Heat, Dust, and Human Error: Reconstructing Rooftop Photovoltaic Degradation in Semi-Arid Cities as a Coupled Field Problem

Abstract

Rooftop photovoltaic performance in semi-arid cities is conventionally predicted by treating intrinsic degradation, soiling, and maintenance as separable problems, each studied on its own bench and largely outside the urban rooftop context. This article argues that such fragmentation systematically mis-predicts real rooftop performance. Drawing exclusively on a secondary analysis of values reported across the retrieved field literature — and presenting no new measurements — it defends three linked claims. First, heat and soiling are coupled rather than additive: deposited dust raises module temperature by roughly 1.5 °C, and degradation rates roughly double per 10 °C, so soiling enters the power balance twice, optically and thermally, a structure made explicit in an author-derived power equation. Second, the semi-arid city rooftop is a distinct regime, since mixed adhesive aerosols, building-governed worst-case tilt, hotter flush-mounted baselines, and unreliable rain that can cement rather than clean all bias transferred desert-plant curves in the same optimistic — and financially dangerous — direction. Third, and most consequentially, the human maintenance regime is the dominant modifiable variable: the gap between an optimal cleaning cadence and a neglected one exceeds the spread of any module or climate lever the technical literature works to optimize, while mitigation technologies relocate rather than remove the behavioral variable. The article is candid that its most pointed inferences are the author’s synthesis rather than measured findings, and that no retrieved study measures cleaning behavior on rooftops — itself the identified gap. It therefore closes with a fully specified, proposed field protocol that instruments all three pathways on the same devices and stratifies rooftops by governance class to test three pre-registered hypotheses.

Introduction

Photovoltaic (PV) modules are sold on datasheet efficiency and warrantied on a promise of slow, orderly decline. The reality on a rooftop in a hot, dusty city is neither slow nor orderly. Two decades of field synthesis have taught the community what to expect on average: the analytical review of reported degradation rates assembled by Jordan and Kurtz established the first defensible baseline for how fast fielded modules lose power [1], and the larger compendium that followed refined it into the reference distribution against which later work is benchmarked [2]. The most recent field compilation updates that picture and, tellingly, revises it upward — a global median degradation rate of 1.00%/year and a mean of 1.27%/year, both higher than the values reported in the 2016 and 2013 syntheses, with strong variation by technology, climate, and geography [3]. Climate is not a minor covariate in this story. A weighted meta-regression across the degradation literature finds that modules exposed in desert conditions degrade, on average, 0.642 percentage points per year faster than otherwise comparable modules [4]. For a nominal 1%/year baseline, that is a difference of roughly two-thirds, compounded over a 25-year design life.

Soiling adds a second, faster loss channel on top of degradation. Field measurements in Qatar show that, without cleaning, soiling reduced output power by 43% after six months at an average ambient dust density of 0.7 mg/m³ [5]. Modeled soiling losses reach as high as 35% across northern Africa and the Middle East where particulate loading is high and rain is scarce [6], and spatial field campaigns confirm the chemistry and heterogeneity of the deposited dust down to the level of individual cities [7]. These are not marginal corrections. They are first-order determinants of whether a rooftop array meets its financial case.

Yet the evidentiary base has an uneven shape. The overwhelming majority of soiling research has been conducted in open semi-arid or desert terrain, at ground-mounted test benches or utility plants; comprehensive measurement of soiling loss in dense urban environments is, as one of the few dedicated urban studies puts it plainly, rarely reported [8]. This matters because the semi-arid city is not a scaled-down desert plant. It is a place where dust chemistry is mixed with combustion and construction particulates, where shading and airflow are governed by surrounding buildings, where module temperatures are elevated by roof microclimates, and — decisively — where cleaning and maintenance are performed (or neglected) by building owners, tenants, and contractors whose behavior no anti-soiling coating can standardize. The rare field studies that instrument rooftops directly capture exactly this coupling: a recent North Indian rooftop experiment attributed a 9.6% output decline to temperature alone while monthly soiling losses reached 11.7% in the dry season, with monsoon rain partially resetting the system [9].

