on seasonal conditions whether to apply it. In a wet year, the second application supports rapid uptake during the nitrogen-hungry reproductive stages. In a drought year, farmers can skip or reduce the second application, saving money on fertilizer and fuel while avoiding the accumulation of unused nitrate that could leach into groundwater or run off into the Gulf of Mexico. With nitrogen fertilizer prices having risen sharply in recent years, the economic argument aligns neatly with the environmental one.

The authors caution that the differences in seasonal precipitation limited their ability to test nitrogen strategies under ideal conditions in both years, and they recommend further validation in non-drought seasons. Still, the core message is clear and actionable: adaptive nitrogen management—splitting applications, monitoring the weather, and adjusting rates in real time—can maintain productivity, enhance nitrogen use efficiency, and reduce environmental risk in dryland maize production. In an era when agriculture must feed a growing population while shrinking its footprint on water and climate, getting the timing and amount of nitrogen right may be one of the most powerful levers available.

Subject of Research: Effects of split and reduced nitrogen fertilizer application on maize growth, yield, and nitrogen use efficiency under variable rainfall conditions

Article Title: Split nitrogen application to soil improves maize agronomic performance and nitrogen use efficiency

Article References: Bhattarai, K., Salehin, S. M. U., Rajan, N., & Schnell, R. W. (2026). Split nitrogen application to soil improves maize agronomic performance and nitrogen use efficiency. Discover Soil, 3(1), Article 144. https://doi.org/10.1007/s44378-026-00295-w

Image Credits: AI Generated

DOI: 10.1007/s44378-026-00295-w

Keywords: maize, nitrogen fertilizer, split application, nitrogen use efficiency, drought, dryland agriculture, residual soil nitrate, grain yield, chlorophyll content, leaf area index, agronomic efficiency, sustainable farming