My argument in this article is that these three loss pathways — heat, dust, and human maintenance behavior — have been studied largely in isolation and predominantly outside the urban rooftop context in which they actually interact, and that this fragmentation systematically mis-predicts real rooftop performance in semi-arid cities. I defend three linked claims. First, that heat and soiling are not additive but coupled: dust changes module temperature and temperature accelerates the degradation that soiling losses are superimposed on. Second, that the urban rooftop is a distinct regime whose loss structure cannot be inferred from desert-plant soiling curves without large error. Third — and this is the claim I take to be most consequential and most neglected — that the human maintenance regime is the dominant modifiable variable in the whole system, so that a field science which treats cleaning frequency as an exogenous input rather than a behavioral object is measuring the wrong thing. I do not present new field data; this pipeline collected none. Instead I offer a secondary analysis of the values reported across the retrieved literature and a fully specified field protocol designed to close the gap those values reveal.

The article proceeds as follows. I first build a conceptual frame that separates the three pathways and specifies how they couple. I then examine what the field literature genuinely establishes about soiling in semi-arid settings, and where its geometry and measurement conventions constrain what it can tell us. Next I develop the thermal–degradation coupling and the specifically urban rooftop regime. I then take up maintenance behavior and cleaning economics as the decisive lever. Finally I set out a research design that would let the coupled problem be measured honestly, and I close by confronting the strongest counterarguments to my thesis and the limits of the evidence I rely on.

Three loss pathways and why they must be modeled together

It is analytically convenient to decompose the gap between a module’s datasheet power and its delivered power into separable terms. Convenience, however, has quietly shaped the field into three research communities that rarely meet on the same roof. It is worth stating the decomposition precisely before arguing that it is misleading when the terms are estimated independently.

Definitions and the standard additive picture

The first pathway is intrinsic degradation: the slow, largely irreversible loss of rated capacity through material and interface aging. Field syntheses express it as an annual percentage rate, with the current global median at 1.00%/year and mean at 1.27%/year [3], modulated upward in hot, arid climates [4]. The second is soiling: the reversible attenuation of incident light by deposited dust, expressed either as an instantaneous soiling ratio or as an energy loss over an exposure interval — up to 43% over six months in the Qatari field case [5]. The third, which I will argue has been under-theorized, is the maintenance regime: the schedule, method, and reliability with which human agents remove soiling and respond to faults. In most models this appears only implicitly, as a cleaning frequency that resets the soiling term.

The standard picture treats delivered energy as the datasheet energy discounted multiplicatively by a degradation factor and a soiling factor, with maintenance entering only as the reset cadence of the latter. That picture is not wrong so much as incomplete, because it assumes the three terms are independent. They are not.

Where the couplings live

The couplings are physical and behavioral, and the digest’s sources document both. Physically, deposited dust alters the module’s thermal balance. A year-long Moroccan study found a dust-induced average temperature rise of 1.5 °C for poly-silicon modules and 1.3 °C for CdTe, the difference attributed to uneven dust distribution having less thermal influence on the thin-film device [10]. Temperature, in turn, drives degradation: field and materials work reports that the degradation rate roughly doubles for every 10 °C rise in operating temperature, and that cells operating above 25 °C lose about 0.4% to 0.65% of instantaneous output per additional degree [11]. So soiling is not merely an optical loss sitting on top of an independent aging process; by raising operating temperature it feeds the aging process itself. The additive decomposition hides this feedback.

Behaviorally, the maintenance term determines the time-integral of both other terms. A soiling loss of 11.7% in a single dry month [9] is catastrophic if cleaning is annual and trivial if cleaning is weekly; the same physical dust deposition produces radically different energy outcomes depending on a human schedule. And because elevated soiling raises temperature, an infrequent cleaning regime does not only cost the optical losses it fails to remove — it also lengthens the exposure to the thermal acceleration of permanent degradation. The maintenance regime therefore couples the reversible loss to the irreversible one.

The practical consequence, which I develop through the rest of the article, is that estimating any one pathway well while treating the other two as fixed parameters will produce biased predictions on a real roof. This is the methodological core of my thesis and the reason a field science organized around three separate benches has struggled to predict urban rooftop performance.

What the semi-arid soiling literature actually establishes

Before criticizing the field’s geometry, it is only fair to credit what it has firmly established. The semi-arid and desert soiling literature is quantitatively strong, internally consistent on its main effects, and — within its own siting assumptions — highly actionable.

Magnitudes and rates

The headline result is that unmanaged soiling in arid climates is large and fast. A detailed analysis of soiling and temperature reports that dust deposition can reduce annual plant performance by 10–30%, with losses exceeding 30% documented in desert environments [10]. Consistent with this, industry modeling places worst-case uncleaned soiling losses as high as 35% across northern Africa and the Middle East, and cites capacity-factor reductions above 30% in Middle Eastern and Saharan regions despite abundant irradiance [6]. The Qatari field campaign gives the sharpest single figure: a 43% power reduction after six months without cleaning at 0.7 mg/m³ ambient dust, and, for a utility-scale plant under typical local dust densities, roughly a 10% decrease over six months translating to about 11,000 QAR/hour of lost revenue [5]. The economic weight of these losses at scale is spelled out in the electrostatic-cleaning work of Panat and Varanasi, which estimates that even a 1% power reduction on a 150-MW installation implies roughly a $200,000 annual revenue loss, and that a global 3–4% soiling loss corresponds to between $3.3 billion and $5.5 billion [13].

A second robust finding is the front-loaded character of the loss. The same electrostatic-cleaning study reports that the dropoff in output happens steeply at the very beginning of dust accumulation and can reach a 30% reduction after just one month [13]. This is important for the maintenance argument later: because most of the harm accrues early, cleaning cadence dominates the outcome and there is little benefit in letting dust “build to a threshold.”

Geometry: tilt, particle size, and deposition

The literature is unusually clear on tilt dependence, which is a design lever available even on constrained rooftops. A semi-arid case study found that soiling loss decreases as tilt angle increases, with maximum deposition rates of 14.28%, 13.53%, 6.79%, and 9.78% at 25°, 40°, 140°, and 155° respectively — deposition rising as the panel approaches horizontal — and reported deposited particles around 150 µm in diameter across inclinations [14]. The same work notes, following Aissa and colleagues, that at small tilt angles airborne dust accumulates disproportionately at the lower edges of the module in Middle Eastern conditions [14]. The dust-storm response is even more dramatic: a review anchors this on an eight-module semi-arid experiment in which, after a dust storm, daily average energy output fell by 58.2%, 27.8%, 21.7%, and 20.7% at tilt angles of 0°, 15°, 30°, and 45° respectively [15]. The 0° case losing nearly three times as much as the 45° case is a decisive argument against near-horizontal rooftop mounting in dusty cities — and, as I will argue, urban rooftops are precisely where near-horizontal mounting is most common for architectural and structural reasons.

The measurement conventions and their blind spot

What the field establishes, it establishes through a particular measurement culture, and that culture has a known blind spot. Optical-characterization work argues that standard methods either measure the soiling impact directly on a reference device or estimate losses from the broadband transmittance of a soiled glass plate, but in doing so ignore the optical properties, composition, and particle-size distribution of the deposited dust — properties that govern adhesion and therefore matter for the effectiveness of cleaning methods and anti-soiling coatings [19]. This is not a pedantic point. If two cities present the same soiling ratio but different dust mineralogy, the cleaning technology and cadence appropriate to each will differ, and a transmittance-only metric will not tell you so. The Qatari spatial campaign underlines the relevance of composition: field measurements at 15 locations mapped silicon, magnesium, sodium, and chlorine as the dominant elements, and their spatial distribution, using GIS [7]. Chloride and salt-bearing dust behaves differently on adhesion and cementation than inert silica, particularly after the light rainfall events discussed below. The literature that reports large, clean soiling numbers has, in effect, measured the how much far better than the what kind, and the urban rooftop is where the what kind starts to dominate.

The thermal–degradation coupling: why hot dust ages modules faster

The soiling literature summarized above measures a reversible optical loss. The degradation literature measures an irreversible capacity loss. My second claim is that in semi-arid cities these two are joined at the module’s operating temperature, and that the joint effect is under-counted when they are estimated on separate benches.

The temperature sensitivity of output and of aging

Two distinct temperature effects must be kept apart. The first is instantaneous and reversible: hotter cells produce less power now. Desert-adapted module work quantifies this as a loss of about 0.4% to 0.65% of output per degree above 25 °C [11]. The second is cumulative and irreversible: thermal stress drives expansion and contraction of encapsulant and interconnect materials, and the degradation rate roughly doubles for every 10 °C rise in operating temperature [11]. This Arrhenius-like acceleration is the mechanism that turns the desert climate penalty of +0.642 pp/year [4] from a statistical association into a physical expectation. The North Indian rooftop study makes the instantaneous effect concrete: cell temperatures peaking at 64.0 °C produced an average daily efficiency reduction of 12.0%, of which 9.6% of output decline was attributed directly to temperature, alongside 0–10.2% from soiling [9].

To make the coupling explicit as the author’s own derivation — not as a fitted result of any study — consider the instantaneous power of a soiled, hot module relative to its clean, reference-temperature power. Writing the soiling transmission factor as , the temperature coefficient as (with in the range 0.004–0.0065 per °C reported in [11]), and the dust-induced temperature rise as (about 1.5 °C for poly-Si in [10]):

(1)

where . The point of writing (1) is not to claim a new measurement but to show structurally that soiling enters twice: once directly through , and once indirectly by raising through . An additive accounting that estimates the soiling loss on a temperature-controlled bench and the temperature loss on a clean module will miss the cross term, and it will do so in the direction that flatters the array’s projected performance.

Long-run field degradation in hot climates

The cumulative side of the coupling is visible in long-term field studies. A study of 24 mono-crystalline modules on an Egyptian rooftop after roughly 25 years of outdoor operation reports a mean maximum-power loss of 23.3%, with short-circuit current and maximum current down 12.16% and 7.2%, open-circuit voltage, maximum voltage, and fill factor down 2.28%, 12.16%, and 15.3%, and an overall system performance ratio of 85.9% [18]. The same work reports an average annual power degradation rate near 1.55%/year for mono-crystalline modules after 11 years and 1.28%/year for multi-crystalline modules after 12 years [18] — both above the current global median of 1.00%/year [3], consistent with the desert penalty [4]. At the more severe end, an in-situ aging assessment of five PV systems in desert conditions found degradation rates up to 2.7%/year, with visual inspection revealing snail trails, delamination, and discoloration, and infrared and electroluminescence imaging identifying hot spots, potential-induced degradation on p-type crystalline modules, and micro-cracks [17].

Two of those failure modes are directly implicated in the coupling I am arguing for. Hot spots — localized overheating, often at partially shaded or soiled cells — are precisely the mechanism by which uneven dust deposition converts into accelerated local aging; the digest’s optical work reminds us that dust is not deposited uniformly [19] and the Moroccan thermal study attributes CdTe’s smaller temperature rise to uneven dust distribution [10]. Delamination and discoloration are thermally driven encapsulant failures whose rate is governed by the same operating-temperature history that soiling elevates. The field record, in other words, does not merely show that hot climates degrade modules faster; it shows the specific signatures one would expect if soiling-driven heating were feeding the degradation.

The urban rooftop as a distinct regime

My third claim concerns geography of a finer grain than “arid versus temperate.” The semi-arid city rooftop differs from the open desert plant in ways that make direct transfer of plant-derived soiling and degradation curves unreliable. The urban soiling study that motivates this section states the gap without ambiguity: most existing soiling research focuses on semi-arid or arid areas, and comprehensive measurement of soiling loss in urban environments — especially high-density cities — is rarely reported; the study accordingly introduces a field method using PV and glass scale models, combining weekly field exposure with in-lab measurement, precisely because the complex urban aerosol and airflow environment is not captured by existing benches [8]. I take this as the empirical charter for treating the urban rooftop as its own regime, and I identify four structural differences.

Aerosol composition and loading are mixed, not mineral

Desert-plant soiling is dominated by wind-blown mineral dust of the kind mapped in Qatar — silicon, magnesium, sodium, chlorine [7]. Urban rooftop soiling superimposes on that background a combustion and construction aerosol: soot, hygroscopic sulfates and nitrates, and cementitious particulates. Because adhesion and cementation depend on composition and particle size, not merely on mass loading [19], the cleaning cadence and method calibrated on mineral desert dust may under-serve an urban deposit that includes sticky, hygroscopic, or oily fractions. The digest does not provide a decomposed urban dust chemistry, and I mark this as a reasoned inference from the composition-sensitivity established in [19] and the urban measurement gap identified in [8] rather than as a measured result.

Airflow, shading, and mounting geometry are building-governed

On a utility plant, tilt is chosen to optimize yield and can be set steep enough to shed dust, exploiting the strong tilt dependence documented in the semi-arid literature — recall the 58.2% loss at 0° against 20.7% at 45° after a dust storm [15], and the monotonic rise in deposition toward horizontal [14]. Urban rooftops routinely violate this optimum: parapets, plant rooms, and architectural constraints push arrays toward low tilt and toward flush or near-horizontal mounting, and surrounding buildings create both intermittent shading and turbulent, sheltered pockets where dust settles and does not blow off. The consequence, which I advance as the author’s synthesis of [14] and [15], is that the urban rooftop tends to sit at the worst end of the tilt–soiling relationship the desert literature has characterized, so plant-derived soiling rates measured at favorable tilt will understate urban rooftop losses.

Roof microclimate raises the temperature baseline

The thermal coupling of the previous section is intensified on a roof. Flush-mounted or low-standoff arrays over a hot membrane experience reduced rear ventilation and a higher baseline before any dust-induced is added. The rooftop study that reported a 9.6% temperature-attributable output decline and 64.0 °C peak cell temperature is itself a rooftop composite-climate measurement [9], not a ground bench, and its magnitudes are a warning that rooftop thermal losses are not the same as open-rack ones. Combining this with the degradation-doubling-per-10 °C relationship [11], the urban roof plausibly experiences both larger instantaneous losses and faster aging than an open plant at the same latitude — again, a coupling that separate benches miss.

Natural cleaning is unreliable and can backfire

Utility plants in truly arid zones expect little rain and plan for it; the urban semi-arid rooftop is often in a climate with occasional rainfall that owners assume will clean the array. The review literature is blunt that this assumption is dangerous: rainfall is the most effective natural cleaning mechanism, but in arid and semi-arid zones it is irregular and low, occasional desert rainfall cannot clean modules, and light rainfall can even form slick mud puddles on the surface [20]. On a low-tilt urban roof — where water pools rather than runs off — light rain onto a mixed hygroscopic deposit is a plausible recipe for cementation rather than cleaning. The North Indian rooftop record shows the benign side of this, with monsoon rain lowering soiling losses [9]; the review shows the malign side [20]. Which one obtains depends on tilt, dust chemistry, and rain intensity — exactly the variables that differ between plant and city.

Taken together, these four differences do not merely add uncertainty; they bias the transfer in a consistent direction. Mixed adhesive dust, worst-case tilt, a hotter baseline, and unreliable rain all push urban rooftop losses above what plant-derived curves predict. If that directional bias is real, then the current practice of sizing and warranting urban rooftop projects on desert-plant or lab soiling data is not conservatively wrong but optimistically wrong — the more dangerous failure for an investor and a grid planner.

The maintenance regime: the neglected and decisive variable

I have left the argument I consider most important for last in the analytical development, because it depends on the preceding ones. If soiling is front-loaded [13], if it feeds temperature-driven aging [10], [11], and if urban rooftops sit at the unfavorable end of every physical lever [8], [14], [15], [20], then the single variable that most determines lifetime energy yield is not the module, the coating, or even the climate — it is the human cleaning and maintenance regime. And that variable is precisely the one the field has treated as an exogenous knob rather than an object of study.

Cleaning cadence dominates the economics

The techno-economic evidence is unusually clean on this point. A year-long Moroccan study using TraCS and DustIQ soiling sensors found that, with monthly cleaning, total annual energy loss was 1.95% for PV versus 17.76% for CSP, with an average daily optical soiling loss of 0.24%/day for PV against 1.21%/day for CSP, and — the operative result — a three-week cleaning frequency delivered the highest profitability for both technologies [16]. The lesson is not the specific interval, which is site-specific, but that an optimum exists and that it is set by trading cleaning cost against the front-loaded loss curve. Clean too rarely and the integrated soiling loss (and its thermal knock-on) dominates; clean too often and labor and water costs dominate. The MDPI soiling review states the governing principle directly: the soiling effect is site-specific and requires an optimal cleaning procedure tailored to the local climate while considering cost-effectiveness [20].

Why cadence is a behavioral, not a physical, quantity

Here is the crux of my claim. The optimum cleaning interval is a physical and economic quantity, but the realized interval on an urban rooftop is a behavioral one. A utility plant has an O&M contract, monitoring, and staff whose job is to hit the optimum. A building rooftop is cleaned — if at all — by an owner who may not monitor output, a tenant with no stake in generation, or a contractor whose visit frequency is set by cost and access, not by the soiling curve. The 43%-in-six-months Qatari figure [5] and the 11.7%-per-month dry-season figure [9] describe what happens when the human reset does not arrive. The gap between the optimal three-week cadence [16] and a realized annual or never cadence is, in energy terms, the difference between a 2% and a 30%+ soiling loss — a factor exceeding the entire spread of module efficiencies or degradation rates the technical literature works so hard to optimize. This is why I call maintenance the decisive lever: no plausible improvement in coatings or cell technology can recover the yield lost to a neglected cleaning schedule.

The digest does not contain a study that measures maintenance behavior — schedules, decision rules, monitoring adoption — on urban rooftops. That absence is itself the gap. The field has excellent physics of soiling and degradation and excellent economics of optimal cleaning, but almost no empirical characterization of the human process that determines whether the optimum is ever approached. “Human error” in my title is not rhetorical: it names the difference between the engineered optimum and the behavioral realization, and it is currently unmeasured.

Mitigation technologies do not remove the human variable

One might hope that automated or passive mitigation would render the human regime moot. The mitigation literature suggests otherwise, in an instructive way. A comparative study of self-cleaning mechanisms on roof-mounted panels found that as dust density rose from 7.5 to 18.15 g/m², power output fell about 23% for a self-cleaning wiper, 33% for a nanocoated panel, and 37% for an uncoated reference [21] — with the useful diagnostic detail that open-circuit voltage was unaffected while short-circuit current fell, confirming the loss is optical shading rather than an electrical fault [21]. Read carefully, this is a partial-mitigation result: the best self-cleaning approach still lost 23%, cutting the reference loss by less than half. Even the electrostatic, waterless removal approach of Panat and Varanasi is motivated by the water cost of manual cleaning in arid regions [13], and its deployment is itself a maintenance decision. Nanocoatings degrade, wipers fail, electrostatic systems need power and upkeep — each mitigation substitutes one maintenance task for another rather than eliminating the human regime. The behavioral variable is not designed away; it is relocated.

There is also a water dimension that ties maintenance to the resource base of semi-arid cities. Frequent wet cleaning at the optimal three-week cadence [16] consumes water in exactly the places where water is scarce, which is one reason the waterless electrostatic approach is being pursued [13]. The maintenance optimum for energy is therefore not the same as the maintenance optimum for water, and in a semi-arid city the two resource systems are coupled through the cleaning schedule — a further reason the human regime deserves study as a system variable rather than a fixed input.

A field research design for the coupled problem

The preceding sections indict the fragmentation of the evidence base. Honesty requires that I specify how the coupled problem could actually be measured, since this pipeline executed no measurement. What follows is a research design — a protocol that could be executed — built to capture heat, dust, and maintenance behavior together on real urban rooftops, and to test the directional bias hypothesized in Figure 2. I present it as a design, with each element justified by the sources that reveal its necessity.

Objectives and hypotheses

The design would test three pre-registered hypotheses, each following directly from the analysis above:

• H1 (regime bias): Under matched cleaning cadence, urban low-tilt rooftop arrays exhibit higher integrated soiling loss than open-rack references at optimized tilt in the same city — testing the directional bias built from [8], [14], [15], [20].
• H2 (thermal coupling): Soiling-induced temperature rise contributes a measurable increment to both instantaneous loss and multi-year degradation beyond the additive soiling-plus-temperature prediction — testing the cross term in equation (1), grounded in [9], [10], [11].
• H3 (behavioral dominance): Across a portfolio of rooftops, variation in realized maintenance cadence explains more of the variance in delivered annual yield than variation in module technology, tilt, or intrinsic degradation rate — the central claim, for which the digest supplies motivating magnitudes ([5], [9], [16]) but no direct test.

Instrumentation and measurement matrix

The design’s distinctive feature is that it instruments all three pathways simultaneously on the same devices, rather than importing two of them from separate benches. Table 2 sets out the measurement matrix. Optical soiling would be characterized not by transmittance alone but with dust sampling for composition and particle-size distribution, following the explicit critique in [19] and the compositional mapping precedent in [7]. Module temperature would be logged at the cell and rear-surface level to resolve as in [10]. Degradation would be tracked by periodic I–V characterization plus electroluminescence and infrared imaging to catch the hot-spot, PID, delamination, and micro-crack modes documented in desert aging [17]. Maintenance behavior would be recorded as an observed variable — actual cleaning dates, methods, water use, and monitoring practice — the element absent from the entire retrieved literature.

Site sampling and the behavioral stratification

The sampling frame is what makes this an urban rooftop study rather than another plant study. The design would stratify rooftops across a single semi-arid city by (i) tilt class, capturing the near-horizontal-to-steep range that determines deposition [14], [15]; (ii) mounting standoff, capturing the roof-microclimate thermal effect [9]; and, crucially, (iii) governance class — owner-occupied, tenanted, and professionally managed — as a proxy for the maintenance-behavior regime that H3 targets. Co-located open-rack reference modules at optimized tilt in the same city would anchor H1. The scale-model field method validated for complex urban environments in [8] offers a low-cost way to expand spatial coverage of the soiling term across many rooftops without a full array on each.

Analysis plan and honest treatment of confounds

The proposed analysis would decompose delivered energy per site into degradation, soiling, and temperature terms using equation (1) as the structural model, then test H2 by asking whether an additive fit leaves a systematic residual correlated with . H3 would be tested with a variance decomposition across the site portfolio, with realized cleaning cadence entered as the behavioral regressor. The known confounds must be stated: dust chemistry varies within a city [7]; rain events can either clean or cement depending on tilt and composition [20], so precipitation must be logged and interacted with tilt rather than treated as a uniform reset; and the meta-analytic desert penalty [4] should inform priors on the degradation term so that a short study is not over-interpreted. The design borrows the methodological caution of the degradation meta-analysis — study-level weighting and clustered standard errors [4] — for any pooling across sites or seasons. I emphasize that no such fit has been performed here; the plan specifies what would be estimated and how, not what was found.

Counterarguments, competing interpretations, and the limits of the evidence

A thesis this pointed invites strong objections. I take the three most serious in their strongest form before responding, and then state plainly what my evidence base cannot support.

Objection 1: The physics is already well characterized; only the parameters change

The strongest version of this objection is that soiling optics, temperature coefficients, and degradation kinetics are universal physics, so a well-parameterized plant or lab model transfers to any site once local dust loading and irradiance are supplied — the urban rooftop is just another parameter set, not a new regime. This is a serious position, and the internal consistency of the semi-arid literature — tilt dependence [14], [15], front-loaded accumulation [13], temperature sensitivity [11] — lends it weight.

My response is that the objection concedes the mechanism but underestimates the interaction terms and the behavioral variable. Equation (1) shows that the soiling and temperature terms are not independent parameters but coupled ones through [10], so “supplying local parameters” to an additive model is structurally biased, not merely imprecise. More decisively, the parameter that dominates lifetime yield — realized cleaning cadence — is not a physical parameter at all but a behavioral outcome, and the dedicated urban study exists precisely because existing methods do not capture the urban environment’s soiling behavior [8]. Universal physics with mis-specified coupling and an unmeasured behavioral driver does not yield reliable urban rooftop predictions.

Objection 2: Soiling is reversible, so it cannot matter for degradation warranties

A second objection separates the concerns: soiling is a reversible operational loss that cleaning fully recovers, while degradation is the irreversible quantity that warranties and lifetime models care about; conflating them, the objection runs, confuses O&M with reliability. The self-cleaning study’s finding that open-circuit voltage is unaffected while short-circuit current falls under dust [21] supports the reversibility of the optical loss.

I grant the reversibility of the optical loss but deny the separation. The thermal coupling means that time spent soiled is time spent hotter, and the digest’s own numbers — degradation roughly doubling per 10 °C [11], dust-induced rises of ~1.5 °C [10], and hot-spot and delamination modes in desert aging [17] — indicate that the reversible loss leaves an irreversible thermal fingerprint. A module that is chronically soiled because of a lax maintenance regime is not merely losing recoverable energy; it is accumulating irreversible degradation faster. The Egyptian 25-year field record of 23.3% Pmax loss and 1.55%/year rates [18], sitting above the global median [3] in a hot climate consistent with the desert penalty [4], is what this looks like at end of life. Reversible and irreversible are coupled through operating temperature, so the O&M/reliability firewall the objection relies on does not hold in hot climates.

Objection 3: The maintenance argument is untestable hand-waving

The sharpest objection to my central claim is that “human error” is not a measurable engineering variable and that elevating it above module technology is rhetoric dressed as analysis. This deserves a direct answer because, on the current evidence base, it has a point: the digest contains no study that measures maintenance behavior on rooftops, and H3 is motivated rather than demonstrated.

My response is twofold. First, the maintenance variable is measurable — cleaning dates, methods, water use, and monitoring adoption are observable facts, and Table 2 specifies how a study would record them; the techno-economic literature already treats cleaning frequency as a quantitative decision variable with an optimum [16], so the object exists and merely awaits behavioral measurement. Second, the magnitude argument is not hand-waving but arithmetic on cited values: the spread between the ~2% annual soiling loss at optimal cadence [16] and the 30–43% losses under neglect [5], [10] exceeds the spread of any other lever in the system. That the dominant lever has not yet been measured behaviorally is an indictment of the field, not of the claim. I present H3 honestly as a hypothesis my proposed design would test, not as an established result.

Limits of this article’s evidence base

Candor about the evidence is part of the argument. Several limits constrain what I can claim. This article reports no new measurements; every numeric value is a result reported by a cited source, and the coupled decomposition in equation (1) is my own structural reasoning, not a fitted model. The digest itself carries an integrity caveat: for a number of the field studies I rely on (notably [9], [16]–[19], [21]), the retrieved records exposed findings, venues, years, and locators but not the full author lists, which the source record flags for verification before final citation; I have cited them as recorded and marked no author I could not attribute. The canonical degradation syntheses [1], [2] are cited for their role as baselines, and their exact median figures were not in the retrieved record — I have therefore leaned on the more recent compilation’s explicit numbers [3] for quantitative claims about degradation rates. Some framing figures come from industry tooling and trade press [6], [7], which I have used only for spatial context and confirmed against primary field studies where a quantitative claim is made. Finally, several of my most pointed inferences — the mixed urban dust chemistry, the “worst-tilt” position of urban rooftops, and the behavioral dominance hypothesis — are explicitly the author’s synthesis of the cited physics and economics, not measured findings, and I have marked them as such wherever they appear. The correct reading of this article is as a structured argument and an executable design grounded in the reported literature, not as a report of executed field work.

Conclusion

This article set out to argue that rooftop photovoltaic performance in semi-arid cities cannot be predicted by treating heat, dust, and human maintenance as three separable problems studied on three separate benches. The field literature is quantitatively strong within its own siting assumptions: it establishes that unmanaged soiling in arid climates is large and front-loaded, reaching 43% over six months in Qatar [5] and up to 30% in a single month [13]; that the degradation baseline is rising, with a global median of 1.00%/year [3] and a documented desert penalty of +0.642 percentage points per year [4]; and that cleaning cadence has an economic optimum, with monthly cleaning holding annual PV soiling loss to 1.95% in Morocco [16]. My thesis was not that these findings are wrong but that estimating any one pathway well while holding the other two fixed produces biased predictions on a real roof.

The three linked claims stand on that evidence. Heat and soiling are coupled rather than additive: dust raises module temperature by roughly 1.5 °C [10], and the degradation rate roughly doubles per 10 °C [11], so soiling enters the power balance twice, as I made explicit in equation (1) — once optically and once thermally. The urban rooftop is a distinct regime because mixed adhesive aerosols, building-governed worst-case tilt [14], [15], hotter flush-mounted baselines [9], and unreliable rain that can cement rather than clean [20] all bias transferred plant curves in the same optimistic direction, precisely the direction most dangerous to investors and grid planners. And the maintenance regime is the dominant modifiable variable: the gap between an optimal three-week cadence [16] and a neglected annual or never cadence [5], [9] exceeds the entire spread of module efficiencies and degradation rates the technical literature works to optimize, while mitigation technologies relocate the human variable rather than removing it [13], [21].

I have been candid throughout about what this evidence base cannot do. This article reports no new measurements; every numeric value is a result reported by a cited source, and the coupled decomposition in equation (1) is my own structural reasoning, not a fitted model. The most pointed inferences — mixed urban dust chemistry, the worst-tilt position of urban rooftops, and the behavioral-dominance hypothesis H3 — are explicitly the author’s synthesis rather than measured findings. Most tellingly, the retrieved literature contains no study that measures maintenance behavior on rooftops, and that absence is itself the gap the argument identifies. I have therefore presented the field research design as a proposed protocol, not as executed work: its measurement matrix, behavioral stratification, and three pre-registered hypotheses specify what would be estimated and how, not what was found.

The open problems follow directly. First, the field needs an instrumented urban rooftop study that logs all three pathways on the same devices, pairing soiling composition and particle-size measurement [7], [19] with cell-level thermal logging [10] and defect imaging [17], so that the cross term in equation (1) can be tested against an additive null. Second, and most neglected, it needs an empirical characterization of realized cleaning cadence as a behavioral object stratified by governance class, so that H3 can be moved from motivated hypothesis to measured result. Third, the coupling between the energy-optimal and water-optimal cleaning schedules in water-scarce cities deserves study as a system problem [13], [16]. Until the human maintenance regime is measured rather than assumed exogenous, a field science organized around three separate benches will keep measuring the wrong thing — and sizing urban rooftop projects on curves that flatter their real performance.

References